Alexander Nix: From Mad Men to Math Men | OMR Festival 2017 - Hamburg, Germany | #OMR17

OMR

AI-assisted summary of Alexander Nix: From Mad Men to Math Men | OMR Festival 2017 - Hamburg, Germany | #OMR17 — original source. Check the original for full context.

Source published: 2017-03-10

Key ideas

Cambridge Analytica CEO Alexander Nix outlines the industry shift from intuition-driven 'Mad Men' to algorithmic 'Math Men', anchored by behavioral science (the OCEAN model), big data analytics, and addressable ad tech. (0:00)
During Donald Trump's 2016 campaign, Cambridge Analytica ran 300,000 to 400,000 monthly polls nationwide to power predictive modeling and directed more than $100 million in digital advertising. (12:16)
An independent Google brand lift study of targeted campaign videos demonstrated an 18% lift in search volume, an 11% increase in personal favorability, and an 8% increase in voter intent for Donald Trump. (17:05)
Nix claims nearly two million individuals voluntarily took personality quizzes to build the training datasets used to predict psychological traits across wider electorate databases. (23:51)

Detailed notes

## Background Published on YouTube by OMR on March 10, 2017, this video captures Cambridge Analytica CEO Alexander Nix delivering a keynote at the OMR Festival in Hamburg, Germany. The presentation explores the shift from traditional advertising intuition to algorithmic data analytics, detailing the behavioral methodologies and big data strategies Cambridge Analytica utilized during Donald Trump's 2016 presidential campaign.

## TL;DR Alexander Nix presents Cambridge Analytica's methodology for modern communications, built on the three pillars of behavioral science (the OCEAN personality framework), big data analytics, and addressable advertising technology. He details the firm's real-world deployment on Donald Trump’s 2016 presidential campaign, highlighting how intensive weekly polling, voter segmentation dashboards, and over $100 million in targeted digital media drove fundraising, rally turnout, and voter persuasion in key swing states like Wisconsin. While Nix insists that data technology is not a panacea capable of manufacturing victory for a bad candidate, he frames it as an indispensable competitive edge. The session culminates in an unsparing Q&A addressing European privacy regulations, media backlash, and a heated audience confrontation challenging the corporate ethics of electing Donald Trump.

## Atmosphere & style The keynote begins as a slick, hyper-articulate corporate masterclass, with Nix projecting disciplined technocratic calm as he systematically deconstructs legacy advertising through crisp slides and data charts. However, the emotional arc shifts dramatically during the Q&A, transforming from an industry marketing talk into an intense, politically charged reckoning. When an audience member bluntly challenges Nix over his moral accountability for Trump's presidency, the room crackles with friction, contrasting the audience's raw indignation against Nix's unflappable, strictly professional detachment.

## Bulletin - [00:00](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=0s) Cambridge Analytica CEO Alexander Nix outlines the industry shift from intuition-driven 'Mad Men' to algorithmic 'Math Men', anchored by behavioral science (the OCEAN model), big data analytics, and addressable ad tech. - [12:16](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=736s) During Donald Trump's 2016 campaign, Cambridge Analytica ran 300,000 to 400,000 monthly polls nationwide to power predictive modeling and directed more than $100 million in digital advertising. - [17:05](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1025s) An independent Google brand lift study of targeted campaign videos demonstrated an 18% lift in search volume, an 11% increase in personal favorability, and an 8% increase in voter intent for Donald Trump. - [23:51](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1431s) Nix claims nearly two million individuals voluntarily took personality quizzes to build the training datasets used to predict psychological traits across wider electorate databases.

## Key takeaways - [00:00](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=0s) Modern marketing and political communication are shifting from top-down intuition to three quantitative pillars: behavioral science, big data analytics, and addressable ad tech. - [01:12](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=72s) Legacy audience segmentation based purely on demographics like gender or age is fundamentally flawed because personality is the primary driver of behavior. - [01:12](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=72s) Cambridge Analytica operationalizes behavioral science using the academic OCEAN psychometric framework (Openness, Conscientiousness, Extroversion, Agreeableness, Neuroticism) to customize creative messaging. - [05:26](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=326s) Big data analytics unifies demographic, attitudinal, and behavioral data into a single repository, identifying what an audience needs to hear before creative drafting begins. - [09:00](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=540s) Addressable ad tech renders blanket advertising obsolete by matching individual offline records with set-top box data and hyperlocal cable delivery. - [12:16](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=736s) Cambridge Analytica joined the Trump campaign as an end-to-end technology partner, leveraging 14 months of prior primary research to support a lean 30-person staff. - [13:40](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=820s) Real-time weekly tracking polling allowed campaign managers to allocate resources precisely during crises, such as the hot-mic controversy, and uncover hidden persuadable voters in states like Wisconsin. - [15:23](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=923s) Predictive modeling scored every US adult across donor probability, candidate preference, voter turnout, and issue salience to deliver micro-targeted messaging. - [17:05](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1025s) Early small-dollar fundraising efforts generated roughly $27 million in the campaign's first month and collected approximately 1 million supporter email addresses. - [17:05](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1025s) Nix emphasizes that data science is not a panacea that can make a bad candidate into a good one, but serves as a decisive edge in knife-edge elections. - [19:46](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1186s) Nix rejects ideological pigeonholing, asserting that Cambridge Analytica works on 8 to 10 national campaigns globally each year across both left- and right-of-center parties. - [23:51](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1431s) Psychometric baseline data is acquired through voluntary online personality quizzes, which are then modeled probabilistically against broad consumer and voter databases. - [23:51](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1431s) European opt-in data protection regulations prevent direct commercial record buying as seen in the US, but primary OCEAN research and legal first-party brand databases remain viable. - [26:30](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1590s) When challenged by an audience member over electing Donald Trump, Nix defended the outcome as democratic self-determination and evaluated his firm's role strictly on professional execution. - [27:59](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1679s) Nix maintains that providing data and communication services to political campaigns is an essential civic duty to uphold fair democratic representation.

## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=0s) The Shift from "Mad Men" to "Math Men"

Alexander Nix takes the stage at the OMR Festival in Hamburg to address the fundamental revolution underway across marketing and communications: the shift from the intuition-driven "Mad Men" era of the 1950s to the algorithmic rigor of modern "Math Men." Nix connects this shift directly to the real-world political arena, noting Cambridge Analytica’s controversial and high-profile work on Donald Trump’s 2016 presidential run. He acknowledges Trump’s own public skepticism about tech wizardry, agreeing with the candidate's core assertion about what actually drives an election across the finish line: “Throughout the presidential cycle in 2016 Trump was vocal about the fact that it is candidates and not technologies that win campaigns and in this regard we would absolutely agree with him.”

Even so, Nix argues that political races and commercial brand campaigns are being fundamentally transformed by three interconnected technical methodologies: Behavioral Science, Big Data Analytics, and Addressable Ad Tech. Together, these tools allow campaign operatives and brand marketers to replace crude assumptions with scientific precision, moving away from broad demographics to individual human psychology.

```mermaid flowchart TD accTitle: Cambridge Analytica Three Pillars Methodology accDescr: Diagram showing the three foundational methodologies used to target individuals. subgraph Pillars ["Three Methodological Pillars"] A["Behavioral Science (OCEAN Psychographics)"] B["Big Data Analytics (Unified Database of Record)"] C["Addressable Ad Tech (Hyperlocal & Match Targeting)"] end Pillars --> D["Individualized Micro-Targeted Messaging"] ```

---

[01:12](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=72s) Behavioral Science and the OCEAN Personality Model

To illustrate the difference between conventional messaging and behavioral science, Nix presents a thought experiment about keeping swimmers off a private beach. An informational notice simply tells people not to swim, seeking to alter surface-level attitudes. A behavioral sign, however, flashes the warning of shark-infested waters to trigger deep, visceral fear. Nix contends that most contemporary agencies remain hopelessly outdated because they rely exclusively on demographic proxies like age, gender, geography, and media habits. He rejects this legacy model as fundamentally illogical: “When you stop for just a moment and think about this, the idea of segmenting audiences say based on gender is very ridiculous.” Treating every woman, young person, or minority as an undifferentiated voting or purchasing bloc ignores the primary driver of human action: personality.

To capture these deeper drivers, Cambridge Analytica deployed a comprehensive quantitative psychometric instrument grounded in the academic standard of the "Big Five" or OCEAN personality framework: - **O — Openness**: How open an individual is to novel experiences and intellect. - **C — Conscientiousness**: How much someone values order, habit, duty, and meticulous planning. - **E — Extroversion**: How outgoing, socially active, and externally stimulated someone is. - **A — Agreeableness**: Whether one prioritizes community and social harmony over individual needs. - **N — Neuroticism**: A person's baseline level of worry, anxiety, and emotional vulnerability.

To demonstrate how psychographics dictates creative execution, Nix contrasts two hypothetical car buyers. Both share identical demographics: male, similar middle-class incomes, and matching consumer profiles. However, their psychometric scores are polar opposites. The first individual scores high in conscientiousness and agreeableness; he is drawn to environmental impact (such as an electric engine) and rational, fact-dense specifications. The second is an extrovert who cares little for spec sheets, requiring bold imagery of a sports car designed to evoke excitement, social status, and raw sensory experience.

---

[05:26](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=326s) Big Data Analytics and the Unified Record

The second leg of Nix's framework is Big Data Analytics. In the classic agency model of the mid-20th century, creative directors sat in boardrooms and devised top-down concepts, broadcasting them widely in the hope that something would land. Data analytics flips that equation upside down by discovering what an audience needs to hear before copywriters write a single word: “Today we can use big data to understand exactly what messages each specific group within a target audience need to hear way before the creative process is even started.”

Nix defines big data as aggregating every accessible record into a single, centralized database: - **Factual & Demographic Data**: Age, gender, ethnicity, religion, and household structure. - **Attitudinal & Lifestyle Data**: Vehicle ownership, magazine readership, social media engagement, church attendance, and civic or social club memberships. - **Behavioral & Psychographic Data**: Inferred traits that map how individuals perceive and process risk, logic, and emotion.

Using a commercial toothpaste campaign to demonstrate, Nix displays an analytics dashboard mapping consumers along two axes: price sensitivity versus brand preference. Budget-conscious consumers receive discount promotions, whereas feature-sensitive buyers receive performance arguments tailored to stain removal, teeth whitening, or gum health. When combined with psychographics, targeting becomes surgical: an ad aimed at single women aged 25–30 who index high in neuroticism delivers an anxiety-relieving emotional hook on Facebook, warning them to eliminate stains before they ruin their smile.

---

[09:00](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=540s) Addressable Ad Tech and Industry Transformation

The final pillar is addressable ad tech, which Nix describes as the distribution mechanism that renders mass media obsolete. He makes a bold proclamation about the end of common media experiences: “I've said it before and I'll say it again: I'm very much of the belief that blanket advertising is dead.” In the emerging landscape, communication is fully individualized, allowing a husband and wife sitting in the exact same living room to receive distinct advertisements for the very same product based on their individual psychological profiles.

Nix cites advertising researcher Bill Harvey to reinforce the financial imperative: “On average a 10% increase in purchase target density will generate a 7% increase in incremental sales.” Marketers can achieve this by pairing offline data records with online cookies, set-top box cable viewing feeds, and localized cable infrastructure: - **Set-Top Box (STB) Data Matching**: Cable providers capture exact viewership records, enabling advertisers to place inventory strictly in the programs watched by specific psychological segments. - **Hyperlocal Television**: Buying airtime not on national networks or metro-wide regional broadcasts, but carved up by local cable headends down to small clusters of zip codes.

To prove that this revolution is not hypothetical, Nix highlights massive corporate investments: - **Retail & Automotive Giants**: Walmart and Ford have poured hundreds of millions of dollars into internal analytics infrastructure and software acquisitions. - **Agency Pivots**: Ogilvy launched its own Center for Behavioral Science, while Kroger acquired British data science firm dunnhumby. - **Holding Company Moves**: WPP chief Sir Martin Sorrell spent years publicly forecasting the transition from creative intuition to hard quantitative data, mirrored by comparable strategic acquisitions at Havas.

Nix concludes this strategic overview by turning to the presentation's primary case study: how Cambridge Analytica deployed these exact techniques when they joined the Donald Trump presidential campaign in June 2016.

[12:16](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=736s) Joining the Trump Campaign and Establishing End-to-End Infrastructure

When Cambridge Analytica joined Donald Trump's presidential campaign in 2016, the operation was lean to an extreme degree: there were only 30 full-time employees on staff. This stood in stark contrast to Hillary Clinton's campaign, which employed roughly 800 people. Cambridge Analytica presented an attractive proposition because the firm had just spent 14 months working on the primary campaigns of Ben Carson and Ted Cruz, accumulating massive voter data repositories, primary research, and an entire technological infrastructure ready to be handed directly to Trump's team. Consequently, Cambridge Analytica was contracted as a complete, end-to-end technology provider.

Their work began with an aggressive research cadence: running roughly 1,500 polls per week in every key state, translating to between 300,000 and 400,000 polls per month across the country. These surveys fed real-time reporting dashboards designed to direct campaign resource and asset allocation, as well as powering predictive audience models. These models segmented the electorate, which in turn dictated messaging and engagement. Ultimately, Cambridge Analytica placed over $100 million in digital advertising buys for the Trump campaign, alongside tens of millions of dollars in data-driven television spending.

```mermaid flowchart TD accTitle: Cambridge Analytica Campaign Operations Pipeline accDescr: Flow diagram showing how polling data feeds dashboards and predictive models, leading to voter segmentation and large-scale media buying.

subgraph DataCollection["Data Collection"] A["1,500 Weekly Polls per State (300k–400k Monthly)"] end

subgraph Analytics["Analytics & Strategy"] B["Reporting Dashboards (Asset & Resource Allocation)"] C["Predictive Audience Models"] D["Voter Segmentation & Sub-segmentation"] end

subgraph MediaExecution["Media Placement"] E["$100M+ Digital Ad Placements"] F["Tens of Millions in Data-Driven TV"] end

A --> B A --> C C --> D D --> E D --> F ```

- Trump campaign staffing at CA's arrival: 30 full-time employees (versus approximately 800 on Clinton's team). - Prior foundation: 14 months of research and data infrastructure built during the Ben Carson and Ted Cruz primary bids. - Polling throughput: ~1,500 polls per state weekly, totaling 300,000 to 400,000 polls nationwide each month. - Media placement scale: Over $100 million in digital advertising and tens of millions in data-driven TV.

---

[13:40](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=820s) Real-Time Crisis Tracking and Targeting Swing States

Weekly polling supplied campaign managers with rapid, granular tracking of electorate shifts. Alexander Nix illustrates this utility through the fallout of the infamous "hot mic" incident involving Trump's recorded locker-room banter. Conventional wisdom assumed nationwide disaster, but Cambridge Analytica’s real-time polling painted a far more complex picture: while Trump's approval plummeted sharply in several states, his ratings actually ticked upward in others. This rapid feedback allowed campaign managers to allocate resources precisely rather than panicking uniformly.

A second crucial dashboard visualized heat maps of persuadable voters—citizens who remained genuinely undecided and receptive to persuasive messaging. Nix points to Wisconsin as the prime example of this strategy in action. Wisconsin had historically been viewed as a safely Democratic stronghold; in fact, the Clinton campaign never visited the state once throughout the general election campaign. However, Cambridge Analytica's data revealed concentrated pockets of persuadable voters who could be swayed to Trump. Acting on this intelligence, the campaign staged five major rallies in Wisconsin, exposing Trump in person to an estimated 60,000 to 70,000 voters. When the final tallies rolled in, Trump carried Wisconsin by a razor-thin margin of roughly 50,000 votes, demonstrating the impact of data-driven field operations.

---

[15:23](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=923s) Predictive Modeling and Audience Segmentation

To construct an actionable targeting apparatus, Cambridge Analytica deployed predictive models across four primary domains: donor modeling, candidate preference, voter turnout, and individual issue salience. In the initial month, the paramount priority was small-dollar fundraising (solicitations of $5 to $20), raising essential early cash before handing that apparatus over to the Republican National Committee (RNC). From there, the analytics team concentrated on preference models (discovering which candidates resonated with specific demographics), turnout models (determining who was likely to show up at polling places and the potential influence of unregistered citizens), and comprehensive issue modeling.

The team modeled social and political preferences so exhaustively that they assigned predicted issue scores to every single adult in the United States. They organized these insights into a behavioral matrix: - **Core Trump Supporters (Top Right):** Best candidates for financial solicitations and donation requests. - **Get Out the Vote (GOTV):** Committed Republicans who required logistical encouragement to turn out and cast ballots. - **Persuasion Audiences:** Centrist voters requiring subtle issue-based pushes to swing toward the conservative ticket.

Once a target segment was isolated, the firm sub-segmented individuals according to the specific policy issues they prioritized, serving tailored creative tailored to those exact concerns.

---

[17:05](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1025s) Campaign Execution Metrics and the Google Brand Lift Study

The operational implementation of this modeling yielded massive reach across fundraising, rally turnout, digital persuasion, and early voting drives: - **Small-Dollar Fundraising:** Generated nearly $27 million in the campaign's first month and captured approximately 1 million supporter email addresses. - **Event Promotion:** Advertised nearly 100 rallies and prompted at least 35,000 downloads of the official Trump campaign smartphone application. - **Digital Persuasion:** Served 1.4 billion impressions across 4,000 discrete advertising campaigns over four months, producing an estimated net favorability lift of 3% for Trump. - **GOTV & Absentee Ballots:** Delivered approximately 80 million impressions, lifting absentee ballot submissions by roughly 2%.

During the campaign, Google carried out an independent brand lift study to assess Cambridge Analytica's digital video campaigns. Testing a cohort of roughly 150,000 people who received tailored digital video content against an unexposed control group receiving generic messaging, Google measured striking performance gains among targeted viewers: - An 18% lift in campaign-related search volume. - An 11% increase in personal favorability toward Donald Trump. - An 8% bump in stated voter intent for Trump.

Nix concludes his prepared remarks by addressing the broader implications of tech in politics, offering a measured defense: > “Clearly the outcome of the election is well known and a lot has been written about it, but if there's one singular takeaway from this event, that is that data science technology is not a Panacea.”

He cautions that “you cannot make a bad candidate into a good one; you cannot swing an election,” but insists that in modern, knife-edge contests, data-driven methodology gives candidates an indispensable edge.

---

[19:46](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1186s) Q&A: Electoral Attribution, Das Magazin, German Politics, and OCEAN Profiles

Stepping into the Q&A session, the moderator asks Nix whether Donald Trump would have become president without Cambridge Analytica’s involvement. Nix pushes back against sweeping claims, stating: > “We're a long way away from a situation where elections can be won or lost simply on the basis of targeted engagement.”

He stresses that in modern political races decided by razor-thin margins, technology is one of several vital components. The moderator then notes that Cambridge Analytica staff had been mingling at the conference over prior days, and asks if the company works exclusively for right-wing candidates. Nix explains that the firm engages in 8 to 10 national campaigns for presidents and prime ministers worldwide each year, working as frequently for left-of-center parties as right-of-center ones: > “We try and leave our ideological—personal ideologies at the door in order to provide the most professional service.”

The interview pivots to the explosive investigative article published by the Swiss magazine *Das Magazin* (widely shared across Facebook in Germany under the title *Ich habe nur gezeigt, dass es die Bombe gibt* / *The Data That Turned the World Upside Down*). When asked for his perspective on the piece, Nix remarks that while some baseline technological capabilities were accurately described, the journalists drew exaggerated conclusions without proper data, resulting in "enormous falsehoods." He notes that because the firm was focused on finishing up work in the United States, the article created a bigger media sensation in Europe than in the US.

Turning to German federal politics ahead of the September 2017 elections, the moderator asks if Cambridge Analytica is advising German campaigns, joking whether Nix could "make Angela Merkel great again." Nix reveals they are not currently working with any German political parties, citing grueling fatigue after two years of high-pressure work in the United States, the need for staff downtime, and the reality that the compressed timeline before the German election limited what they could realistically deliver.

Finally, an audience member queries the core mechanics of psychographic modeling: how does Cambridge Analytica actually collect people's OCEAN personality profiles (Openness, Conscientiousness, Extraversion, Agreeableness, Neuroticism), and why would ordinary citizens volunteer sensitive psychographic data? Nix clarifies that no citizen is ever forced or obligated to share their personal information. Instead, hundreds of thousands—and ultimately millions—of individuals in the United States voluntarily completed online surveys and questionnaires. Once those baseline OCEAN scores were gathered voluntarily, Cambridge Analytica could map and match them against broad commercial consumer databases, predicting psychological traits across the wider electorate.

[23:51](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1431s) Psychometric Data Collection and European Privacy Laws

The discussion continues with the mechanics of how Cambridge Analytica gathers personality data to train its predictive models. Alexander Nix explains that the company uses data gathered from individuals who have completed personality inventories to create a training dataset, which is then deployed to make probabilistic predictions across broader populations. An interviewer questions the underlying incentive structure, asking why any average user would spend five minutes answering extensive questionnaires. Nix responds that people take social media surveys for a variety of reasons, chief among them being curiosity and personal introspection: “I think some people are fundamentally interested to understand what their personality is and what that means, so self-inquiry is obviously driving that.” He emphasizes the scale of this engagement, noting that just shy of two million people have taken their tests, demonstrating a strong intrinsic motivation.

```mermaid flowchart LR accTitle: First-Party Personality Data Modeling Pipeline accDescr: Flow showing how voluntary survey data is gathered and converted into predictive models. subgraph Collection["Data Acquisition"] A["Online Personality Surveys / Quizzes"] --> B["Voluntary User Participation"] end subgraph Processing["Model Building"] B --> C["OCEAN Psychometric Profiles"] C --> D["Training Dataset"] end subgraph Deployment["Application"] D --> E["Predictive Behavioral Models for Wider Audiences"] end ```

When the interviewer presses whether this operates essentially as a Facebook game where users surrender personal data while playing, Nix pushes back against any sinister framing, insisting: “They're voluntarily giving up their data; they're doing this in full knowledge of what's happening here—this is not anything Machiavellian or untoward.” The conversation shifts to the strict legal boundaries of data protection in Germany and across Europe compared to the United States. Nix contrasts the two regulatory environments, highlighting that the American framework operates largely on an opt-out philosophy, whereas European regimes mandate explicit opt-in consent. While foreign firms cannot simply purchase consumer records off the shelf in Europe as they do in the US, Nix points out that organizations can still legally collect primary research—such as OCEAN psychometric surveys—from willing participants or leverage the extensive first-party customer databases that European commercial enterprises have already legally accumulated.

- Nix reports that nearly two million individuals have completed Cambridge Analytica's personality surveys voluntarily. - The United States relies primarily on an opt-out regulatory framework for acquiring data, while European jurisdictions require strict opt-in compliance. - Modeling in Europe remains viable when built on direct, first-party psychometric research or a brand's internal proprietary customer data.

---

[26:30](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1590s) Confrontation Over Electing Donald Trump

The tone of the session shifts dramatically when an audience member at the bar takes the microphone to challenge Nix directly over the ethical consequences of his political work. With palpable emotion and intermittent applause from the crowd, the questioner asks: “If it's really true that you helped Trump get into office, are you happy putting a misogynist buffoon in the highest, the most powerful job in the world, who's now fucking up our world—are you happy?”

Nix remains composed and rejects the premise of the question, arguing that external observers have no standing to override democratic processes: “I think you need to step backwards a little bit and understand that I don't think it's for people sitting in other countries to cast their opinions on the self-determination of the United States.” He contextualizes the 2016 election cycle by noting that Donald Trump competed against seventeen Republican contenders in the largest primary field in modern political history, won the nomination fairly, and ultimately secured the presidency. When the questioner repeats the inquiry, demanding to know whether Nix feels personal happiness about the result, Nix draws a strict professional boundary: “I'm happy with the job that we did.”

- Donald Trump defeated 16 primary challengers in what Nix highlights as the largest candidate field in presidential primary history. - Nix frames the election outcome as the democratic self-determination and will of the American electorate, which outside observers ought to respect. - Rather than defending the candidate personally, Nix evaluates his work strictly through the lens of professional campaign execution.

---

[27:59](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1679s) Commercial vs. Political Campaigns and Closing Remarks

The questioner follows up by examining the corporate ethics of Cambridge Analytica, asking whether the agency makes any distinction between selling consumer products like toothpaste and managing high-stakes political figures like Donald Trump. Nix acknowledges that electoral work carries unique reputational stakes: “Obviously working on a political campaign by its very nature is divisive, and we have to be very careful that the candidates or parties that we work for don't cause any reputational damage to the company, that aren't going to upset our staff or employees, and that obviously aren't going to upset any of our other clients or customers.” Nevertheless, Nix maintains that so long as elections are democratic and lawful, political consulting is a fundamental civic function: “As long as they're free and fair elections and democracy is being upheld, we think it's the duty of our firm to provide candidates with the best possible service and fair representation—that's the only way you can have fair elections.”

When pressed on whether the firm has ever refused a client, Nix remarks that while hypothetical red lines exist, he cannot recall an instance where they declined a contract. The questioner takes a final jab, asking how the firm could work for someone famous for bragging about grabbing women by the genitals. Nix declines to re-litigate the controversy, stating: “Well, I can't comment on that; I think that was covered in the news at the time, and I think I've heard both sides of the argument.”

The moderator then steps in to bring the keynote session to a close. Thanking Nix for presenting his work transparently at the OMR Festival, the host notes that while audience members will inevitably disagree on the politics and morality of the methods, understanding the real-world application of behavioral data science is essential for industry professionals. Nix receives thanks from the host for contributing substantive insight to the festival's platform before departing the stage.

- Nix defends political consulting as an essential service for democratic representation, provided elections meet free and fair standards. - Nix states he cannot recall any client or candidate Cambridge Analytica has formally turned down. - The OMR organizers conclude by emphasizing that presenting controversial data methods is critical for industry education and transparent technical discourse.

## Notable quotes > [09:00](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=540s) I've said it before and I'll say it again: I'm very much of the belief that blanket advertising is dead. — Alexander Nix

> [19:46](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1186s) We're a long way away from a situation where elections can be won or lost simply on the basis of targeted engagement. — Alexander Nix

> [05:26](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=326s) Today we can use big data to understand exactly what messages each specific group within a target audience need to hear way before the creative process is even started. — Alexander Nix

> [01:12](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=72s) When you stop for just a moment and think about this, the idea of segmenting audiences say based on gender is very ridiculous. — Alexander Nix

## Mentioned Alexander Nix, Cambridge Analytica, Donald Trump, Hillary Clinton, Ted Cruz, Ben Carson, Angela Merkel, Bill Harvey, Sir Martin Sorrell, Google, Facebook, Republican National Committee, Ogilvy, dunnhumby, Kroger, Walmart, Ford, WPP, Havas, Das Magazin, OMR Festival

## Video description From mad men to math men. Cambridge Analytica first gained notoriety for their role in Donald Trump's presidential campaign. In the Online Marketing Rockstars Keynote | OMR17, Cambridge Analytica CEO Alexander Nix discusses big data, the methodologies and strategies used to identify people's behavior and how data analytics factors into political campaigns.

---

#OMR20 is approaching fast! Be part of the OMR Festival 2020, Europe's leading digital marketing event, on May 12 & 13, 2020 in Hamburg, Germany: http://bit.ly/OMRfestival20

---

Subscribe to our OMR YouTube channel: http://bit.ly/OMRyoutube

---

At OMR, we view ourselves as a platform for industry professionals by industry professionals. We bring together players from every corner of the digital marketing stratosphere: marketing makers with global visionaries, head hunters with prime candidates and information seekers with experts. Our range of products includes the annual OMR Festival with 52,000 visitors in 2019, big-name international speakers, stellar networking opportunities and unique industry insights, our German and English-language blogs, our weekly podcasts, monthly reports, jobs portal, and many more. Welcome to the home of Digital Marketing: http://bit.ly/OMRhomeEn

---

OMR Festival: http://bit.ly/OMRfestival20 OMR Daily: http://bit.ly/OMRdailyEn OMR Events: http://bit.ly/OMReventsEn OMR Podcast: http://bit.ly/OMRpodcastEn

---

OMR Facebook: http://bit.ly/OMRfacebookEn OMR LinkedIn: http://bit.ly/OMR-linkedin OMR Instagram: http://bit.ly/OMRinsta OMR Twitter: http://bit.ly/OMRtwitter

[Source](https://www.youtube.com/watch?v=6bG5ps5KdDo)

中文简报

亚历山大·尼克斯：从广告狂人到数学怪才 | 2017年OMR数字营销峰会（德国汉堡）| #OMR17 亚历山大·尼克斯介绍了剑桥分析公司现代传播方法论，该方法立足于三大支柱：行为科学（OCEAN人格模型框架）、大数据分析以及可定向广告技术。他详细阐述了该公司在唐纳德·特朗普2016年总统竞选中的实际应用，重点介绍了密集周常民调、选民细分仪表板以及超过1亿美元的定向数字媒体投放，是如何在威斯康星州等关键摇摆州推动筹款、集会投票率和选民说服工作的。尽管尼克斯坚称数据技术并非能为糟糕候选人捏造胜利的灵丹妙药，但他将其定义为不可或缺的竞争优势。会议最后进行了毫不避讳的问答环节，涉及欧洲隐私法规、媒体反弹，以及现场观众就助选唐纳德·特朗普的企业伦理问题提出的激烈质问。剑桥分析公司首席执行官亚历山大·尼克斯概述了该行业从直觉驱动的“广告狂人”向算法驱动的“数学怪才”的转变，并以行为科学（OCEAN模型）、大数据分析和可定向广告技术为基石。 (0:00) 在唐纳德·特朗普2016年竞选期间，剑桥分析公司在全美范围内每月开展30万至40万次民调，为预测模型提供支持，并指导了超过1亿美元的数字广告投放。 (12:16) 谷歌针对定向竞选视频进行的一项独立品牌提升研究表明，唐纳德·特朗普的搜索量提升了18%，个人好感度提升了11%，投票意向提升了8%。 (17:05) 尼克斯声称有近200万人自愿参加了性格测试，以构建用于预测更广泛选民数据库中心理特征的训练数据集。 (23:51)

Reading summary

Alexander Nix of Cambridge Analytica argues that political campaigns and commercial marketing have shifted from intuition-driven creative approaches to algorithmic data science combining psychographics, centralized data records, and addressable ad tech. While acknowledging that candidates rather than technologies win campaigns, Nix asserts that granular behavioral micro-targeting provides an indispensable competitive advantage in modern, razor-thin elections. The Three Pillars of Modern Behavioral Targeting (0:00) Nix outlines a framework resting on three interconnected components: behavioral science, big data analytics, and addressable ad technology. He criticizes standard demographic segmentation—such as grouping audiences solely by age or gender—as an illogical proxy for predicting human behavior. Instead, Cambridge Analytica measures personality traits through the academic OCEAN framework (Openness, Conscientiousness, Extroversion, Agreeableness, and Neuroticism), tailoring creative execution to underlying psychological triggers such as risk aversion or status seeking. This psychographic profiling is linked to a centralized database that aggregates factual demographics, lifestyle habits, and inferred behavioral tendencies to ascertain what messages target groups need to hear before copy is written. Nix asserts that blanket mass advertising is dead, highlighting addressable tools such as set-top box viewer matching and hyperlocal cable headends that allow advertisers to deliver distinct, individualized messages to different members of the same household. Operational Scale in the 2016 Trump Campaign (12:16) Cambridge Analytica joined Donald Trump's 2016 presidential effort when the operation was lean, employing just 30 full-time staff compared to Hillary Clinton's roughly 800-person team. Capitalizing on 14 months of prior voter research and technical infrastructure developed during the primary bids of Ben Carson and Ted Cruz, Cambridge Analytica operated as an end-to-end technology partner. The firm executed approximately 1,500 polls per key state weekly—totaling 300,000 to 400,000 monthly nationwide—to feed live reporting dashboards and predictive voter models, ultimately directing over $100 million in digital advertising and tens of millions in data-driven television. This real-time tracking guided resource allocation during high-stakes campaign moments. Following the release of the candidate's controversial audio recording, state-level tracking revealed surprising pockets of upward mobility despite national backlash, preventing broad strategic panic. Similarly, geospatial heat maps identified concentrations of persuadable voters in historically Democratic Wisconsin, a state Clinton did not visit. Acting on this data, the Trump campaign staged five rallies reaching 60,000 to 70,000 voters in person, preceding a statewide victory margin of roughly 50,000 votes. Predictive Modeling and Measured Campaign Lift (15:23) The campaign deployed predictive models across four domains: small-dollar fundraising, candidate preference, turnout propensity, and issue salience. In its opening month, targeted small-dollar solicitations raised nearly $27 million and gathered one million email addresses. Issue modeling predicted social and policy priorities for every adult in the United States, allowing operatives to segment the electorate into core donors, get-out-the-vote segments, and centrist persuasion targets who received customized policy messaging. Over four months, Cambridge Analytica delivered 1.4 billion digital impressions across 4,000 discrete ad campaigns, reporting an estimated 3% favorability lift and a 2% bump in absentee ballot submissions. Nix cited an independent Google brand lift study of 150,000 targeted video viewers that recorded an 18% increase in campaign search queries, an 11% increase in personal favorability toward Trump, and an 8% lift in stated voter intent compared to an unexposed control group. Data Gathering Mechanics, Privacy, and Corporate Ethics (19:46) Addressing the collection of psychometric profiles, Nix explained that Cambridge Analytica gathered baseline training data from voluntary personality tests completed by nearly two million US participants, which were then mapped onto commercial consumer databases. Contrasting regulatory frameworks, he noted that while US data acquisition relies heavily on an opt-out standard, European deployment requires strict opt-in consent, limiting European models to voluntary primary research or a company's internal proprietary customer files. Facing audience pushback regarding the ethics of aiding Donald Trump's victory, Nix maintained that Cambridge Analytica operates as a neutral commercial service, conducting 8 to 10 national political campaigns annually for both left-of-center and right-of-center parties worldwide. Nix argued that providing technical capability within legal, democratic elections is a valid commercial practice, judging the firm's engagement strictly by its professional execution rather than the personal conduct or ideology of the candidate. Nix concludes that while behavioral data science grants an essential tactical advantage in competitive races, technology is not a panacea capable of transforming a flawed candidate into a winner or deciding an election on its own. The presentation highlights the growing deployment of individual psychographic profiling in politics, while leaving open unresolved tensions surrounding the ethical boundaries and democratic consequences of applying corporate marketing machinery to civic elections.

精炼阅读摘要

Cambridge Analytica首席执行官Alexander Nix在OMR 2017大会上提出，现代营销与政治竞选已从依赖经验直觉的传统广告时代，转向由行为科学、大数据分析与精准投放技术驱动的算法时代；但他同时强调，数据技术并非万能灵药，选举的决定性因素仍在候选人本身而非技术。驱动传播转型的三大技术支柱 (0:00) Nix认为，传统的受众细分方式仅仅依赖年龄、性别、地域等人口统计学变量，这种粗放假设忽视了个性才是人类行为的根本驱动力。Cambridge Analytica的方法论建立在三大互为支撑的支柱之上：第一是基于学术标准OCEAN人格模型（开放性、尽责性、外向性、宜人性、神经质）的行为科学，通过评估深层心理特质来指导内容创意的定制；第二是构建涵盖事实人口属性、生活方式偏好及心理画像的大数据统一数据库，在创意启动前先行锁定受众需求；第三是可寻址广告技术，结合机顶盒收视数据与本地化网络，实现家庭层面的差异化精准投放，彻底终结了无差别的“地毯式广告”。 2016年特朗普竞选的底层数据工程 (12:16) 在2016年加入特朗普竞选团队时，面对仅有30名全职雇员的精简团队（希拉里团队约有800人），Cambridge Analytica迅速移植了此前在初选期间积累的14个月选民数据基础设施，作为端到端技术提供商介入竞选。团队建立了高密度的民意监测机制，在关键州每周开展约1500次民意调查，全美每月追踪总量达30万至40万次，借此向动态仪表盘输入实时数据，指导资源调配。例如，在特朗普录音丑闻爆发后，实时民调显示其支持率在部分州暴跌但在另一些州有所反弹，使竞选管理层避免了全面恐慌。同时，团队开发了涵盖捐赠者、候选人偏好、投票率与议题关注度的预测模型，向全美所有成年人赋予预测性的议题评分，并将选民划分为核心支持者、投票动员对象与中立说服群体。基于对摇摆选民分布热力图的分析，团队发现历来被视作民主党阵地的威斯康星州存在大量可说服选民；特朗普团队据此在该州举行了5场大型集会，直接触达6至7万名选民，最终以约5万票的微弱优势赢得该州。投放转化表现与技术效用的现实边界 (17:05) 这一数据建模系统支撑了大规模的线上线下协同执行。在竞选首月，团队通过小额筹款模型募得近2700万美元，并获取了约100万个支持者邮箱地址；在四个月内投放了4000个独立广告项目，累计产生14亿次曝光，将特朗普的净好感度推升约3%。在整个竞选期间，该机构负责了超过1亿美元的数字广告投放以及数千万元的数据驱动型电视广告。 Google开展的一项包含约15万人的独立品牌提升研究显示，相较于接受通用信息的对照组，接触定向视频内容的受众在竞选相关搜索量上提升了18%，对特朗普的个人好感度上升了11%，投票意向增加了8%。然而，Nix在演讲中明确划定了技术能力的边界，驳斥了“算法决定选举”的夸大说法；他指出数据科学无法化腐朽为神奇，不能将劣质候选人变为优秀候选人，但在势均力敌的选举中，精细化的数据运营能为竞选提供关键竞争优势。心理画像的数据采集与法规模拟 (23:51) 针对外界对心理学数据获取渠道的质疑，Nix解释称，系统训练数据并非通过强制手段搜集，而是依赖大量用户出于好奇心与自我探索的主观意愿，主动参与社交媒体上的在线测试和问卷调查。据Nix披露，全美有近200万人自愿完成了OCEAN性格测试；分析团队随后将这些样本得分与商业消费数据库进行交叉匹配，利用机器学习建立针对全美选民的概率预测模型。在跨国业务拓展方面，Nix对比了欧美之间的数据监管环境差异。他指出，美国主要遵循选择退出（opt-out）机制，允许企业在公开市场上直接购买商业消费者档案；而欧洲与德国则实行严格的明确授权（opt-in）法规。因此，直接购买消费者数据的模式在欧洲不可行，但机构仍可通过自愿参与的原始问卷调研，或依托欧洲商业企业自身已合法积累的第一方客户数据库来开展行为科学建模。政治咨询的中立性辩护与伦理拷问 (26:30) 在问答环节，现场观众情绪激烈地质疑该机构帮助争议性候选人执政的道德责任。Nix在回应中拒绝对候选人个人言论置评，强调外部观察者应当尊重美国选民在民主体制下的自决权。他指出特朗普是在击败了16位党内竞争对手后通过合法选举入主白宫，公司的职责仅在于提供专业的竞选服务，并对自身执行的工作成果表示满意。针对外界将政治咨询等同于推广消费品（如牙膏）的批评，Nix承认政治竞选天生具有撕裂性，机构必须谨慎评估候选人对企业声誉、员工及其他客户的潜在波及。但他辩护称，Cambridge Analytica每年在全球协助8至10场国家级元首竞选，业务覆盖左翼与右翼阵营；只要选举遵循自由公正的民主程序，为候选人提供专业的数据咨询就是保障其正当代表权的合法服务，且公司迄今未曾正式拒绝过任何客户。 Alexander Nix将算法化微观定向描绘为政治与商业传播的不可逆进化，试图将数据武器化的伦理指责归化为客观的专业服务。然而，这场演讲仍留下了未解的核心矛盾：在算法能够隐蔽地针对人性弱点（如神经质倾向的焦虑感）施加情感刺激并重塑选民意向时，单纯以程序合规与选民自决为由推卸技术背后的道德与社会责任，并未真正消除公众对数据操纵民主进程的深层疑虑。

Detailed notes — zh

## Background 该视频由 OMR 于 2017 年 3 月 10 日发布在 YouTube 平台上。视频记录了剑桥分析公司（Cambridge Analytica）首席执行官亚历山大·尼克斯（Alexander Nix）在德国汉堡举行的 2017 年 OMR 营销盛会上的主题演讲，内容围绕大数据、行为识别方法以及数据分析如何深度介入唐纳德·特朗普（Donald Trump）的总统竞选活动展开。 ## TL;DR 亚历山大·尼克斯系统阐述了营销传播正从创意主导的《广告狂人》时代转向由行为科学、大数据分析与可寻址广告技术构成的《数学狂人》时代。他拆解了基于 OCEAN 大五人格模型的心智画像体系，并全面复盘了剑桥分析在 2016 年特朗普竞选期间所主导的民调建模、小额募资、广告测试与精准人群说服等实战细节。在随后的问答环节中，尼克斯还就欧美数据隐私法规差异、选民数据获取的合规性，以及协助特朗普当选所引发的道德与民主争议做出了正面回应。 ## Atmosphere & style 整场演讲开篇呈现出严谨、沉稳且极富说服力的商业技术发布会质感，尼克斯以冷静而自信的精英姿态系统拆解复杂的受众建模理论。然而进入问答环节后，现场气氛骤然升温，充斥着浓烈的针锋相对与思想碰撞，观众抛出的道德质问尖锐而直白，引发台下骚动与阵阵掌声。尼克斯始终保持克制、逻辑严密但略带防御姿态的应答，让整场活动在尖锐的技术探讨与政治伦理交锋中呈现出极具张力的戏剧性波澜。 ## Bulletin - [12:16](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=736s) 剑桥分析（Cambridge Analytica）在 2016 年特朗普竞选期间投放了逾 1 亿美元数字广告和数千万美元精准电视广告，每月在全美开展 30 万至 40 万份周度民调以动态指引资源分配。 - [15:23](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=923s) 通过结合 OCEAN 大五人格模型与微观受众建模，剑桥分析实现了为全美每一个成年人打上最关切的议题标签，并打破传统粗放式无差别泛投放广告。 - [17:05](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1025s) Google 针对剑桥分析投放成效的独立品牌提升测试显示，接收定制化视频的受众在搜索量上提升 18%，对候选人好感度增加 11%，投票意愿净增 8%。 - [23:51](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1431s) 针对欧洲与美国的数据合规差异，亚历山大·尼克斯指出欧洲严格要求“选择进入”（Opt-in），不能像美国（Opt-out）那样批量购买第三方数据，但仍可依靠合法自愿的一手调研及企业私域数据建模。 ## Key takeaways - [00:00](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=0s) 现代营销传播正经历从传统自上而下的《广告狂人》（Mad Men）向数据与算法驱动的《数学狂人》（Math Men）时代深刻转变。 - [00:00](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=0s) 剑桥分析的核心方法论建立在三大技术支柱上：行为科学（深层动机）、大数据分析（受众建模）以及可寻址广告技术（点对点触达）。 - [02:00](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=120s) 传统人口统计学细分（如单纯依据性别、种族或年龄）已过时，驱动个体决策的最核心底层引擎是心理画像与性格机制。 - [03:00](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=180s) 公司采用实验心理学中的 OCEAN 大五人格模型（开放性、尽责性、外向性、宜人性、神经质）来定制完全不同的信息诉求与视觉创意。 - [09:01](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=541s) 无差别的泛投放广告（Blanket Advertising）已经死亡，未来趋势是同一屋檐下的夫妻也会收到完全差异化的定制内容。 - [09:30](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=570s) 数据与可寻址技术的威力受到巨头资本背书：目标受众密度每提高 10% 可带来 7% 的增量销售额，沃尔玛、福特及 WPP 等行业巨头正全力加码数据能力。 - [12:16](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=736s) 介入特朗普竞选初期，团队面对的是仅 30 人的精简班子（对比希拉里团队约 800 人），并依托此前初选中沉淀的技术基建快速展开全流程技术服务。 - [13:40](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=820s) 周度动态民调让团队能够精准研判公关危机（如“热麦克风事件”）的真实地域影响，并识别出传统蓝州（如威斯康星州）隐藏的大量可说服选民，从而实现边缘逆转。 - [15:23](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=923s) 竞选模型体系涵盖候选人偏好、出门投票率和议题偏好三大维度，形成了将选民精准分为核心支持者、催票人群及重点说服群体的二维矩阵。 - [18:00](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1080s) 尼克斯强调数据科学并非万灵药，无法凭空将差劲的候选人包装为优秀候选人，但能在势均力敌的拉锯战中起到关键的边缘推动作用。 - [19:46](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1186s) 剑桥分析坚称在政治合作上保持中立，全球服务横跨左翼与右翼阵营，其 OCEAN 心理测试数据源自近两百万名知情且自愿参与的线上用户。 - [26:30](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1590s) 面对现场关于帮助争议候选人胜选的激烈道德质问，尼克斯回应称特朗普是通过自由公正的民主选举获胜，技术服务商的职责是为客户提供平等的专业服务而非操纵民意。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=0s) 从《广告狂人》到《数学狂人》：行为科学与心理画像亚历山大·尼克斯（Alexander Nix）开门见山地向汉堡的听众致意，指出营销与传播领域正在经历一场从《广告狂人》（Mad Men）时代向《数学狂人》（Math Men）时代的深刻演变，而这正是剑桥分析公司（Cambridge Analytica）在 2016 年唐纳德·特朗普（Donald Trump）美国总统竞选活动中所深度参与的实践。尼克斯提到，特朗普在整个竞选周期中多次公开强调，赢得竞选的是候选人本身，而不是技术；针对这一点，尼克斯由衷表示赞同，但他也随即指出，有三项关键技术与方法论正在彻底重塑当今的政治竞选、品牌推广和商业传播。 ```mermaid flowchart TD accTitle: Cambridge Analytica 核心传播三角架构 accDescr: 展示行为科学、大数据分析与可寻址广告技术如何协同构建个性化传播流程 subgraph Science["1. 行为科学 (Behavioral Science)"] OCEAN["OCEAN 心理性格模型\n(开放/尽责/外向/宜人/神经质)"] Motivation["深层动机挖掘\n(超越传统人口统计标签)"] OCEAN --> Motivation end subgraph Data["2. 大数据分析 (Data Analytics)"] Agg["多维数据聚合\n(人口学/生活方式/行为数据)"] Model["微观受众建模\n(价格敏感度 vs 品牌偏好)"] Agg --> Model end subgraph AdTech["3. 可寻址广告 (Addressable AdTech)"] CookieMatch["线下数据与 Cookie 匹配"] HyperTV["机顶盒数据与超本地化电视投放"] CookieMatch --> PersonalMsg["一对一个性化触达"] HyperTV --> PersonalMsg end Motivation --> Agg Model --> CookieMatch ``` 尼克斯谈到的第一项技术是行为科学（Behavioral Science）。为了生动阐明传统信息传播与行为传播的区别，他展示了一个海滩警示牌的例子：如果你拥有一处私人海滩并希望阻止人们下海游泳，左侧传统的“信息型传播”（Informational Communication）只会挂出写着“禁止游泳”的告示牌，试图说服人们的态度；而右侧的“行为型传播”（Behavioral Communication）则直接立起一块画着大鲨鱼张开血盆大口的警示牌。显然，被鲨鱼吃掉的致命威胁能直接触及更深层的恐惧与本能动机，形成极具威力的震慑。尼克斯严厉批判了当前绝大多数广告与公关机构依然奉为圭臬的细分方式。这些机构仍旧基于人口统计学（Demographics）、地理位置（Geographics）、消费与生活方式属性以及媒体消费习惯来划分受众。尼克斯指出这种做法只要稍微深思就显得荒唐可笑：“仅仅因为性别就认为所有女性都应该接收完全相同的讯息，或者认为所有非裔美国人只因种族就毫无二致，亦或把所有老年人、富人或年轻人都按人口统计学标签一刀切，这根本毫无逻辑。”他强调，人口和地理属性固然影响世界观，但决定人类决策的最关键底层引擎其实是性格（Personality），也就是所谓的心理画像（Psychographics），因为真正驱动人们投票或购买行为的正是性格机制。剑桥分析引入了实验心理学前沿的“OCEAN 大五人格模型”（Five-Factor Model），通过长篇量化工具精准测算个体的五大特质： - **开放性（Openness）**：个体对新鲜事物的接纳程度； - **尽责性（Conscientiousness）**：个体对生活秩序、习惯和周密规划的看重程度； - **外向性（Extroversion）**：个体的社交取向与精力外倾程度； - **宜人性（Agreeableness）**：个体更倾向于优先考虑自身利益，还是优先顾及社会与社区福祉； - **神经质（Neuroticism）**：衡量个体焦虑、担忧与情绪波动倾向的指标。为了展示性格定位的商业威力，尼克斯举了两个人口学特征几乎完全一致的男性潜在购车客户：两人拥有相同的性别、相似的高收入水平以及大致相当的生活方式。如果单纯按照传统人口统计学，他们会收到同一套汽车广告。然而，数据揭示左侧的男士性格高度尽责且宜人性极高，而右侧的男士则极度外向。针对宜人性高且尽责的人群，广告应当强调契合当今社会责任的品质（如环保电动引擎），并辅以理性、注重事实依据的文案说服；反观那位外向型人群，营销策略则应直接推销跑车，并在画面与措辞上全方位激发其对驾驶体验和激情的感官渴望。 --- ### [05:28](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=328s) 大数据分析：从创意猜想到受众建模尼克斯阐述的第二项支柱技术是大数据分析（Data Analytics）。他回顾了 20 世纪 50 年代《广告狂人》的黄金时代，那时的创意过程完全是“自上而下”（Top-Down）的——几位才华横溢的天才创意总监在办公室里构思出精妙的概念，随后生硬地推向大众市场，祈祷能够引起受众的共鸣。然而在当今时代，大数据彻底逆转了这一逻辑。营销人员可以在创意构思尚未动笔之前，就精准获知目标群体中每一个微观小群体究竟渴望听到什么样的内容。所谓大数据，本质上就是尽可能穷尽并聚合个人生活中的所有数据点。这些数据跨越三个维度：一是事实性与人口统计学数据（年龄、性别、种族、宗教信仰等）；二是态度与生活方式数据（驾驶什么车型、订阅什么杂志、参与何种社交媒体互动、前往哪座教堂做礼拜、属于哪家高尔夫俱乐部）；三是行为与心理性格数据。剑桥分析将所有维度的海量数据沉淀并合成为一个权威数据底座（Database of Record），在此基础上训练出极其强大复杂的受众模型，从而主导高度细分的一对一传播。尼克斯借助一个针对日常消费品——牙膏的营销仪表盘展示了这一建模逻辑： - **受众参与地图的二维划分**：横纵坐标分别区分了“价格敏感型”（易受经济学促销诱导）与“品牌偏好型”（易受品牌认同驱动）的人群。 - **差异化诉求切入点**：仪表盘右侧的价格敏感型群体更容易被直接的折扣优惠打动；而注重品牌偏好的人群则必须依靠详实的产品功能特征（如去除牙渍、美白牙齿、清新口气、保护牙龈健康等）来进行逻辑说服。 - **媒介习惯与购买力叠加**：仪表盘还综合了受众最常活跃的媒体渠道、预期收入及消费潜力，并叠加心理性格特征。当这一整套数据模型落地到具体的广告创意制作时，其精度展现无遗：如果模型显示目标群体以 25 至 30 岁、单身、中等收入的年轻女性为主，创意的视觉形象与文案语调就必须贴合这一形象；数据明确显示该群体在 Facebook 上高度活跃，因此 Facebook 便成为首选投放渠道；在产品卖点上，模型判定她们最关切的是“去除牙渍”；最后，心理画像显示该人群具有较高的神经质（情绪焦虑倾向），这就意味着广告必须诉诸情绪化的共鸣表达。因此，一句“在污渍毁掉你的笑容之前消除它们”的警示性口号，便构成了击中该人群心理痛点的致命创意。 --- ### [09:01](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=541s) 可寻址广告技术与巨头资本的全面转向尼克斯提出的第三大支柱是可寻址广告技术（Addressable AdTech）。他旗帜鲜明地下了断言：“我以前说过，今天我还要再说一遍：我坚信无差别的泛投放广告（Blanket Advertising）已经死亡。”他认为，向数以百万计的人群播放同一支广告的历史很快就会变成下一代人完全无法理解的陈旧遗迹。现代传播正以前所未有的速度迈向高度个体化时代，受众与品牌之间正在建立起点对点的私人关联。无论是通过纸质直邮、电子邮件还是短信，品牌都在与活生生的个人对话——甚至在同一个屋檐下的夫妻二人，面对同一家公司的同一款产品，也会在各自的屏幕上收到截然不同的广告内容。为了证明精准受众密度的商业价值，尼克斯援引了行业权威的比尔·哈维（Bill Harvey）的经典论断：“平均而言，购买目标受众密度每提高 10%，就会带来 7% 的增量销售额增长。” > “Bill Harvey famously said, on average a 10% increase in purchase target density will generate a 7% increase in incremental sales.” 在实操落地中，可寻址技术主要依托以下手段实现极高密度的精准触达： - **数字身份打通**：通过将线下的目标受众细分数据与线上 Cookie 进行交叉比对与上传，营销团队能够锁定最密集的核心目标受众并为其定向推送定制内容。 - **机顶盒与有线电视数据匹配**：在美国市场，团队能够将线下的目标受众标签直接与电视机顶盒（STB）的收视行为数据相匹配。用户的每一次收看记录都会回传给有线电视运营商，分析师由此清晰掌握不同人群的荧幕偏好，从而精准购买目标受众密度最高的电视节目广告时段。 - **超本地化电视投放（Hyperlocal TV）**：打破全美甚至全州的大规模广播覆盖，直接向当地特定的有线电视服务商采购广告位，将投放范围精确限制在仅有少数几个邮政编码（Zip Codes）的极小地理网格内，锁定关键核心受众。尼克斯提醒现场听众，如果对这股数据科学转型的浪潮仍有迟疑，不妨看看当今巨头企业的巨额资本流向。零售巨头沃尔玛（Walmart）和汽车巨头福特（Ford）在过去几年均砸下数亿美元自建或并购数据分析能力；广告业巨擘奥美（Ogilvy）就在上周正式成立了行为科学中心（Center for Behavioral Science），全力探索心理学在营销中的落地；零售巨头克罗格（Kroger）则果断收购了英国数据分析先锋企业 dunnhumby；全球传播巨头 WPP 的掌舵人马丁·索雷尔（Martin Sorrell）早在至少四年前就不断高调宣示广告重心正从传统创意全面倒向科学与数据；法国宏盟传媒集团（Havas）等全球机构亦在亦步亦趋地跟进。在完成了三大方法论的铺垫后，尼克斯宣布将利用接下来的演示幻灯片，向全场详细复盘剑桥分析公司自 2016 年 6 月正式入驻并主导特朗普总统竞选活动以来的真实运作内幕。 ### [12:16](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=736s) 临危受命：Cambridge Analytica 接入特朗普竞选团队在 2016 年 Cambridge Analytica（剑桥分析公司）受聘加入特朗普（Donald Trump）竞选团队时，整个竞选团队的全职员工仅有区区 30 人；相比之下，希拉里·克林顿（Hillary Clinton）团队当时拥有近 800 名工作人员，双方规模悬殊。Alexander Nix 强调，Cambridge Analytica 之所以成为极具吸引力的合作方，是因为团队在此前的 14 个月里深度参与了本·卡森（Ben Carson）和特德·克鲁兹（Ted Cruz）的党内初选，积累了海量的选民数据与深度调研成果，并已构建起一套成熟的技术基础设施，能直接移交给特朗普团队使用。公司被聘为特朗普团队的全流程端到端（end-to-end）技术服务商。业务首先从海量民调调研切入：团队每周在每个州开展约 1,500 份民意调查，全美每月调研总量高达 30 万至 40 万份。这些数据被实时汇入监控仪表板，指导竞选资金与资源的动态分配，并不断优化预测模型。接着，模型推导出精细的用户分群，最终落地为极具针对性的触达方案。在整个竞选期间，Cambridge Analytica 为特朗普团队投放了超过 1 亿美元的数字广告，以及数千万美元的数据驱动型精准电视广告。 ```mermaid flowchart TD accTitle: Cambridge Analytica 竞选数据驱动闭环流程 accDescr: 展现从全美大规模周度民意调研、构建预测模型、选民受众细分到跨渠道精准触达的完整闭环。 subgraph DataCollection["数据收集与建模"] A["每周各州约 1,500 份民调 （每月 30-40 万份）"] --> B["动态数据仪表板 （资源配置决策）"] B --> C["多维预测模型 （捐款/候选人偏好/投票率/议题）"] end subgraph TargetingAndExecution["精准分群与执行"] C --> D["人群矩阵细分 （核心支持者/催票人群/摇摆人群）"] D --> E["按关注议题进行二级子分群"] E --> F["精准媒介执行 （1 亿+ 美元数字广告 / 数千万美元定向电视）"] end ``` - 特朗普竞选团队初期全职雇员仅 30 人，希拉里团队约为 800 人。 - 团队在此前服务 Ben Carson 和 Ted Cruz 的 14 个月中已沉淀了底层技术与数据基础。 - 调研频率达到每周每州约 1,500 份，月均 300,000 至 400,000 份。 - 投放规模包括超 1 亿美元数字广告和数千万美元的精准定向电视广告。 --- ### [13:40](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=820s) 动态公关应对、摇摆选民热力图与威斯康星州战例 Alexander Nix 详细展示了数据仪表板在实战中的两大核心价值：动态追踪选民变化与发现关键摇摆人群。周度民调让团队能够极速应对突发危机。典型案例是特朗普的“热麦克风事件”（hot mic incident，即录音泄露引发的更衣室污言秽语风波）。主流舆论普遍认为该事件对特朗普造成毁灭性打击，但监测数据显示，特朗普的支持率虽然在部分州暴跌，却在另外几个州意外上升。这让竞选团队得以避开恐慌情绪，将公关和媒体资源高度聚焦在受损严重的特定州。另一个关键应用是通过仪表板生成“可被说服选民”（persuadable voters）的热力图。Alexander Nix 以威斯康星州（Wisconsin）为例：该州传统上是坚挺的民主党票仓，以至于希拉里在整个大选竞选期间甚至从未踏足过该州一步。然而，Cambridge Analytica 的数据模型精准识别出威斯康星州隐藏着规模巨大的可说服选民群体。特朗普团队据此果断在该州连续举行了 5 场大型集会，直接面对面触达了约 6 万至 7 万名选民。最终特朗普在该州仅以约 5 万票的微弱优势胜出，强有力地印证了精准数据驱动策略的实际威力。 --- ### [15:23](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=923s) 预测模型矩阵、小额筹款与“给全美每个成年人打标签” 任何政治竞选都离不开资金机器。竞选最初的一个月，团队的核心任务是识别潜在的小额捐助者，募集 5 美元、20 美元或更小额度的捐款，为后续媒介采购积累弹药。该系统运转成熟后直接移交给了共和党全国委员会（RNC）。随后，数据团队迅速转向四大核心模型： 1. **候选人偏好模型（Candidate Preference Models）**：分析哪些候选人对特定选民最具吸引力。 2. **投票率模型（Turnout Models）**：测算选民的实际出门投票概率，并重点估算未登记选民可能带来的冲击。 3. **议题偏好模型（Issue Models）**：预测选民对特定社会与政治议题的敏感度。Alexander Nix 披露：“We actually assigned different issues to every adult in the entire United States.”（我们实际上给全美每一个成年人都打上了他们最关心的议题标签。） 4. **二维受众细分矩阵**：将选民按意向与行为分类，位于右上角的核心支持者用于捐款转化；中间及偏右的人群属于催票对象（Get Out The Vote）；中间偏左的选民则是需要向右拉动的关键说服对象（Persuasion Audiences）。在确定主群组后，再结合其特定关心的议题进行精细子分群，实现“千人千面”的信息定制。 --- ### [17:05](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1025s) 实战战报与 Google 独立品牌提升研究 Alexander Nix 列举了一系列具体的实战成果，全面展示了这套技术体系在执行层面的交付能力： - **数字筹款与私域沉淀**：启动首月即募集近 2,700 万美元小额捐款，收集并登记了约 100 万个支持者邮箱地址。 - **集会动员与 App 下载**：直接推广了近 100 场线下集会，驱动至少 3.5 万次特朗普竞选官方 App 下载。 - **精准劝票（Persuasion）**：在短短 4 个月内，跨 4,000 个独立的广告系列总共投放了惊人的 14 亿次曝光（impressions），推动特朗普的好感度净增长约 3%。 - **缺席投票与投票动员（GOTV）**：实现约 8,000 万次曝光，使得缺席选票提交率（absentee ballot submissions）提升了约 2%。 - **超细分地理媒介采购**：利用数据指引，广告采购没有像希拉里团队那样采取昂贵且粗放的“广谱撒网”（broad brush），而是精准锁定向极小规模的特朗普支持者聚集区。在此期间，Google 针对 Cambridge Analytica 的作业成效进行了一项独立的品牌提升研究（Brand Lift Study）。测试选取了约 15 万名用户，实验组接收高度定制的定向数字视频，对照组则仅接收泛化通用信息。Google 的测试报告显示：收到定制沟通的受众在搜索量上获得了 18% 的增长提升，对特朗普的好感度增长了 11%，投票给特朗普的意愿（voter intent）净增长了 8%。在全场掌声响起前，Alexander Nix 给出了他对数据技术的定位与总结，留下一句警醒：“If there's one singular takeaway from this event that is that data science technology is not a panacea. You cannot make a bad candidate into a good one. You cannot swing an election.”（如果说从这次大选中能吸取什么核心教训，那就是：数据科学技术绝非万灵药。你不可能把一个差劲的候选人包装成优秀的候选人，你也不可能单凭技术逆转一场选举。） --- ### [19:46](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1186s) 现场追问：中立立场、争议调查报道与大五人格（OCEAN）数据源起演讲结束后，主持人登台开启问答，首度抛出锋利质问：这是否意味着即便没有 Cambridge Analytica，特朗普也能当选？Alexander Nix 回应称，当今的选举往往在极微弱的差距中决出胜负，技术固然不可能在劣势悬殊时单凭自身左右大局，但在微弱差距的拉锯战中，精准技术确实能起到决定胜负的边缘推动作用。面对团队是否有政治倾向的提问，Alexander Nix 澄清公司每年在全美乃至全球承接 8 到 10 场国家元首或总理级别的竞选服务，其中左翼与右翼阵营各占一半。作为专业技术提供商，他们始终坚持“把个人政治意识形态留在门外”，只要在合理法律与商业框架内，面向任何主流政党开放合作。当被问及为何未参与即将在同年 9 月举行的德国大选、默克尔（Angela Merkel）来电是否会合作时，他笑称若默克尔致电将是自己的荣幸，但团队刚经历长达两年高压强度的美国大选，全员处于休整期，繁复的时间节点使得他们目前无法抽身支援德国竞选。主持人进一步追问了瑞士杂志《Das Magazin》撰写、在德国社交媒体（特别是 Facebook）上引发刷屏狂潮的深度调查文章。Alexander Nix 对此回应：文章作者在缺乏全貌信息的情况下生硬拼接因果关系，文中充斥着严重的失实与谬误（“enormous falsehoods”），但他也承认文章中提到的部分技术工具与理念确实是团队所使用的。最后，针对全场最为好奇的底层机制——大五人格模型（OCEAN Profile，即开放性、尽责性、外向性、宜人性、神经质）的数据来源与 Cookie 匹配问题，观众提问选民为何愿意交出如此隐私的心理画像数据。Alexander Nix 回应道，公民完全没有义务透露个人信息，但团队在美国通过用户主动参与的调研与测试，合法收集了数以百万计的自愿提交的 OCEAN 心理画像数据，并进一步将这些心理画像与多维选民属性数据库进行了深度交叉匹配。 ### [23:51](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1431s) 心理画像数据的收集机制与欧美数据合规差异现场主持人向 Alexander Nix 提问，探究剑桥分析公司（Cambridge Analytica）究竟是如何获取这些用于训练模型的心理测试数据的，用户花五分钟甚至更长时间回答大量问题到底图什么。Nix 解释称，受众参与社交媒体问卷的动机各有不同，许多人纯粹是出于对自我探索（Self-inquiry）的强烈兴趣，渴望了解自己的性格特质及其含义。他指出，已有将近两百万人参与了这项测试，这证明此类自我认知驱动具有强大的吸引力。面对主持人“这是否就像一款让人一边回答问题了解自己、一边将数据拱手相让的 Facebook 游戏”的追问，Nix 坚决反驳了任何暗箱操作的指控，强调受访者是在完全知情的前提下自愿提供数据，整个过程不存在任何恶意或不正当行为。随后，主持人提出了欧洲与德国极为严苛的数据隐私法规与美国法律体系的对比，质疑剑桥分析在美采用的运作模式在德国法律框架下是否具有可行性。Nix 从宏观法制文化层面进行了剖析，指出美国的数据合规文化本质上是“选择退出”（Opt-out）机制，而欧洲则严格遵循“选择进入”（Opt-in）原则。因此，在欧洲显然不能像在美国那样直接在公开市场上批量采买第三方数据，但依然可以通过合法途径获取一手研究数据（First-party research，如大五人格 OCEAN 模型问卷）；只要用户自愿授权将其数据用于市场营销与传播，就可以在此基础上构建预测模型。此外，许多大型企业自身已经沉淀了海量私域数据，这些数据完全可以在其内部合法用于建模和营销转化。 ```mermaid flowchart TD accTitle: 心理学数据获取与跨国合规流程 accDescr: 描述剑桥分析获取问卷数据并根据欧美不同隐私法规进行建模与应用的路径。 subgraph Collection["数据获取端 (自愿参与)"] A["社交媒体性格测试 / OCEAN 问卷"] --> B["用户动机：自我探索与个性诊断"] B --> C["沉淀近200万样本的一手训练集"] end subgraph Legal["法制与合规环境对比"] C --> D{"法律适用区域"} D -->|"美国市场 (Opt-out 机制)"| E["直接采买第三方数据 + 结合训练集预测全量受众"] D -->|"欧洲/德国 (Opt-in 机制)"| F["必须依靠用户主动授权一手数据 / 客户企业私域数据建模"] end subgraph Output["商业与政治应用"] E --> G["个性化精准政治及商业营销"] F --> G end ``` - **受试样本规模**：接近 200 万人完成了性格测评，构成了大规模预测算法的基础训练集。 - **监管与文化差异**：美国采用 Opt-out（默认收集，用户可选择退出），欧洲与德国强制要求 Opt-in（事先明确授权同意）。 - Nix 强调受众数据来源的合法性与透明度：“They're voluntarily giving up their data they're doing this in in in you know in full knowledge of what's happening here this is this is not anything Macchiavellian or untoward.” --- ### [26:30](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1590s) 尖锐对质：帮助特朗普胜选的道德争议与民主辩护现场提问环节爆发了全场最激烈的交锋。一位提问者直截了当地向 Alexander Nix 抛出发难，台下顿时引起骚动与掌声：“I just actually have only one question so if it's really true that you help Trump get into office are you happy putting a misogynist buffoon the highest the the most powerful job in the world who's now messing up our world are you happy?” 面对听众关于把一个“仇女的小丑”送上全球最有权势的宝座是否感到良心不安的逼问，现场爆发出一阵阵热烈的掌声与紧张的情绪。 Nix 表现得十分克制而冷静，他反驳称外界应当退一步看清本质，其他国家的人无权对美国的民族自决指手画脚。他辩护道，特朗普当时是在共和党党内初选 17 位候选人这一历史上规模最大的阵容中竞争，完全是通过自由公正的选举堂堂正正赢得了党内提名，并在随后的大选中赢下了总统宝座。Nix 严正指出：“Trump was one of 17 presidential nominees the largest field ever and he won that nomination fair and square through free and fair elections and then went on to win the election so if that's the will of the people I don't think it's anyone's job in this room to to to go against that.” 提问者紧咬不放再次追问“那你本人究竟是否开心”，Nix 简短而有力地回应：“I'm happy with the job that we did.” - **初选背景数据**：特朗普在 2016 年共和党初选中面对多达 17 位竞争对手，系美国近代政治史上规模最大的总统初选阵容。 - **核心逻辑抗辩**：选举属于民主制度下的民意体现（Will of the people），作为服务商是履行技术支持职责，而不是对民主选举的结果负责。 --- ### [28:02](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1682s) 商业服务的中立性边界与争议底线随后主持人接过了话题，针对政治竞选与商业产品推销之间的界限展开深入探讨。主持人提问，剑桥分析既服务于消费品客户，也服务于政客与政党，对于一家数据服务公司而言，推销一款牙膏和推销特朗普到底有没有本质区别。Nix 承认，参与政治竞选因其固有属性而极具争议性和撕裂性（Divisive），因此公司必须异常审慎，确保所服务的候选人或政党不会对公司声誉造成无法挽回的损害，不会激怒自己的员工团队，也不会冒犯其他商业客户。但除去这些顾虑，只要竞选是自由公正的，民主原则得到了捍卫，公司就有义务为候选人提供最优质的服务与平等的代表权，因为这是实现公平选举的唯一路径。紧接着，现场再次抛出尖锐问题，追问剑桥分析是否有拒接客户的底线，以及他们是否会拒绝那些侮辱女性（暗指特朗普被曝光的“Grab them by the...”音频丑闻）的候选人。Nix 坦言公司肯定存在拒接的标准，但在实际操作中他们会尽量提前规避，不让自己陷入必须说“不”的被动境地，同时表示目前记不得有哪次明确拒绝过业务。针对特朗普录音丑闻的具体攻击，Nix 拒绝正面评价，表示媒体当时已有全方位的报道，自己也听到了正反两方面的不同声音。 - Nix 阐述数据技术公司在民主选举中的使命定位：“Beyond that you know as long as they're free and fair elections and democracy is being upheld we think it's the duty of our firm to provide candidates with the best possible service and fair representation that's the only way you can have fair elections.” - 政治与商业双轨制的风控考量：政治竞选极易波及商业品牌形象与内部团队士气，因此必须权衡声誉损益。 --- ### [29:32](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1772s) 会议闭幕致谢：透视数据技术浪潮的核心价值最后，大会主持人 Phil 走上前台结束了这场充满火药味与技术干货的炉边对谈。Phil 向 Alexander Nix 的到来与公开分享表示由衷感谢，称赞他愿意亲自站上面对德国观众的舞台，毫无保留地讲述剑桥分析在底层到底做了什么。Phil 总结道，尽管人们完全可以就政治倾向、左右立场以及道德是非展开无休止的争辩，但对整个行业而言，最关键、最迫切的事情是认清现实中正在发生什么，以及当下的前沿大数据技术究竟具备了怎样惊人的能力。 Phil 表示，OMR（Online Marketing Rockstars）的宗旨就是搭建一个关于内容、现象解读、前沿学习与信息交换的开放平台，而 Nix 的这场主题演讲与深度答辩，对这个平台而言是一次极为充实且极具价值的贡献，对现场的所有数字营销从业者都具有启发性。整个演讲在掌声与主持人的道谢中落下帷幕。 - 主持人 Phil 对技术探讨与价值争议的总结陈词：“Obviously you can argue left and right and right and wrong but in the end what matters is to see what's happening and what you might be able of doing what you explained to us today and we are here to provide a platform for content for description for learning for information.” ## Notable quotes > [17:05](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1025s) If there's one singular takeaway from this event that is that data science technology is not a panacea. You cannot make a bad candidate into a good one. You cannot swing an election. — Alexander Nix > [28:02](https://www.youtube.com/watch?v=6bG5ps5KdDo&t=1682s) Beyond that you know as long as they're free and fair elections and democracy is being upheld we think it's the duty of our firm to provide candidates with the best possible service and fair representation that's the only way you can have fair elections. — Alexander Nix ## Mentioned Alexander Nix, Cambridge Analytica, Donald Trump, Hillary Clinton, Ben Carson, Ted Cruz, Angela Merkel, Bill Harvey, Martin Sorrell, Phil, Google, Facebook, Walmart, Ford, Ogilvy, Kroger, dunnhumby, WPP, Havas, Republican National Committee (RNC), Online Marketing Rockstars (OMR), Das Magazin, Mad Men, OCEAN 大五人格模型 ## Video description From mad men to math men. Cambridge Analytica first gained notoriety for their role in Donald Trump's presidential campaign. In the Online Marketing Rockstars Keynote | OMR17, Cambridge Analytica CEO Alexander Nix discusses big data, the methodologies and strategies used to identify people's behavior and how data analytics factors into political campaigns. --- #OMR20 is approaching fast! Be part of the OMR Festival 2020, Europe's leading digital marketing event, on May 12 & 13, 2020 in Hamburg, Germany: http://bit.ly/OMRfestival20 --- Subscribe to our OMR YouTube channel: http://bit.ly/OMRyoutube --- At OMR, we view ourselves as a platform for industry professionals by industry professionals. We bring together players from every corner of the digital marketing stratosphere: marketing makers with global visionaries, head hunters with prime candidates and information seekers with experts. Our range of products includes the annual OMR Festival with 52,000 visitors in 2019, big-name international speakers, stellar networking opportunities and unique industry insights, our German and English-language blogs, our weekly podcasts, monthly reports, jobs portal, and many more. Welcome to the home of Digital Marketing: http://bit.ly/OMRhomeEn --- OMR Festival: http://bit.ly/OMRfestival20 OMR Daily: http://bit.ly/OMRdailyEn OMR Events: http://bit.ly/OMReventsEn OMR Podcast: http://bit.ly/OMRpodcastEn --- OMR Facebook: http://bit.ly/OMRfacebookEn OMR LinkedIn: http://bit.ly/OMR-linkedin OMR Instagram: http://bit.ly/OMRinsta OMR Twitter: http://bit.ly/OMRtwitter [Source](https://www.youtube.com/watch?v=6bG5ps5KdDo)

Source quotes

[17:05] Alexander Nix: If there's one singular takeaway from this event that is that data science technology is not a panacea. You cannot make a bad candidate into a good one. You cannot swing an election. [28:02] Alexander Nix: Beyond that you know as long as they're free and fair elections and democracy is being upheld we think it's the duty of our firm to provide candidates with the best possible service and fair representation that's the only way you can have fair elections.

Source quotes

[9:00] Alexander Nix: I've said it before and I'll say it again: I'm very much of the belief that blanket advertising is dead. [19:46] Alexander Nix: We're a long way away from a situation where elections can be won or lost simply on the basis of targeted engagement. [5:26] Alexander Nix: Today we can use big data to understand exactly what messages each specific group within a target audience need to hear way before the creative process is even started. [1:12] Alexander Nix: When you stop for just a moment and think about this, the idea of segmenting audiences say based on gender is very ridiculous.

Original transcript

[0:00] [Music] [0:16] ladies and gentlemen it is a pleasure [0:18] and a privilege to be here in Hamburg to [0:22] speak to you about the evolution taking [0:25] place in marketing and Communications [0:28] from Mad Men to math [0:30] men and particularly in conjunction with [0:34] some of the work that we contributed to [0:36] Donald Trump's presidential [0:40] campaign throughout the presidential [0:42] cycle in 2016 Trump was vocal about the [0:47] fact that it is candidates and not [0:50] technologies that win [0:52] campaigns and in this regard we would [0:55] absolutely agree with [0:57] him that said I'd like to talk today [1:01] about three Technologies three [1:03] methodologies that are shaping the way [1:06] that political campaigns and brand and [1:08] communication campaigns are being [1:12] delivered the first of these is [1:14] Behavioral Science and probably the [1:17] easiest way to explain this is through [1:21] illustration were you lucky enough to [1:24] own a private beach and you wanted to [1:26] stop people from swimming in your piece [1:28] of the ocean you might put up this piece [1:31] of advertising on the [1:34] left this is informational communication [1:38] and seeks to inform [1:40] attitudes conversely you might seek to [1:44] employ the piece of communication on the [1:46] right this is behavioral communication [1:50] and it seeks to probe an altogether [1:52] deeper [1:53] motivation clearly the threat of being [1:56] eaten by a shark is a very compelling [1:59] argument to stop you swimming in the [2:02] sea yet this is extremely difficult to [2:05] do and most advertising and [2:07] communication agencies today still [2:10] segment and Target based on demographics [2:14] geog Geographics uh consumer and [2:17] lifestyle attribution and media [2:20] consumption but when you stop for just a [2:22] moment and think about this the idea of [2:25] segmenting audiences say based on gender [2:28] is very ridiculous the idea that all [2:31] women should get the same message simply [2:33] because of their sex or all [2:35] African-Americans because of their [2:37] ethnicity or all old people or rich [2:39] people or young people because of their [2:41] demographics it simply doesn't make [2:44] sense clearly your demographics your [2:48] Geographics your media consumption your [2:50] consumer and lifestyle attributes will [2:53] influence the way that you see the world [2:56] but as important or possibly more [2:58] important is personality or [3:02] psychographics an understanding of [3:04] personality is key because it is [3:06] personality that informs decision making [3:08] and clearly your decisions uh will drive [3:11] the way that you vote or which products [3:13] and services that you [3:15] purchase so at Cambridge we've rolled [3:18] out a very longform quantitive [3:20] instrument to probe the underlying [3:23] traits that make up your personality we [3:26] use The Cutting Edge in experimental [3:28] psychology this is the ocean five Factor [3:31] personality model ocean being an acronym [3:34] for o openness how open you are to new [3:39] experiences conscientiousness how much [3:41] you care about order and habits and [3:44] planning in your life extroversion how [3:47] socially minded you are agreeableness [3:51] whether you tend to put uh your needs [3:54] ahead of society and Community or vice [3:56] versa and finally neuroticism a measure [4:00] of how much you tend to [4:02] worry why is this important it's [4:05] important because if you're targeting [4:08] audiences purely on demographics you [4:10] might end up sending the same message to [4:12] people who really have very different [4:14] World Views in this example these two [4:18] these two individuals both the same [4:20] gender both the same income both the [4:22] same lifestyle attributes [4:24] broadly but we can see that the [4:26] gentleman on the left here has a very [4:29] very uh conscientious and agreeable [4:32] personality whereas on this side the [4:35] gentleman's very [4:36] extroverted so if we knew that both [4:39] these individuals were in the market for [4:41] an automotive we might want to use [4:43] psychographics to Target and message [4:46] them differently [4:49] specifically a highly agreeable person [4:52] might be more interested in a product [4:55] that had uh qualities that were more [4:58] relevant in today Society such as an [5:01] electronic engine a conscientious person [5:04] could be persuaded by a rational [5:06] fact-based argument such as the one [5:08] that's set out in this piece of creative [5:11] this is in contrast to an extrovert to [5:14] whom we might want to Market uh a sports [5:17] car and to use language and IM imagery [5:20] that evokes the emotion and the [5:22] experience that buying such a car would [5:26] provide the second technology I want to [5:29] talk about is data [5:31] analytics big data is fundamentally [5:34] changing the way that Communications are [5:37] undertaken back in the 1950s the era of [5:40] Mad Men creative was top down that is [5:44] Brilliant Minds came up with Creative [5:46] Concepts and they push them onto an [5:49] audience in the hope that they would [5:51] resonate but today we can use big data [5:54] to understand exactly what messages each [5:58] specific group group within a target [6:00] audience need to hear way before the [6:03] creative process is even [6:05] started so what is Big Data big data is [6:09] really the aggregation of as many data [6:11] points that you can possibly get your [6:13] hands on this includes factual data uh [6:17] demographic data such as age gender [6:19] ethnicity religion and so forth uh [6:22] attitudinal data what car you drive what [6:25] magazines you read what social media you [6:27] engage with uh what church you attend [6:29] what golf club you belong to and so [6:31] forth and of course Behavioral or [6:34] personality data an understanding of [6:36] your world [6:38] view so we can take all these data and [6:41] synthesize them into one database of [6:43] record and use this to build complex but [6:46] very powerful models about the audiences [6:49] we're seeking to engage with and then [6:51] use these models to drive a much more [6:54] individualized uh communic sorry [6:57] communication engagement [7:00] let me illustrate this with a rather [7:01] simple example for a dental product [7:04] toothpaste you can see uh a data [7:07] dashboard uh the engagement map up here [7:10] is divided between those people who are [7:13] more likely to be influenced by an [7:15] economic argument price sensitivity and [7:19] on the other axis we see those people [7:21] more likely to engage with a brand [7:24] preference so clearly the people on the [7:27] right shaded blue um might respond [7:30] better to an offer for a discount that [7:32] could be a positive way to engage them [7:36] conversely the people who are more [7:38] stimulated by brand preference will need [7:40] a persuasive argument based around [7:42] product features such as the ones listed [7:45] here removes stains whitens teeth bad [7:48] breath gum health Etc but this dashboard [7:51] also informs us about which media [7:53] channels different audiences are likely [7:55] to engage with and earning potentials [7:58] therefore spend potenti IAL and basic [8:01] demographics overlaid on this is an [8:04] understanding of Personality or [8:05] psychographics so we can start to Nuance [8:07] our messages uh taking this into account [8:11] so how does this manifest itself in a [8:13] piece of [8:14] creative if the audience we're seeking [8:17] to engage with is largely female 25 to [8:21] 30 single middle inome and so forth that [8:24] needs to be reflected in our creative [8:28] both in imagery in the language that we [8:30] use we can see that this person or this [8:33] group is active on Facebook so that's [8:35] going to be our channel of communication [8:37] most importantly we know they need a [8:39] persuasion message built around the fact [8:42] that uh this toothpaste removes stains [8:46] and finally the personality of this [8:48] group is neurotic so we need an [8:50] emotional message so the fact that [8:52] stains might stop um um before they [8:56] spoil your smile is a very powerful [8:58] message in [9:00] indeed the last leg of the stool is [9:02] addressable adte I've said it before and [9:05] I'll say it again I'm very much of the [9:07] belief that blanket advertising is dead [9:11] the idea that millions of people are [9:13] going to receive one piece of [9:15] communication is something that our [9:17] children will never understand [9:19] communication is getting ever more [9:22] individualized we are moving to an age [9:25] where we are going to have a personal [9:27] relationship with Brands so whether they [9:29] communicate with you through mail email [9:33] text they're going to be writing to you [9:35] as individuals such that a husband and a [9:37] wife in the same household can receive [9:40] different Communications even if they're [9:42] from the same company and on the same [9:45] product Bill Harvey famously said on [9:48] average a 10% increase in purchase [9:51] Target density will generate a 7% [9:54] increase in incremental [9:57] sales we can use digital marketing to [10:01] upload offline Target segments with [10:04] cookies in order to identify and engage [10:08] the densest possible audiences for [10:10] specific [10:11] messages in America we can also take [10:14] offline Target segments and match them [10:17] to set toop box viewing data that is [10:19] cable data um I'm sure you're aware but [10:22] every time you watch a cable television [10:25] program the the the programs that you're [10:27] watching are being recorded and sent [10:28] back to the cable provider and this [10:30] gives us a very clear understanding of [10:33] what different audiences uh watch on [10:35] television we can match these viewing [10:38] habits uh against the audiences that [10:40] we're interested in marketing to such [10:42] that we can purchase inventory in the [10:45] programs that have the highest density [10:47] of the people we want to engage with [10:50] similarly in America we can uh undertake [10:53] hyperlocal television that is buying not [10:55] National broadcast or even Regional [10:57] broadcast but buying by local cable [11:00] provider which allows you to purchase by [11:03] just a handful of zip codes or postcodes [11:06] which allows you to Target very very [11:07] specifically key audience [11:10] groups so if you're in any doubt about [11:13] what I'm saying have a look at where the [11:15] money is going have a look at where the [11:16] big Brands the big advertising agencies [11:19] are investing Walmart and Ford both have [11:22] invested hundreds of millions of dollars [11:25] in the last couples of years in building [11:28] analytics capability I or acquiring them [11:31] Ogie the advertising giant just last [11:35] week launched their Center for [11:37] Behavioral Science in an attempt to [11:39] understand uh the impact of psychology [11:42] on marketing Kroger the the the retail [11:45] giant acquired danami the British [11:47] analytics firm Martin Sorel the the guru [11:51] of advertising for at least four years [11:53] has been talking about the shift away [11:56] from creative and towards science and [11:58] data and again Havas the French [12:01] conglomerate is moving in the same [12:03] direction so just to illustrate how some [12:06] of these Technologies can manifest [12:07] himself uh a few slides on uh the work [12:10] we undertook for the Trump Campaign [12:12] which started back in June [12:16] 2016 when Cambridge analytica my firm [12:19] joined the Trump campaign there were [12:21] only 30 full-time employees on The [12:24] Campaign which is in stark contrast to [12:28] the Clinton campaign campign which had [12:30] somewhere close to 800 people working on [12:33] it clearly um we were an attractive [12:37] proposition we just spent 14 months [12:39] working on the Ben Carson and Ted Cruz [12:42] campaign and had collected a huge amount [12:44] of voter data and research and have [12:47] built an entire technological [12:48] infrastructure which we were able to [12:50] hand over to the Trump team we were [12:54] therefore contracted as an endtoend [12:55] technology provider starting with [12:58] research uh where we were undertaking [13:00] somewhere in the region of 1,500 polls [13:04] each week in each state that is 300 or [13:08] 400,000 polls a month in order to inform [13:11] reporting dashboards to help the [13:14] campaign understand its asset allocation [13:17] and resource allocation and also to [13:19] inform our models our predictive models [13:21] about the audiences we needed to engage [13:24] with these models then inform the [13:26] segmentation which in turn formed the [13:29] engagement we placed over aund million [13:33] of digital buy for the Trump campaign [13:35] and tens of millions of dollars of [13:38] datadriven [13:40] Television so starting with research [13:43] what sort of things did we provide to [13:44] the campaign well these weekly polls [13:47] provided a very powerful insight into [13:50] tracking the electorate this was [13:52] particularly relevant when uh incidents [13:55] took place that needed responding to uh [13:58] particularly good example might be the [14:00] hot mic incident where Trump's uh locker [14:04] room banter was broadcast nationally [14:07] what we were able to tell the campaign [14:09] is that whilst in some states his [14:12] ratings plummeted in others they [14:14] actually increased this allowed the [14:17] campaign to allocate resources in a very [14:19] targeted [14:21] way another type of dashboard is the one [14:24] over here which shows heat maps of [14:27] persuadable Voters that is voters who [14:29] still hadn't made up their decision who [14:32] they were going to vote for and were [14:34] likely targets for us a good example of [14:37] this would be Wisconsin Wisconsin is a [14:39] state that is traditionally a very safe [14:42] Democrat [14:44] seat so much so that the Clinton [14:46] campaign never visited Wisconsin once in [14:49] the entire election but we were able to [14:52] use data to identify that there was very [14:54] large quantities of persuadable Voters [14:57] there that could be influen uence to [14:59] vote for the Trump campaign and so much [15:02] so the Trump campaign had five rallies [15:05] there that five those five rallies [15:08] probably gave Trump contact with some 60 [15:11] or 70,000 voters and when you think that [15:14] this state was won by a margin of only [15:17] 50,000 voters that gives you some [15:20] understanding of how powerful this data [15:23] is predictive models we use a lot of [15:25] research to build models obviously donor [15:28] models of fun fundamentally important uh [15:30] every campaign needs cash to drive its [15:32] media engagement and for the first month [15:35] we were really trying to raise as much [15:37] small dollar donation that is $520 [15:40] donations or less from Trump supporters [15:43] so we need to identify these people and [15:45] then to engage them in the most [15:47] compelling way thereafter we handed this [15:49] capability to the RNC and we focused on [15:53] preference models candidate preference [15:55] models which candidates appeal to which [15:58] voters [15:59] turnout models understanding whom was [16:01] going to vote and who wasn't and of [16:04] course understanding the impact of [16:06] unregistered vote that was going to be [16:08] very [16:09] important finally we started to look at [16:12] issue models predicting which issues [16:14] social and political appeal to which [16:17] members of the target audience which [16:19] voters we actually assigned different [16:21] issues to every adult in the entire [16:23] United States we could then take these [16:26] models and put them into a matrix a [16:29] little bit like the dental health [16:30] example where we can categorize people [16:33] or segment them according to how they're [16:36] likely to behave core Trump supporters [16:38] top right may be uh more susceptible to [16:41] a donation solicitation get out the vote [16:45] people who are going to vote Republican [16:47] but they need persuading to do so [16:49] persuasion audiences people who need [16:52] shifting a little bit from the center [16:54] towards the right once we've identified [16:57] a segment we can then sub segment them [16:59] by the issues that are most relevant to [17:02] them and then start to Target them with [17:03] specific [17:05] messages so how did this manifest itself [17:08] well in terms of digital execution as I [17:11] mentioned we raised nearly $27 million [17:14] just in the first month of the campaign [17:16] and registered some million email [17:19] addresses from supporters event [17:21] promotion we promoted nearly a 100 [17:24] rallies uh and drove at least 35,000 [17:27] people to download the Trump app in [17:31] terms of [17:32] persuasion we delivered a staggering 1.4 [17:36] billion impressions in just four months [17:39] across 4,000 separate ad campaigns this [17:43] increased or had a net increase of about [17:46] 3% on favorability towards Trump and [17:49] finally in terms of absentee ballot and [17:52] get out the vote we delivered about 80 [17:55] million impressions increasing U [17:57] absentee balance submissions by about [18:01] 2% during this period Google undertook [18:03] an independent brand lift study to [18:06] understand the impact of our work we uh [18:09] contacted about 150,000 people and sent [18:13] them highly targeted digital video [18:15] content and this was as opposed to a [18:17] control group who just received generic [18:20] messaging Google reported that the [18:23] people who received targeted [18:25] Communications uh responded 18% in terms [18:29] of the uplift through search volume this [18:32] manifested itself in 11% increase in [18:35] favorability and an 8% in voter intent [18:39] for [18:40] Trump finally we use these data to [18:43] optimize very targeted geographically [18:46] based media buys going after very small [18:49] pockets of trump supporters this [18:51] attracted a huge amount of attention [18:53] because it was in contrast to Trump uh [18:55] to Hillary's approach which was much [18:58] more Broad rush and obviously less [19:03] economical clearly the outcome of the [19:05] election is well known and a lot has [19:07] been written about it but if there's one [19:10] singular takeaway from this event that [19:13] is that data science technology is not a [19:18] Panacea you cannot make a bad candidate [19:21] into a good one you cannot swing an [19:24] election but in political campaigns and [19:27] Commercial and brand camp campaigns [19:29] these sorts of Technologies can make a [19:31] huge difference and will continue to do [19:34] so for many years to come thank [19:38] [Applause] [19:44] you [19:46] um are you saying that without your help [19:50] I mean it didn't matter in the end if [19:51] you helped him or not I would have [19:53] Donald Trump become president even [19:54] without your [19:56] help well obviously I would stand on [19:59] stage and say that but I think the point [20:02] here is that um that elections we're a [20:05] long way away from a situation where [20:08] elections can be won or lost simply on [20:10] the basis of uh targeted engagement um I [20:14] think that in elections that are [20:16] increasingly won by very small margins [20:19] uh technology can play a role and I [20:21] think we saw that manifest itself in [20:23] this last election and [20:26] I guys from your company were already [20:28] here yesterday and the the day before [20:30] and I I already know that but I want to [20:31] ask for for the for the audience um you [20:34] are open to work for any party it's not [20:36] just the Republicans you're open to work [20:38] for whoever asks you is that correct [20:40] well I think most people are unaware [20:42] that we probably work on eight or 10 [20:45] National campaigns for prime minister [20:46] and president each year all over the [20:49] world and as many of these are left of [20:51] center as they are right of Center we [20:53] try and leave our ideological personal [20:55] ideologies at the door in order to [20:57] provide the most profession profal [20:58] service so within a within a framework [21:02] uh we will consider working for most [21:04] mainstream parties and maybe can you say [21:07] a few words before I open up the bar on [21:10] the article that everybody in Germany [21:12] read it was the most read article on [21:13] Facebook [21:16] V was the name of the article by Swiss [21:19] magazine I'm sure you felt the [21:22] consequence of this article how did it [21:24] come about is it correct what's your [21:27] thinking about this article today [21:29] well look I think some of the facts in [21:31] the article were correct I think the the [21:33] journalists tried to connect the dots um [21:35] without certainly without all the [21:37] information at their fingertips um I [21:40] read a translation of it we were still [21:42] you know heavily involved in in wrapping [21:44] up in work in the states so whilst I [21:46] think it had more of an impact in Europe [21:48] uh we were less affected uh in [21:51] America so so it's not is it is it [21:53] correct incorrect what we could read [21:55] there I mean how do you how do you no I [21:58] had it had some some uh enormous [22:01] falsehoods in it but then as I just [22:04] explained to you some of the [22:05] technologies that we were using were [22:07] attributed correctly [22:11] okay so we have elections uh in Germany [22:14] in September do you already work with [22:17] German [22:18] parties uh we're not currently working [22:21] with uh any parties in Germany why not [22:24] so if you if Angela Merkel is calling [22:27] you you can make her GRE [22:30] again look if if Angela Merkel was to to [22:34] to call me I'd be privileged um the [22:37] timelines were complex we'd just come [22:38] out of two years of work in the United [22:40] States uh this is extremely high [22:43] pressure uh long hours and after the US [22:47] election you know we we all took a break [22:49] for a while and and obviously the the [22:51] German elections are upon us so the the [22:55] uh the amount of help that we could [22:56] contribute at this stage would limited [22:59] by that [23:03] timeline um so I understand that as soon [23:05] as you have an ocean profile of a [23:07] customer or potential customer and maybe [23:08] connected with a cookie you can Target [23:10] them but how do you actually get to the [23:12] ocean profile I guess there are kind of [23:14] is kind of a survey a person has to do [23:16] and why should people kind of give you [23:19] those complex information in order to [23:21] you make an ocean profile CU I [23:24] understand that kind of this is kind of [23:25] the basis of uh kind of this granular [23:28] targeting right certainly so there's no [23:30] reason why people should give up any [23:32] data at all and there's certainly no [23:34] obligation to um we've managed to [23:37] collect millions or hundreds of [23:39] thousands and millions I should say of [23:41] ocean profiles in the United States by [23:44] people volunteering this information to [23:46] us we can then match these ocean [23:49] profiles against the data attributes [23:51] that we hold on those individuals who've [23:53] taken these personality inventories and [23:56] we can use that to build a training set [23:58] to to make predictions about the wider [24:00] audience now in other countries we don't [24:02] have this vast Insight we would have to [24:05] collect it and whether people wish to [24:07] give up this information of course is [24:09] entirely up to them but how do you get [24:11] the information and is it kind of a game [24:13] or I mean what is my incentive to give [24:15] you to take to kind of spend five [24:17] minutes or so of my time to give you [24:18] those informations because I guess it's [24:20] there are several questions behind uh [24:22] such model right well I think people to [24:25] undertake different uh social media um [24:28] uh surveys for different reasons I think [24:30] some people are fundamentally interested [24:32] to understand what their personality is [24:34] and what that means um so self-inquiry [24:36] is obviously driving that um I can't [24:40] speak for other audiences but um we've [24:42] had somewhere shy of a couple of million [24:44] people take this test so there's [24:46] obviously a powerful motivation uh or [24:48] many powerful motivations for people to [24:50] do so so it's it's like a Facebook game [24:52] it's like Facebook game where you answer [24:54] questions and you find out things about [24:56] yourself and while the person is playing [24:58] that game you're giving he's giving you [25:00] the the data well they're voluntarily [25:03] giving up their data they're doing this [25:06] in in in you know in full knowledge uh [25:09] of what's Happening Here this is this is [25:10] not anything mackan R toward no no I'm [25:14] I'm not saying that I'm just trying to [25:15] understand that through a Facebook game [25:17] or through a Facebook survey you are [25:19] finding that out okay exactly okay um [25:21] before you go it goes back to you I want [25:23] to just know [25:25] um by the I mean Germany we have very [25:27] strict data rules protection data [25:31] protection uh different legal system [25:33] obviously than in the US would it from a [25:36] from a legal perspective what you were [25:38] able to do in the US would that be [25:40] possible in Germany just on a legal uh [25:44] well at a very high level the the [25:47] culture in in the United States is very [25:49] much opt out whereas in Europe it tends [25:51] to be more opt in in terms of data so [25:54] clearly um you can't just go out and [25:57] acquire data in the same way that you [25:58] can in the United States but you you can [26:02] um collect first party research uh such [26:04] as the ocean research and if people wish [26:07] to volunteer their information such that [26:09] it could be incorporated into marketing [26:11] and Communications then you can begin to [26:14] uh build these sorts of models equally [26:16] there are a number of you know companies [26:18] out there who already have amassed very [26:20] M very large quantities of their own [26:22] data which can then be used and modeled [26:25] for their own internal marketing [26:27] purposes [26:30] I uh I just actually have only one [26:33] question so if it's really true that you [26:37] help Trump get into [26:40] office are you happy putting a [26:48] [Applause] [26:51] mogist a mogist [26:54] buffoon the highest the the most [26:57] powerful job in the world who's now [27:00] up our world are you [27:04] [Applause] [27:08] happy well I think you need to step [27:11] backwards a little bit and understand [27:13] that I don't think it's for people [27:15] sitting in other countries to um to to [27:19] to cast their opinions on the [27:20] self-determination of the United States [27:23] Trump was one of 17 presidential [27:25] nominees the largest field ever and he [27:28] won that nomination fair and square [27:30] through free and fair elections and then [27:33] went on to win the election so if that's [27:36] the will of the people I don't think [27:38] it's anyone's job in this room to to to [27:41] to go against [27:45] that so are you [27:47] [Applause] [27:50] happy my question I'm happy with the job [27:52] that we [27:53] did fair [27:57] enough [27:59] uh I accept that um you said you work [28:02] for Consumer products and you work for [28:04] political parties for you as a service [28:06] company doesn't do any difference if you [28:08] go like for a toothpaste or for [28:11] Trump well obviously working on a [28:15] political campaign by its very nature is [28:17] divisive and we have to be very careful [28:19] that um the candidates or parties that [28:22] we work for um don't cause any [28:25] reputational damage to the company that [28:27] aren't get going to upset our staff or [28:29] employees um and that obviously aren't [28:32] going to upset any of our other clients [28:34] or customers but beyond that you know as [28:37] long as they're free and fair elections [28:39] and democracy is being upheld we think [28:42] it's uh the duty of our firm to provide [28:45] candidates with the best possible [28:46] service um and fair representation [28:49] that's the only way you can have Fair [28:54] elections are there any clients you [28:56] would refuse to work for I'm sure there [28:59] are but um we we try not to to you know [29:02] to go get ourselves into a position [29:04] where we have to say [29:06] no there was no situation where you say [29:10] no right now uh not that I can recall no [29:15] but clearly someone who likes to grab [29:17] women by the is absolutely all [29:19] right just well I can't comment on that [29:23] um I think that was covered in the news [29:24] at the time and I think I've heard both [29:27] sides of the argument [29:32] any further questions from the [29:33] bar that's it um Alex thank you very [29:37] much for doing this thank you very much [29:38] for showing up here for telling us what [29:40] you did I think it's really important to [29:42] understand what's going on I mean [29:44] obviously you can argue left and right [29:48] and right and wrong but in the end what [29:50] matters is to see what's happening and [29:52] what you might be able of doing what you [29:54] explained to us today and we are here to [29:56] provide a platform for content for [29:58] description for learning for information [30:00] and I think you contributed to that [30:02] platform quite a bit with what you said [30:04] and how you explained that to us and I [30:06] think that's very helpful for me and [30:08] hopefully as well for audience so thank [30:10] you very much for coming by thanks [30:16] Phil