Christopher Wylie describes in detail how Cambridge Analytica used ML and Facebook data to micro target US voters, creating 253 predictions per profiled record
Whistleblower Christopher Wylie explains the science behind Cambridge Analytica's mission to transform surveys and Facebook data into a political messaging weapon
Context & Ripple Effects
This piece is the technical sequel to the March 2018 bombshell, when documents sourced by Christopher Wylie revealed the harvesting of personal information from 50M Facebook profiles to build psychological profiles of US voters. Where that coverage established what was taken, Wylie's new account explains how it worked: surveys and Facebook data fed into machine-learning models producing 253 predictions per profiled record.
The detail matters because it converts an abstract privacy scandal into a documented methodology. Earlier reporting already tied the operation to Steve Bannon, who approved the firm's 2014 Facebook data collection according to Wylie, and to voter suppression campaigns targeting African-American voters with disengagement tactics. A precise description of the modeling pipeline gives regulators and platforms something specific to police rather than a general allegation.
First-order effects
- Cambridge Analytica's core sales claim — psychographic microtargeting at scale — is now described in operational detail by its own former employee, hardening the case facing the firm and Facebook from lawmakers and the FTC.
- Facebook's data-access practices move from reputational damage to a concrete policy problem: the article documents exactly which inputs (surveys plus profile data) powered the models built on its platform.
Second-order effects
- Rival political-data firms face forced scrutiny of their own targeting claims, since Wylie's account sets a public benchmark for what 'data-driven campaigning' actually entailed in 2014–2016.
- Platforms and app developers supplying demographic data to third parties come under pricing and contractual pressure as advertisers and campaigns reassess partners exposed to the same harvesting pattern.
Third-order effects
- If the pattern holds, political advertising becomes a regulated category distinct from commercial ads, with disclosure rules for the data sources and models behind voter targeting — the direction later leaks reinforced when 100K+ leaked Cambridge Analytica docs showed the operation spanned 68 countries.
- Social platforms structurally tighten third-party data access, trading the open-API growth model of the 2010s for walled, audited data sharing.
The trend: Political persuasion is shifting from broad messaging to individually modeled targeting, and each whistleblower-level disclosure pushes regulators and platforms toward treating voter data as infrastructure requiring audit rather than a free input.