📖 Business
Prediction Products
Prediction products are the manufactured goods of surveillance capitalism — algorithmic outputs that forecast what a particular person or group will do in a specific context at a specific time. Zuboff explains that behavioral surplus is the raw material, machine intelligence is the manufacturing process, and prediction products are the finished goods sold in what she calls "behavioral futures markets." These markets operate like financial futures: buyers (advertisers, insurers, employers, political campaigns) purchase predictions about future human behavior — who will click an ad, who will develop a health condition, who will quit a job, who will vote for a candidate. The better the prediction, the higher the price, creating an economic incentive to make predictions ever more accurate by collecting ever more data.
2
Minutes
2
Concepts
+45
XP
1
How It Works
- The behavioral futures market — Prediction products are traded in automated markets. Google's ad auction is the archetype: advertisers bid for predictions about which users will click their ads. The prediction "this user is 73% likely to click an ad for running shoes" is more valuable than "this user might be interested in sports." Precision drives price. Facebook, Amazon, and programmatic advertising exchanges operate similar markets.
- From observation to prediction to intervention — Zuboff identifies three stages of prediction product evolution:
- Stage 1: Observation — Predict behavior by observing past behavior (search history → ad targeting)
- Stage 2: Behavioral modification — Nudge users toward predicted outcomes to increase prediction accuracy (algorithmic feeds, notification timing, social pressure features)
- Stage 3: Guaranteed outcomes — Move beyond prediction to actual behavior modification, selling guaranteed behavioral outcomes rather than probabilistic predictions
- The prediction imperative — Competition among surveillance capitalists drives a race for prediction accuracy. More accurate predictions command higher prices, which funds more data collection, which produces better predictions. This creates a flywheel: data → predictions → revenue → more data collection → better predictions. The competitive moat is not the algorithm but the volume and variety of behavioral surplus.
- Supply chain structure — Raw behavioral surplus flows from users through data pipelines to prediction factories (machine learning systems), then to behavioral futures markets (ad auctions, data brokers), and finally to business customers who act on the predictions. Users are the supply — the source of raw material — not the customers. The customers are the businesses buying predictions.
- The asymmetry problem — Prediction products work because users don't know what's being predicted about them or who's buying those predictions. The market operates in one direction: surveillance capitalists know what users will do; users don't know what's being done with their data. This information asymmetry is not a bug — it's a structural requirement of the business model, because informed users would likely restrict data access.