── ── Cognitive bias
Voltage Effect
The voltage effect is John A. List's name for what happens to most promising ideas when they scale: they lose voltage — they fail, shrink, or reverse — because the conditions that produced the small-scale win do not survive being scaled. The discipline is to predict scalability before you scale, by interrogating an early win against five vital signs: (1)…
Run Voltage Effect on a real problem
Bring something you're actually deciding — free, in the browser.
How it works
Stop rule: If the pilot's effect size, sample size, and selection method are unknown, stop at vital sign #1 and name the data gap — a result you cannot characterize cannot be scaled, only gambled.
1. False-positive test (is the result real?). Get the effect size, sample size, and how many things were tested. Ask: was it powered? Was one winner cherry-picked from many comparisons (p-hacking)? Has it replicated independently at least once? Gate: a single, unreplicated, underpowered, or multiple-comparison result fails — demand a confirmatory test before proceeding. 2. Population representativeness (who did it work on?). Characterize the pilot participants vs. the population at scale. Were they early adopters, volunteers, a hand-picked site, or a founder-adjacent group? Gate: if the scale population differs materially from the pilot population on any driver of the result, the effect is presumed to shrink — quantify the expected dilution. 3. Situation representativeness (where/how did it work?). Characterize the pilot context — market, channel, season, operator skill, novelty. Gate: if the winning condition (a star operator, a single channel, a launch-buzz moment) will not replicate across all scaled sites, the effect is presumed to shrink. 4. Spillover test (what only appears at scale?). Ask what changes when everyone does this: congestion, market saturation, price/wage response, competitive imitation, cannibalization, general-equilibrium feedback. Gate: if a plausible spillover reverses or materially erodes the effect at volume, model it — do not assume the per-unit pilot effect is additive. 5. Supply-side cost trap (do the economics hold at volume?). Trace marginal cost and the scarce input (talent, supervision, inputs, ops) as volume grows. Does the pilot rely on non-scalable inputs (a heroic founder, a subsidized input, a scarce specialist)? Gate: if marginal cost rises with volume or the scarce input cannot be reproduced, the idea is unscalable even if demand is real. 6. Scale-decision + levers. For each surviving vital sign, name the lever that preserves voltage: high-fidelity implementation (so the scaled version = the pilot), marginal thinking (decide on the margin, not the pilot average), and design-for-scale (rebuild the test at representative population, situation, and cost before betting).
When to use it
- The voltage effect is John A. List's name for what happens to most promising ideas when they scale: they lose voltage — they fail, shrink, or reverse — because the…
When not to use it
When the decision is routine and reversible, applying a formal method costs more than it returns.
Worked example
Groupon and the Merchant-Side Spillover (2008–2012)
Groupon is the marketplace case, and its voltage drop lives in vital sign #4 (spillovers). The consumer side scaled spectacularly — Groupon went from launch in 2008 to one of the fastest revenue ramps in history and a November 2011 IPO. But a daily-deals marketplace has two sides, and the small-scale result on the merchant side did not survive scale: the general-equilibrium effect on merchants only appeared once the model was everywhere.
Install this skill (free, MIT)
npx skills add deciqAI/knowledge-skillsUseful? Star the repo — stars help other builders find it.
Related mental models
Wu wei (无为) — Daoist principle (Daodejing, ~4th century BCE): act without forcing, not without acting.
Zero-sum means total value is fixed — one player's gain is another's exact loss.
We seek and weight evidence that confirms what we already believe. Hunt for what would prove you wrong.
Money and effort already spent are gone. Decide only on the marginal cost and benefit from here forward.