Framework · Behavioral science
Realistic effect sizes for behavioral levers
Single behavioral tweaks usually move a metric by low single digits. Across 126 nudge-unit trials (23 million people) the average lift was 1.4 percentage points, about 8% relative, versus 8.7 points in journals; and most program effects fade once the program ends. Plan bundled changes and durability checks.
Plain label: what to expect from a single changeAlso known as: nudge units, effect size anchors, megastudy, impact anchors
The anchors printed on every experiment card.
- Nudges at scale: +1.4 pp (2022) on average across 126 RCTs from two US nudge units (about +8% over control), against +8.7 pp in published papers; roughly 70% of the gap came from selective publication and low power Evidence grade A. Robust evidence: replicated, peer-reviewed research..
- Bias-corrected meta-analyses of choice architecture shrink average effects toward zero; defaults remain the main exception Evidence grade C. Limited evidence: few or small studies..
- In the 24 Hour Fitness megastudy, about 45% of 54 four-week programs raised gym visits during the program, but only about 8% left a significant effect after it ended, and forecasters could not pick the winners Evidence grade A. Robust evidence: replicated, peer-reviewed research..
- A large replication project found replication effect sizes about half the original ones Evidence grade A. Robust evidence: replicated, peer-reviewed research..
How it plugs in. Cards show an impact band (S under 3%, M 3-10%, L over 10% relative), a prior range and the evidence grade, never a point forecast. Small products should test bigger, bundled changes. A single tweak claimed to lift more than about 20% triggers a bug check before anyone celebrates.
Watch-outs. Book headline numbers are single-study effects ("in the original study"), not forecasts.
Sources
RCTs to Scale: Comprehensive Evidence From Two Nudge Units · DellaVigna and Linos, Econometrica, 2022
Evidence grade A. Robust evidence: replicated, peer-reviewed research.Megastudies improve the impact of applied behavioural science · Milkman et al., Nature, 2021
Evidence grade A. Robust evidence: replicated, peer-reviewed research.Estimating the reproducibility of psychological science · Open Science Collaboration, Science, 2015
Evidence grade A. Robust evidence: replicated, peer-reviewed research.
Book content is paraphrased, never quoted at length. Last verified October 24, 2026.
Your product
Find the stage your product is stuck in
Gamify It reads your product, picks the one move that matters at its stage, and sizes the test to your traffic. Connecting is free; a Pro subscription starts it.