Nudge theory
Small changes to how choices are presented can move behaviour without removing options.
A nudge: A change to how choices are presented that shifts behaviour without removing any option. alters behaviour predictably without forbidding any option or meaningfully changing costs. Thaler and Sunstein pair it with libertarian paternalism: steer toward what people would choose on reflection, and leave the exit cheap. The label covers a wide range of interventions whose effect sizes differ by an order of magnitude.
How it shows up in software
Product teams nudge constantly: a suggested amount, a highlighted plan, a progress bar, a prompt at the moment of intent. The software version has an advantage over policy nudges, which is that you can measure it on your own users. It also has a hazard, which is that the same machinery serves the quarter's revenue target.
Using it well
- Name the behaviour you want and ask whether the user would endorse it if you described the nudge to them. If not, stop.
- Forecast off field effect sizes, around one to two percentage points, not off headline numbers in papers.
- Run the nudge as a real experiment with a holdout, and publish the null results internally so the team's own literature is not biased.
- Keep the alternative one click away and equally legible. A nudge that hides the other option has become something else.
Where it turns manipulative
- Nudging toward what the business wants while calling it help for the user is the most common failure, and it is hard to detect from inside the team.
- Stacking many small nudges on one decision adds up to coercion even when each one is defensible alone.
- Using nudge language to avoid fixing the underlying product. If people abandon a flow because it is confusing, a prompt is a patch on a wound.
Where you have seen it
Duolingo streak
A daily counter and a repair option that make the cost of a missed day visible, with nothing lost but the number.
UK HMRC tax reminder letters
Behavioural Insights Team trials added a line saying most people in your area have already paid, which raised on-time payment.
Google Maps eco-friendly routing
The lower fuel route is shown first when travel times are close, with the faster route still one tap away.
What the research says
- Thaler and Sunstein, 2008 (Nudge)Well evidenced
Defines the concept and collects the case studies that made it policy currency, from automatic enrolment to cafeteria layout.
The book is a synthesis and an argument. Several of its illustrations rest on single studies that have aged badly.
- DellaVigna and Linos, 2022 (Econometrica)Well evidenced
Analysed 126 trials run by two US government nudge units covering over 23 million people. Average effect was about 1.4 percentage points, against roughly 8.7 points in comparable published academic work.
The gap is publication bias and selective reporting, not fraud. Nudges work, and they work about six times less than the literature suggests. Plan your forecasts off the field numbers.
- Mertens et al., 2022 and Maier et al., 2022 (PNAS)Contested
A meta-analysis reported a moderate average nudge effect. A reanalysis correcting for publication bias found no evidence of an effect once small-study bias was accounted for.
This exchange is the honest state of the field. Specific, well tested nudges such as automatic enrolment and reminders hold up. The generic claim that nudging works does not.
Grades are a judgement about the evidence, not about how useful the idea is. Plenty of contested effects are still worth knowing, as long as you do not cite them as settled.