Social norm messaging
Telling people what others actually do moves behaviour toward that number, in both directions.
A descriptive norm states what most people do. An injunctive norm states what is approved. Descriptive norms pull behaviour toward the stated average, which helps if the person is below it and backfires if they are above it. The fix is to pair the number with approval for those already doing the right thing.
How it shows up in software
Norm messaging shows up as counts of other users doing a thing, activity feeds, comparison dashboards and copy such as most teams enable this. It is cheap to add and easy to get wrong, because products usually apply it to behaviours where the majority is not on the desired side.
Using it well
- Only state a descriptive norm when the majority genuinely does the desirable thing. If they do not, use an injunctive or trend framing instead.
- Add approval alongside the comparison so people already above average are not pulled back down.
- Make the reference group narrow and true, such as teams of your size, and be able to show the query behind the number.
- Pair the norm with the action it should trigger, on the same screen, so noticing and doing are one step.
Where it turns manipulative
- Inventing or rounding the number is straightforward fraud, and fabricated counts of viewers or recent bookings have drawn regulatory action in several countries.
- Applying norm pressure to private decisions, such as sharing location or contacts, converts peer comparison into coercion about something the person may reasonably refuse.
- Using a norm to make a minority feel deviant, rather than to inform, is shaming with a statistic attached.
Where you have seen it
OPOWER Home Energy Reports
A bill insert compares the household with similar nearby homes and adds a smiley for efficient households, which is the injunctive correction for the boomerang.
GitHub repository stars and contributor counts
A real, verifiable count of adoption shown where a developer is deciding whether to depend on a library.
Spotify Wrapped percentile ranks
Listening compared against the wider population, computed from actual play data rather than asserted.
What the research says
- Goldstein, Cialdini and Griskevicius, 2008 (Journal of Consumer Research)Mixed evidence
Hotel towel reuse rose when the card said most guests reuse their towels, and rose further when it referred to guests in that specific room.
Widely cited and replicated in direction, though effect sizes in later field tests are smaller than the original.
- Schultz, Nolan, Cialdini, Goldstein and Griskevicius, 2007 (Psychological Science)Well evidenced
Households told their energy use relative to neighbours converged on the average. Below-average households increased their use until an approving note was added.
The boomerang effect is the most useful finding here and the one most often left out when teams copy the pattern.
- Allcott, 2011 and Allcott and Rogers, 2014 (American Economic Review)Well evidenced
Home Energy Reports comparing households with neighbours produced a roughly two percent reduction in consumption across large randomised samples, persisting with repeated mailings.
Very large samples and a small, durable effect. A good calibration point for what a strong social norm intervention actually delivers.
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.