Habits and Friction

What a habit is

A habit is a response that a repeated context triggers with little deliberate thought. A 2016 review paper on the psychology of habit describes habit automaticity as involving activation by recurring context cues, and calls habits "the efficient, default mode of response". The same review cites experience-sampling research in which about 43% of actions were performed almost daily and usually in the same context. Much of daily life, on that evidence, runs on routine rather than fresh decisions.

This is why good intentions are a weak plan. We think intention starts a behavior, while context and ease keep it going. That division of labour is worth designing around, because the context is the part we can change.

How long a habit takes

The most quoted number comes from research at University College London. A summary of the study reports that participants chose a daily health behavior and were tracked for 84 days, and that it took on average 66 days for the habit to form. One person reached peak automaticity in 18 days, while another was forecast to need up to 254 days. The British Psychological Society's digest adds that a single missed day had little impact on later gains, although repeated misses lowered the maximum reached.

The UCL summary treats the popular figure of 21 days as a myth. We would go one step further and say that 66 days is not a deadline either. A critical review of the study notes that it involved 96 volunteers, mostly graduate students, that only 82 supplied enough data and only 39 produced good model fits, that the data were self-reported, and that the scale measured automaticity rather than habit itself.

Friction is the lever we can see

Habits are favoured because they are easy. The same review notes that habits are likely to be favoured "due to the ease with which they can be performed compared with alternatives", and that switching costs discourage people from deviating from them.

The Behavioural Insights Team calls the small obstacles "friction costs": the often neglected, sometimes invisible actions that act as barriers. Its EAST report gives examples. Placing meat substitutes beside meat products raised their sales by 171% in a Belgian supermarket. Making renewable energy the default in Switzerland raised uptake from about 3% to around 80% of households. UK pension auto-enrolment lifted private sector workplace pension coverage from 42% to 85%. These are the report's figures, drawn from different settings and designs, so we read them as consistent in direction rather than as a single effect size.

A worked example: friction compounds

Illustrative, not real data. Suppose a morning routine has five steps, and each step has a 90% chance of being completed on a given day, independently of the others. The chance of completing the whole routine is 0.9 x 0.9 x 0.9 x 0.9 x 0.9 = 0.9^5 = 0.59, or about 59%. Remove three steps by preparing the night before, and the chance becomes 0.9^2 = 0.81, or 81%.

No single step looks costly. Together they decide whether the routine happens. The numbers are invented, and real steps are rarely independent, but the structure is the point: small hurdles multiply.

What the evidence does not show

It does not show that any fixed number of days makes a behavior permanent. It does not show that removing friction works for every behavior or every person. The habit study used behaviors people chose themselves, mainly eating, drinking and exercise, so it says little about complex or social behaviors.

Common mistakes follow. People count days instead of watching whether the context cue is stable. They treat one missed day as failure, although the study found it had little effect. They add motivation when removing a step would work better. And they design for their best day rather than their worst.

Going deeper

Plain-text formula for a chain of steps: Probability of completing the routine = p1 x p2 x ... x pn, where each p is the chance of completing that step, assuming the steps are independent.

One well-studied tool is the implementation intention, a plan in the form "If situation X occurs, then I will do Y". A meta-analysis summarised in a university repository record reports findings from 94 independent tests showing a positive effect of medium-to-large magnitude (d = .65) on goal attainment. We think the format works because it ties a behavior to a cue, which is the same ingredient a habit relies on, but that is our reading and the meta-analysis does not test it directly.

To read the original habit study yourself, look for how many participants produced usable model fits, how automaticity was scored (the critical review reports a seven-item scale scored 0 to 42), and how far predicted times extend beyond the 84 days observed. To read the Behavioural Insights Team figures, check whether each example had a comparison group.

We have linked every source we used. If you think we have read one wrongly, we would welcome the correction.