Recency Heuristic

What happened most recently feels like what happens generally. Three good weeks and the strategy is proven; one bad month and the marriage is failing; the latest quarter defines the company's health. The mind extrapolates from the freshest data because the freshest data is the most vivid.

This heuristic once made sense: in a stable world, the recent is the most relevant sample. But our world is noisy, and noise means the recent is often the least representative — a random spike mistaken for a trend.

The costs are familiar: investors buying after rises and selling after falls; managers redesigning everything after one bad month; dieters abandoning the plan after one heavy week. Each is a rational creature over-reading the latest page of a shuffled book.

The correction is to ask for the older pages: before concluding anything from the recent run, look at the last ten runs. Is this the trend or the noise? The recent deserves attention — it might be news. But it deserves suspicion too. The last page is one page. The trend lives in the whole book.