One unit is however big you make it

Here is an experiment you can run in a lobby. Leave out a bowl of pretzels with a sign inviting people to help themselves, and watch how much they take. Then, on another day, leave out the same pretzels — same everything — except this time each pretzel is broken in half, so the bowl holds twice as many “pieces.” People take more pieces on the second day, but end up eating less pretzel overall. The size of a piece changed. Their sense of how many pieces make a reasonable helping barely did. Something in us reads “one unit” as “an appropriate amount to have,” almost regardless of how big the unit actually is.

Researchers named this unit bias, and the cleaner demonstrations are almost comic. Put out a jar of M&Ms with a small scoop and people serve themselves a modest pile; swap in a scoop four times larger and, jar unchanged, they take dramatically more — because a scoopful still feels like a scoopful, one unit, the natural stopping point. The unit is doing the deciding. We don’t compute the right quantity from hunger and then act; we accept the unit the world hands us as the definition of “one portion,” and one portion is what feels right to finish. The clean plate, the single serving, the one episode — the boundary of the unit becomes the boundary of the behavior.

Why would a brain work this way? Because deciding exactly how much is exhausting and units are a shortcut that’s usually good enough. Cultures and packagers pre-chew the “how much” question into standard servings, and outsourcing the decision to “one of those” saves you from recomputing it every time. The catch is that the shortcut has no built-in sense of scale. It will treat one enormous unit and one tiny unit as equally “one,” and let whoever drew the unit set the amount. That whoever, in a product, is you.

Once you see it as a lever, it’s everywhere in a product — and it is a different lever than the famous ones. It isn’t anchoring (that’s a number you plant as a reference point) and it isn’t the default (that’s an option you pre-select). Unit bias is about the size of one indivisible chunk your user picks up and finishes. A learning app that ships “lessons” decides, by lesson length, how much a motivated beginner does in a sitting — a five-minute lesson and a forty-minute lesson both get finished as “one lesson,” so the short unit produces more days of showing up and the long one produces more quitting mid-unit. A credits product that sells “a pack” sets consumption by pack size; a to-do app that treats “a task” as the unit shapes how big a thing users are willing to write down. Batch imports, message length limits, the granularity of a checklist, the size of a free tier’s “one project” — each is a unit boundary, and each is silently answering “how much?” on the user’s behalf.

So the build decision is: pick the unit size to match the behavior you actually want, then let unit bias do the rest. Want more frequent, sustainable engagement? Make the unit small enough that finishing one is easy and starting another is optional — a short lesson, a single-card review, a two-minute task. Want a sense of substantial progress per session? Make the unit meatier so that “one” feels like real ground covered. Want a free tier that converts? The size of “one free project” is your knob: too big and one is all anyone ever needs; sized right, one is a satisfying taste that runs out.

Two honest cautions. First, unit bias cuts against you as easily as for you: an oversized default unit can make people over-consume and then feel bad about it, and a product that engineers “just one more unit” against a reward circuit is the same machine as the bottomless bag of chips. The ethical version sizes the unit around a good stopping point and makes stopping the natural thing; the predatory version makes the unit end right where the craving peaks. Second, the effect lives on the boundary being legible — people have to perceive where one unit ends. Blur the units together (the infinite feed, the auto-advancing next episode, the streaming scoop) and you don’t get moderate “one unit” behavior at all; you get consumption with no natural brake. Sometimes the most humane design decision is simply to keep the unit visible so the instinct to finish one and pause has something to grab onto.

The science, to look up: unit bias (Geier, Rozin & Doros, 2006, Psychological Science), and the wider portion-size literature showing that the size of a serving strongly predicts how much people consume. Note the mechanism is about the norm of a unit, distinct from anchoring and from default effects.