Optimize for the user's long-term goals, not dwell time
Aliases: success metrics · engagement metrics critique
What it is
Time well spent holds that a product's success metric should be the attainment of the user's own goals, not dwell time — session length, daily actives, time in app. Dwell time is a means that has been promoted to an end: it cannot distinguish "used well" from "couldn't stop," and both read as the same number.
Why it happens
Metrics define the direction of behavior. When a team's success is dwell time, every design decision optimizes for "longer" even after the user's goal is met — the answer was found, the message was sent, the quarter hour of rest is over. Dwell time rewards overuse and effective use identically, and often scores the former higher, so optimization pressure systematically produces the former. Reorienting success to the user's long-term goal internalizes the product's external costs (time, attention, later regret): dwell time books only the benefit side, and user goals restore the cost side — the difference between the two is the value actually created.
Where it stops holding
"User goals" must be operationalized or it degenerates into new rhetoric. Workable proxies include task completion, post-session regret ("was that use worth it?" sampled ratings), and the share of self-initiated sessions. Not every product should minimize time — deep tools see duration and value rise together; the criterion is alignment between duration and goal attainment, not absolute brevity. Advertising-funded business models stand in structural conflict with this metric, a constraint treated separately.
Applying it
Define a "done" signal for each core scenario: message sent, answer found, relaxation the user would describe as achieved. Make it a first-class metric displayed beside dwell time in experiment reports; A/B tests report both, and the combination "dwell up, attainment down" is treated as a regression, not a win. Build a low-frequency regret sample (a weekly one-question probe) and include it in retention modeling. Verification: compare the resilience of regret-driven retention versus dwell-driven retention under competitive pressure — the former should hold markedly steadier.