Organizational evaluation cycles are usually shorter than the time UX harm takes to appear

What it is

Performance, budget, and promotion settle on a quarter or even a month, while spent trust, broken habits, and revoked permissions often take longer to enter countable churn. Evaluation cycle shorter than harm latency means the harm has not appeared at settlement, while the action that produced it has already been rewarded. This is not the general claim that decisions outrun effects. It is that the reward clock is specifically shorter than the harm clock, so short-window harm is institutionally unrecorded.

Why it happens

Evaluation needs a form filled by a deadline. What can enter the form is numbers already realized before the deadline; what cannot is loss that only appears after it. Notification blasts, default renewals, and hard-to-cancel paths book their gains in this quarter’s conversion and their costs in permission revocation and failed renewal two quarters later. The person who collects the reward and the person who pays the bill can even be different cohorts, which stabilizes the mismatch. The organization is not unaware that harm will come; the current clock simply offers no cell for “harm not yet due.” Rational evaluatees therefore choose levers whose latency exceeds the evaluation window, because those levers look like pure gain inside the window.

Studying it

List experience or growth metrics written into evaluation over the past year, estimate median days to appearance for known harms tied to them, and compare with the evaluation window. For harms that have already appeared, look back: had that evaluation already closed, had the reward already been paid. Compare teams on different settlement lengths (monthly versus half-yearly) on their use of high-latency levers. Interviews that ask “which changes would go bad after your evaluation ends” produce the menu of levers the clock protects.

Where it stops holding

Products whose harm is observed immediately (crashes, payment failure on the spot) match the evaluation window, and this claim does not hold. Stretching evaluation without limit weakens accountability. Some harms are always hard to book into personal performance (brand trust) and need organization-level long-window metrics rather than multi-year personal reviews. A short evaluation is not moral permission to use long-latency levers; it only explains why they are systematically chosen.

Applying it

• If a metric in evaluation corresponds to known harm with latency longer than the evaluation window, put that harm in a “not yet due” column on the same form and settle when due.
• High-latency levers (default renewal, notification frequency, cancellation friction) must not pay out on a positive number inside the evaluation window alone.
• Decouple payout from harm review: a failed review claws back or books into the next period; “already paid” is not a close.
• Publish latency estimates to evaluatees so “goes bad after the window” is no longer private knowledge of the designer.

Related

• Same group: Q6.10.2 A long-term holdout that never receives the change is a common way to observe long-term effects · Q6.10.3 Decision-makers systematically underestimate uncertainty in long-term effects · Q6.10.4 Short-term metric gains bought with manipulative design erode long-term trust
• Adjacent: Q6.06 Long term and short term · Q6.09 Metric gaming and proxy distortion
• Search terms: harm latency · evaluation cycle · performance period