Short-term gains can be paid for with long-term harm

What it is

The same change can improve a metric inside a short window and plant harm that only shows later. That is short-term gain long-term harm. Raising notification frequency lifts opens that week; months later, notification permission is turned off and return visits fall. Hiding the free allowance in a hard-to-find place lifts conversion that quarter; at next year’s renewal, trust is overdrawn. A win in the short window and a loss in the long window can come from one design. A good-looking short window is not a declaration of harmlessness.

Why it happens

Much experience harm is delayed. Interrupted habits, spent trust, and learning replaced by shortcuts do not settle in next-day conversion. Short metrics capture immediate reaction: curiosity, being summoned, consent given before there is time to think. Long harm travels through memory, reputation, and substitutes: people remember being tricked once, and leave the next time another product exists. Interface levers are usually more sensitive to short reactions than to long harm, so short gain becomes the default provable result and long harm becomes an unclaimed residual. If ship decisions recognize only the short window, the harm is booked as “the market changed later” or “a competitor stole them,” not as the original short win.

Studying it

Re-read historical short-window blowouts on a longer window: the same cohort’s later retention, permission revocation, complaints, and voluntary uninstalls. Compare ships that passed on a short window with no long tracking against ships that passed short and were still tracked long; if the latter often reverse, short-gain long-harm is not a one-off. Qualitative follow-up should ask for explanations several cycles later, not first impressions in the ship week. Separate novelty, season, and acquisition quality from delayed harm of the design itself.

Where it stops holding

Not every short gain has a long cost: fixing a true error can improve both windows. Not every long harm traces to one short gain; some is accumulated strategy. If the observation window is too short, “we did not see harm” is not no harm. On life-safety or money products the costs are asymmetric, so the evidential bar for a short-window win should be higher.

Applying it

• For proposals that mainly pull short metrics with summons, scarcity, or defaults, force a column of expected delayed harm and its observation window.
• A short-window win must not fully ship on its own; keep some traffic in a longer window, or set a post-full-ship review date.
• At review, use later metrics on the same cohort; do not excuse an old change with mixed totals of later new users.
• If later metrics worsen, mark that lever type as high-delay risk; the next similar proposal may not report the short window alone.

Related

• Same group: Q6.06.2 Long-term effects require long-horizon experiments · Q6.06.3 Mismatch between decision cycle and effect cycle is a common trap
• Adjacent: Q6.10 Long-term effects versus short-term metrics · Q6.04 Experience and business metrics
• Search terms: short-term gain long-term harm · intertemporal trade-off · delayed harm