Reviewing too early has low marginal benefit because forgetting hasn't yet begun

What it is

Scheduling the next review or reminder while a memory is still solid and hasn't begun to decay at all is largely wasted effort — it buys almost no extra long-term retention, and just re-activates a memory that was already easily accessible. This is another way of stating the massed practice problem: holding the total number of reminders fixed, packing them too densely and too early produces a markedly lower marginal return than spreading the same number of reminders out.

Why it happens

How much consolidation benefit a review produces depends on how much effort that particular retrieval took — if the memory hasn't started decaying yet, retrieval is effortless, and the "review" amounts to nothing more than re-recognizing information that was already fully accessible, without triggering any additional reconsolidation. In other words, the number of reviews isn't the variable that determines the payoff; how much the information had decayed at the moment of each review is. Scheduling a review before any decay has occurred wastes what could have been a productive review opportunity on a moment where the payoff is close to zero.

Studying it

The classic design for testing this compares massed and distributed practice: two groups get the same number of reviews/practice attempts, one packed tightly together (leaving almost no room for forgetting between them), the other spread out over time, with long-term retention measured afterward for both. The result is consistently that, at equal total practice, distributed scheduling produces markedly better long-term retention than massed scheduling — showing that simply adding more reviews doesn't linearly improve memory; whether enough room for forgetting exists between reviews is the variable that determines how much each review is actually worth.

Where it stops holding

The claim that "reviewing too early has low benefit" assumes the goal is long-term retention across a meaningful gap. If a task only needs to stay usable for seconds to minutes (typing in a one-time code just received), immediate repetition is of course reasonable and necessary, and doesn't count as "premature" in this sense. It's also important to separate "correcting a poorly encoded item right away" from "prematurely repeating material that was already learned well" — the former is a remedial intervention and this diminishing-returns conclusion doesn't apply to it; only unnecessary, high-frequency repetition of an already-solid memory runs into this problem.

Applying it

• Audit onboarding flows and notification systems for cases where a reminder fires immediately after a user has just successfully used the target feature themselves — if logs show the user independently and successfully completed the same action minutes earlier, that reminder is very likely a premature repetition with near-zero marginal benefit, contributing only interruption cost with no memory payoff.
• Use "was there a recent record of successful independent use" as the basis for deciding whether to send a given reminder, rather than firing on a fixed calendar cadence regardless — if recent successful use is detected, the scheduled reminder should be deferred or skipped.
• Validation: compare a fixed-cadence batch-reminder scheme against a scheme that automatically defers reminders following detected recent successful use, tracking total reminder count, dismissal/opt-out rate, and eventual long-term retention within the same time window — the deferring scheme should achieve comparable or better retention with fewer interruptions.

Related

• Same group: A6.15.1 The optimal spacing interval should expand as memory strength grows, not stay fixed · A6.15.2 A prompt or review is most effective right before forgetting occurs, but not after · A6.15.4 Reminder and re-education timing can directly borrow the spaced-repetition scheduling curve
• Nearby: A5.16 Habituation and cue wear-out · A6.06 Forgetting curve
• Search terms: massed practice · spacing effect · diminishing returns