Volume tier
Past $1M GMV, the leak stops being theoretical
Somewhere past seven figures of GMV, two things happen simultaneously: modelled leakage crosses materiality thresholds denominated in salaries rather than rounding errors, and manual audit rituals stop closing the loop because volume outruns attention. Platforms rarely notice the crossing — they notice its symptoms instead: margins drifting below plan, write-off lines growing, finance asking questions engineering cannot answer quickly. This page puts numbers on the crossing.
Exposure bands by GMV
Using published typical ranges against refund shares, modelled annual stranding lands: $1M GMV → $2–8k depending on refund mix; $5M → $10–40k; $10M → $20–80k. Label everything estimate — the ranges assume average mixes, and your partial-refund share or cross-border load can double or halve them. That variance is exactly why measurement beats modelling the moment it becomes free.
- $1M GMV
- $1.9–7.6k
- $5M GMV
- $9.7–38.9k
- $10M GMV
- $19.4–77.8k
The sampling ceiling
Spot-checks fail structurally here, not lazily: leakage concentrates in specific flows (support refunds, partial-heavy categories), so random samples systematically miss the pockets while reassuring everyone. Coverage requires either exhaustive computation — which events make cheap — or acceptance that unknowns remain unknown.
The ops-cost crossover
Quarterly manual rituals cost analyst-days plus recovery-window decay on everything found. Continuous detection trades a fixed subscription for immediacy: findings land inside payout windows where reversals still reach balances. Past $1M GMV the crossover typically arrives within the first quarter of operation — earlier wherever payout cycles run fast.
Recovery-window mathematics at volume
Backlog age multiplies against payout velocity: weekly-payout platforms decay recoverability roughly 4× faster than monthly ones. At volume, backlog aging is not an administrative detail — it is the difference between recovering half and recovering almost nothing from the same finding set.
What FeeGuard does about it
Free detection removes the excuse economy entirely: baseline in ten minutes, then decide with numbers. Monitor tiers price below the analyst-hours they replace at this scale, and the success-fee structure means recovery economics stay aligned whether volumes are 1M or 50M.
Common questions
Is the range real or marketing?
It is a published-range model, labelled estimate everywhere it appears. The free scan replaces it with your number in minutes — which is the entire point.
What does monitoring cost against these figures?
$49/mo Monitor; Pro $99–149. Against four-figure monthly exposure bands, the arithmetic is not subtle.