Comparison
The price of doing nothing about Connect leakage
Doing nothing has a price; it simply invoices later. Modelled bands by GMV put shape on it, the normalisation loop explains why it persists, and the counter-case for waiting — pre-PMF platforms with trivial refund counts — stays honestly represented here because credibility requires it.
Bands, modelled
Refund share times typical stranding rates across GMV tiers produces annual dollar bands — labelled estimate everywhere, replaceable by measurement anywhere.
The normalisation loop
Quarter of write-offs → budgeted expectation → leak becomes invisible tax. Each cycle deepens acceptance. Interrupting requires one measurement, not a programme.
Non-dollar costs
Diligence surprises at exits. Seller-trust erosion when they notice first. Engineering archaeology later at higher rates. All real; all avoidable cheaply now.
Common questions
We refund rarely — still exposed?
Rare × high-ticket equals the same dollars; frequency misleads, totals decide.
Is this fear-selling?
It is arithmetic plus an invitation: run the scanner and argue with your own numbers instead.