Do not start from the dashboard average
The provider shows one percentage. Inside it sit Visa and Mastercard cards, a wallet, a transfer and retries. On top of that, the deposit and the payout often travel different corridors. An average of 86% can mean the local method sits at 94% while a card from a foreign BIN falls below 60%. Until you split the number by method, country, BIN range and decline code, every PSP change is a lottery.
The second mistake is mixing new volume with recurring. A player who has already passed 3-D Secure and holds a token behaves differently from a first deposit off an affiliate campaign. If marketing has just released cheap traffic, approval will fall regardless of routing.
Read decline codes, not the mood of support
Support will hear “card declined”. The log holds do_not_honor, insufficient_funds, 3ds_failed, restricted_card or an MCC mismatch. Each of those codes pulls a different lever. Restricted and MCC are a conversation with the acquirer. 3ds_failed is friction and a device fingerprint. Insufficient_funds is not a payments problem. It is an offer problem, or a limit the cashier did not show earlier.
We collect codes from 30 days and weight them by volume, not by ticket count. Only then do we decide whether to move 3-D Secure, the BIN list, the method order, or the contract with the provider. Changing PSP without that table moves the same drop onto a new invoice.
3-D Secure: friction you can measure
On a regulated market 3-D Secure is often not optional. You can still measure the step where the player drops off: challenge, timeout, no banking app, a bad descriptor. Sometimes the culprit is a descriptor that does not resemble the brand from the landing page. Sometimes the cashier calls 3DS on an amount the local bank always sends to challenge, while a lower threshold would have gone frictionless.
Descriptor and threshold
The descriptor must match the brand the player saw in the ad and on the site. The amount threshold is agreed with the acquirer, not set “by feel” in the back office. We test it on a sample of BINs from the main geo, not on the team’s own cards.
Failover the cashier does not hide
The most expensive approval drop is a decline with no second chance. If the card fails and the wallet or the transfer sits in an “other” tab, some players close the page. The cashier layout has to show the next lawful method for that geo at once, with a clear message that the first route did not go through. This is not “more buttons”. It is one decision: which method is plan B in this market.
The payout is approval too
Operators count the deposit and forget that an unpaid amount comes back as a ticket, a chargeback or a poor rating in comparison sites. Time-to-payout and the share of payouts inside the SLA are part of the same conversation. Often the deposit goes by a local method and the payout by a card the acquirer does not want. Then “high approval” on the way in hides a bottleneck on the way out.