Where Thai fintech apps lose users between OTP and first transfer

Most onboarding dashboards in Thai fintech collapse a messy hour into a single bar labelled “abandoned KYC” or “never funded.” Conversion drop-off diagnostics start by refusing that label. Between the OTP screen and the first successful transfer sit at least four different clocks, two possible devices, and a SMS rail you do not control.

The first clock is the OTP itself. If your event fires on “code requested” rather than “code accepted,” a user who switches to the SMS app and returns after forty seconds looks identical to someone who closed the product forever. In one lender funnel we read in studio, 31 percent of “OTP drop-offs” completed a transfer inside four hours. They were not lost. They were reading a text message.

Second devices are not edge cases

A surprising number of first transfers in Thailand begin on a phone that did not request the OTP. A partner shows a QR. A parent finishes KYC on a larger screen. If your identity join key is device-first, the original session dies in the funnel while money still moves. The diagnostic move is simple and unfashionable: count funded accounts, then walk backwards to the requesting device instead of declaring the session dead.

LINE return paths make this worse. A user who authenticates inside LINE OA and then opens the native app can appear as two half-people. Before you rewrite onboarding copy, ask whether you are stitching those halves.

What to instrument before you rewrite screens

We ask teams for five timestamps: OTP requested, OTP accepted, KYC submitted, first transfer initiated, first transfer settled. Settled matters because PromptPay and bank rails can look like abandonment when they are merely slow. If you only store “initiated,” every afternoon outage at a partner bank becomes a product story.

Also store the SMS provider, Android version, and whether the keyboard covered the resend control. That last item sounds petty until you watch replays from a specific OEM skin. Rage clicks on “resend” are sometimes a hit-target problem, not an OTP fatigue problem.

A memo shape that survives a quarterly review

Write the claim as: “Among Android 12 users on provider B, OTP-accept to transfer-initiate median is 47 minutes; iOS on the same provider is 6 minutes. Hypothesis: SMS delivery, not copy. Disconfirm by comparing provider A on the same Android build.” That sentence can be attacked. “Users find onboarding confusing” cannot.

If you want practice doing this on your own export, the Conversion Drop-Off Diagnostics Studio spends a full module on Thai rails. Bring the timestamps. Leave the slogans at the door.