BANGKOK — Thailand's central bank is drafting four new regulatory frameworks for buy-now-pay-later lending: minimum user age, product eligibility criteria, spending limits, and interest rate caps. BOT Governor Vitai Ratanakorn has said the rules will take four to five months to finalise, with enforcement expected before the end of 2026. The push follows BNPL accounts growing from 620,000 in 2021 to roughly 5 million today — close to a tenfold rise in five years — against a household debt load that has sat above 86% of GDP for over a year. Officials have specifically flagged purchases as small as a 106-baht bubble tea or a 50-baht plate of khao man gai — Thailand's everyday chicken-and-rice dish — financed over three or four months at 16–18% interest.
There is a second, less-reported detail in this story that matters more than the headline rules. Under current Thai law, BNPL transactions are not classified as "loans" at all. That single definitional gap places them outside the central bank's direct supervision and, more consequentially, outside the National Credit Bureau's reporting requirements. The debt exists. It accrues interest. It gets defaulted on. But until classified otherwise, it is structurally invisible to the system meant to measure a household's total debt exposure — which means every other lender assessing that household's creditworthiness has been working from an incomplete number, not a wrong one. The instrument simply never asked.
The policy story is one thing. The measurement story underneath it is the one worth sitting with.
Every small merchant whose customer base leaned into BNPL over the past four years has been watching a revenue line climb. Foot traffic up, average ticket size up, repeat purchases up. By any conventional reading, that is growth — the kind any owner would frame as evidence the business is working.
It may not be. A revenue chart cannot, by itself, distinguish between two entirely different processes that happen to produce the same upward line: more people wanting the product, and more people being able to finance the product for three or four months at a time. Demand and credit availability are different causal forces. On a simple chart, they are indistinguishable. An XmR chart built on that revenue data — the statistical tool Walter Shewhart and Donald Wheeler developed precisely to separate ordinary variation from a genuine signal — would show a process comfortably in control, trending upward, right up until the moment the financing dries up.
This is the conflation problem identified by Deming's distinction between common-cause and special-cause variation, applied to a credit cycle instead of a factory line. When Thailand's four new frameworks land — minimum age, product eligibility, spending limits, interest rate caps — one of the two causal forces behind that growth line disappears, in some cases overnight. Underlying demand, the part that was never financed, doesn't move. But a merchant who never separated the two signals inside their own numbers will experience the change as a sudden, unexplained collapse, rather than what it actually is: the predictable removal of a force that was never demand in the first place.
This is not a criticism of BNPL providers or of the central bank's intervention. It is an observation about what a growth chart can and cannot tell its owner. A measurement system that records revenue without a mechanism for separating its causes does not warn a merchant when the regulation lands. It only tells them, after the fact, that something changed — at exactly the point when there is no longer enough lead time to act on it.
The fix does not require predicting the regulation. It requires a measurement layer capable of asking, continuously, what is actually driving this number — and flagging the answer before the chart itself goes quiet.
Both problems in this story share the same root: a number that was never required to reveal what was actually inside it. A debt instrument with no obligation to be reported. A revenue line with no mechanism to separate its causes. Regulation can close the first gap. Only a measurement layer built for the purpose can close the second.