Why EMI Bounce Rates Stay High Without eNACH: A Primer For NBFC

Collections reviews at most lending institutions treat bounce rate as a portfolio quality indicator. Rising bounces mean deteriorating credit. Falling bounces mean the book is healthy. The reading is not wrong, but it is incomplete, and it pushes teams towards underwriting when a meaningful share of the failure sits inside the collection mechanism.

The baseline is worse than most portfolios assume

NPCI publishes data on NACH debit presentations and the share that fail. In February 2020, before pandemic distortion entered the series, roughly 31.5 percent of presentations failed in volume terms and 24.9 percent in value terms. That was regarded as a normal month.

Read that again. Close to a third of scheduled auto-debit attempts across the system did not clear, in a period nobody described as stressed. A lender sitting near that level is average rather than failing. A lender materially above it usually has something wrong that underwriting analysis will not surface.

Registration failure is the first leak

Failures counted at the debit stage are visible. Failures at the registration stage frequently are not, because a mandate that never activated produces no presentation and therefore no bounce. The instalment simply goes uncollected and turns up later as a delinquency.

Paper NACH registration involves a physical form, a signature that must match bank records, and a processing cycle measured in weeks. Forms get rejected for mismatched signatures, incomplete fields and account details that do not tally. Through that window, instalments fall due with no mandate in place.

eNACH removes most of it. Authentication happens through net banking, debit card or Aadhaar OTP, and activation typically completes within two to three working days. The mandate exists before the first instalment is due, which is the entire point.

Presentation timing is a lever, not a constant

Many lenders present on the first or the fifth because the loan schedule says so. Borrower balances do not follow loan schedules. They follow salary credits, and in segments such as microfinance and small-ticket personal lending, inflows are irregular and often land mid-month.

Presenting into an empty account produces a bounce, a penal charge and a borrower conversation that damages the relationship without recovering anything. Aligning presentation dates with observed credit patterns, at cohort level rather than account by account, moves success rates without touching credit policy.

Return reason codes are diagnostic

Every failed debit returns a reason. Insufficient funds, mandate not registered, account closed, signature mismatch and technical decline are different problems requiring different responses.

Insufficient funds is a timing or capacity issue and may respond to a re-presentation. Mandate not registered is an operations failure and no amount of retrying will fix it. Account closed requires borrower contact. Institutions that report a single blended bounce percentage upward cannot distinguish between these, so they tend to apply one response to all of them.

Segmenting failures by reason code is usually the cheapest improvement available to a collections function.

Re-presentation needs sequencing

NPCI permits a limited number of retries within a cycle rather than unlimited attempts, and payment service providers are required to throttle execution rates. Retries fired immediately after a failure rarely succeed, because nothing about the account has changed in the intervening minutes.

Spacing attempts around expected credit dates recovers more. Doing that requires knowing when those dates fall, which returns to the earlier point about cohort-level timing.

What this adds up to

Bounce rate is a composite of four things: credit quality, registration integrity, presentation timing and retry design. Only the first is an underwriting matter.

The other three are operational, and they are where eNACH for NBFC collections produces movement inside a quarter rather than across a year. Institutions that separate the four and measure them independently generally find the operational share larger than expected.