Many Indian lenders measure almost every aspect of putting together a loan file except one thing: how many times the same person had to touch it before it sanctioned. Speed, conversion, approval rate, cost per file — every standard origination metric covers some version of these.  

What's missing is a measure of rework, the number of times a file gets handled before it sanctions. In quality frameworks borrowed from manufacturing and BPO operations, this is the territory of First Time Right (FTR) and First Time Not Right (FTNR); measures that track whether a task was completed correctly on the first attempt or had to be redone. 

The reason it has stayed off most dashboards for so long is straightforward: rework has been treated as a normal, expected part of processing a loan file. 

This is not anyone's failure in particular. It is how the process has always run across segments. Nobody measures it because nobody has thought of it as something to be measured.

For much of the last decade, accepting rework as inevitable was a defensible choice. Financial institutions operated with enough margin between deposits and lending yields that the cost of touching the same file multiple times could be absorbed.

Operational inefficiency was real, but it was not a binding constraint on profitability. That's no longer the case. 

Two clocks, one dashboard 

Turn-around time and rework are not the same thing measured differently. They can move in opposite directions, and that is the part most operating reviews miss.

Take parallel processing, for example. A lender takes a five-day sequential workflow — RM, then credit ops, then credit manager — and rearranges it so the three steps happen at once. Calendar TAT collapses from five days to three. The customer sees a faster lender, the dashboard records a win, but the person-hours spent on the file have not changed. They have only been compressed into a shorter window. If anything, when the file gets sent back, it is now more expensive, because three people have to redo their parts instead of one. 

A lender optimising only for TAT will find many ways to make the customer's clock look faster while quietly raising the per-sanctioned-loan cost. The dashboard cannot show what it was never asked to measure. 

Why the standard set looked the way it did 

The metrics on every loan origination dashboard today were built for a different problem. A decade ago, the question every lender was trying to answer was whether they could acquire enough customers and convert enough of them to grow. Approval rate, drop-off rate, cost per file — all these measure different parts of that funnel.  

But the question has changed. The funding side numbers tell the story: 

• Bank net interest margins fell by 25 basis points to 2.89% year-on-year in Q1 FY26, driven by lower deposits, rate cuts, and competition in corporate and home loan segments. 

• NBFC home loan growth, historically one of the more resilient segments, is expected to slow to 12-13% in FY26 from 14 % the previous year, with CRISIL pointing to aggressive pricing by public sector banks pulling customers through balance transfer cases. 

• Even at a lender as operationally efficient as Bajaj Housing Finance, NIM dropped from 4% in Q4 FY25 to 3.8% in Q4 FY26.  

When the funding side gives a lender less room to work with, the cost side has to pull its weight. And the biggest cost nobody is measuring is rework — every file that lands in FTNR territory rather than FTR.

The cost of running rework as a silent process 

The most direct way to see what FTNR is costing is to follow a single file through the system.

A customer applies for a secured loan. The RM collects the document set, runs the initial checks, and forwards the file to credit. Credit comes back with a query: the GST returns don't match the bank statement turnover. The RM goes back to the customer to ask for an explanation, which takes two days. The customer sends a clarification, which raises a second question about a related-party transaction. The RM goes back again. By the time the file reaches the credit manager for sanction, it has been touched eleven times across three roles. The TAT clock will record this as a twelve-day file. The dashboard will compare that to the lender's target and either celebrate or escalate. Neither the dashboard nor the operating review will note that the same file was handled eleven times.

Now multiply this by the file volume of a mid-sized NBFC. A lender doing 12,000 sanctioned files a month across MSME and secured products, with an average of eight role-touches per file when the target should be three, is absorbing roughly 60,000 surplus role-touches every month. Each of those touches has a cost — the RM's time, the credit officer's review, the credit manager's attention. None of it shows up as a measurable inefficiency anywhere on the dashboard, because the dashboard was not asked to look for it.

According to BCG research, banks with fragmented workflows suffer from excessive handoffs, unclear accountability, and limited real-time visibility into applications, with each manual intervention raising the probability of errors, rework, and inconsistent credit decisions. It further recommended that lenders should measure success not just by volumes, but by TAT consistency, rework rates, and early delinquency trends.

What changes when rework comes onto the dashboard 

The decisions every credit and operations head is already making get sharper when FTNR becomes part of the calculation.

Take the file we walked through above. With this measurement on the dashboard, the eleven touches become a number the lender can track, target, and reduce, the same way every other operational variable became improvable once it was measured.

The discrepancy between GST returns and bank statement turnover, which today triggers a back-and-forth between credit, the RM, and the customer, becomes a candidate for capture at the point of sourcing, before the file enters the credit queue. Document gaps that today are discovered three roles deep into the process become gaps that can be flagged at submission.

The credit manager's time, which is spent on files that should have been clean before they arrived, becomes available for the files that genuinely need senior judgment. In other words, more files clear FTR, fewer fall into FTNR, and the operations stop paying for the same work twice.

The window before the standard catches up 

KPI standards do update. They tend to update slowly, usually two or three years behind operational reality, and the lenders who begin measuring something before it becomes industry standard tend to spend that gap building muscle the rest of the market will need to develop later.

This is the problem we have been thinking about. We built Atlas Origin, an agentic AI platform, to sit at the point of capturing, validating, and structuring documents before they ever land in the credit queue, so the file gets to the credit manager first time right . Early numbers showing a 95% first-time-right rate signal what operations teams can do once they stop treating rework at an unavoidable cost.

The cost of rework was always there. It is only now becoming something a lender can no longer afford to leave off the dashboard, because the room to ignore it has gone.

Until next time,

Srijan 
Co-founder 
FinBox