How a mid-sized NBFC cut notice cycle time by 73% in one quarter
An anonymised NBFC with 12,000 active files moved from a manual notice queue to a configured workflow — and rediscovered three months of lost throughput.
The brief
A mid-sized NBFC with a 12,000-file active recovery book ran their entire notice operation through email and a shared drive. Three people, two regions, one shared inbox. Notice volume was capped at roughly 410 per month, not by demand but by capacity. Files older than 90 days were quietly aging in a sub-folder titled "Pending review", and that sub-folder had not been reviewed in seven weeks.
The diagnosis
We did not start with software. We started with a stopwatch. The team timed every step in the notice flow for two weeks: data extraction, template population, legal review, dispatch, acknowledgment capture, and audit log. The bottleneck was not legal review. It was data extraction — pulling outstanding balance, last payment date, and asset description out of three different systems for every notice. Eleven minutes per file, multiplied by 410 files, multiplied by retries on bounce-backs. The team was running at 78% utilisation just to do data entry.
What changed
Three things, in order. First, a one-way nightly sync from the LMS dropped the data extraction step from eleven minutes to thirty seconds. Second, four notice templates (Section 13(2), 13(4), Section 14 affidavit, and statutory demand) were pixel-mapped to the legal team’s existing wording, so legal review compressed from 90 minutes per batch to 8. Third, dispatch moved to a queue with a two-stage approval that could be cleared from a phone. None of this is exotic. The point is they did all three together, not sequentially.
The result, twelve weeks later
Notice volume per month rose from 410 to 1,540. Average cycle time fell from 11.2 days to under 3. Compliance defects (mostly missing service proofs) fell from 38 a month to 4. Headcount stayed the same. The "Pending review" folder was emptied in week 6 and has not refilled.
What we would do differently
We over-engineered the audit log. The first version captured fifteen fields per notice. After three weeks, the team admitted they only ever queried four. We trimmed it. The lesson, again: model what the operation actually does, not what an external auditor might one day ask for.
We did not realise how much of our team's day was lost to copy-paste until it stopped happening. The notices were never the work — the data was.
Notice management looks like the boring part of recovery. It is not — it is where most operational time leaks out, quietly. Plug it and the rest of the funnel breathes.