Feedback That Arrived 45 Days Late Was Not Feedback

At a glance

• Branch feedback was reaching the bank 30 to 45 days after the customer visit.
• Response rates sat at 8 to 9 percent, so the little that arrived was also unrepresentative.
• Rebuilding collection at the branch lifted response rates to 20 to 25 percent, roughly two and a half times higher.
• Complaint resolution moved inside 24 hours, and closed-ticket rates reached 25 to 30 percent.

The situation

The bank ran a large retail branch network and wanted its branches to function as experience centres rather than transaction counters. It had a feedback programme intended to support that.

The programme spanned roughly 4,000 branches and 80 customer touchpoints, covering about 12 million customers a year, with branch interactions alone running to about 2 million a month.

The programme had two problems that compounded each other.

Feedback arrived 30 to 45 days after the visit it described. By the time a branch manager saw a complaint, the customer had lived with it for six weeks, the staff involved had handled hundreds of interactions since, and nobody could reconstruct what happened.

And response rates were 8 to 9 percent, so the delayed information was also thin.

Feedback that arrives 45 days after the visit is not feedback. It is history. Nobody can act on it, and the customer has long since decided what they think of you.

Why latency mattered more than volume

Most feedback programmes treat response rate as the problem to solve and latency as a technical detail. In a branch network the priority is the other way round.

A complaint surfaced within a day can be resolved while the customer still considers the matter open. The same complaint surfaced six weeks later cannot be resolved at all, only apologised for, and an apology six weeks late frequently reopens the wound rather than closing it.

Latency also decides whether frontline staff can learn anything. A branch manager who receives same-day feedback can connect it to a specific queue, a specific process, a specific morning. At 45 days that connection is gone and the feedback becomes a score to be defended rather than information to be used.

What we changed

Three things, in order of impact.

Collection moved to the branch and the moment. Feedback was captured at the point of interaction rather than through a delayed follow-up, which addressed latency and response rate simultaneously.

Complaints were routed to the branch that generated them, with ownership sitting on the frontline rather than escalating into a central queue. The people who could actually fix the issue were the people who saw it first.

Closing the loop became a tracked frontline responsibility, measured as a completion rate rather than left as an aspiration.

The result

Response rates moved from 8 to 9 percent to 20 to 25 percent. That is roughly a two and a half times increase, and depending on where in each range you measure, between 2.2 and 3.1 times.

Complaint resolution moved to within 24 hours, against a baseline where the feedback itself took 30 to 45 days to arrive.

Closed-ticket rates reached 25 to 30 percent. Worth stating plainly: that is a starting position rather than a destination. It represents the share of raised issues formally closed out by the frontline, measured from a base where no such loop existed at all, and there is clear room above it.

The branch manager does not need a better score. They need to know what happened this morning, while the customer is still deciding what they think of you.

Amitayu Basu, CEO and Co-founder, Numr

What transfers to other branch networks

Fix latency before volume. A larger sample of stale feedback is a larger pile of history. Speed is what converts feedback into an operational input.

Push ownership to where the fix lives. Central complaint queues create visibility without capability. The branch that caused the problem is the only unit that can resolve it.

Measure loop completion, not just score. A CX programme that reports a number but cannot say what proportion of raised issues were actually closed is reporting sentiment, not performance.