Chief Operating Officer / India
The exceptions that changed the delivery account
- Operations colleagues
- 780
- Six operating heads across centres and client sites
- Annual operating expense remit
- ₹82 Cr
- Delegated expense, distinct from company revenue and P&L
- Years in records operations
- 39
- Reprographics, conversion, quality and delivery leadership
Decisions & consequences
01Reprocessing a delivery that had already reached its image count
A conversion programme had reached its planned total of 740,000 images. The production report suggested that the capture stage was complete, but a quality review found a recurring problem with faint annotations on part of the source material. Eighty-six thousand images required reprocessing. The files existed and their identifiers were present; the issue was whether details relevant to the agreed record standard remained legible. Counting the files again would not answer that question.
The delivery manager argued that the client could receive the full set on schedule while corrections followed. The commercial team worried that admitting the problem before dispatch would delay acceptance of work that was otherwise usable. I asked for the affected population to be identified and proposed a revised delivery plan that made the correction explicit. We agreed with the client to a three-week extension and accepted ₹29 lakh in additional operating cost. The decision distinguished our capture defect from limitations already present in the originals, which needed a different explanation.
Teams adjusted the capture setup, reprocessed the identified material and reviewed it against the agreed conditions. The operating report continued to show the original population and the rework separately. We did not add repeat images to the accepted total or describe the expense as a productivity investment that had already paid for itself. The corrected delivery addressed the identified legibility issue; it did not certify the engineering statements on the drawings or establish that no future user could discover another defect.
The choice protected the meaning of acceptance at a cost to margin and schedule. It also revealed that our earlier source sampling had not adequately represented the faintly annotated material. I would now bring those characteristics into the preparation review sooner. The lesson was not that every difficult source can be made perfect, but that a known capture problem should be described before a customer is asked to rely on a completion claim.
Holding intake until the source list matched the receipt
A client dispatch list described sixty-eight boxes of industrial records, while the receiving team counted sixty-seven. The transport paperwork repeated the expected quantity and did not settle the difference. A supervisor proposed starting preparation on the boxes present and reconciling the final item later. That would have kept the line occupied, but it risked creating a chain of handling records that assumed the original receipt was complete before anyone had established that fact.
I authorised a pause on the affected intake and asked the client and our receiving team to reconcile the available identifiers and dispatch evidence. We did not announce that a box had been lost or presume that it was on our premises. The distinction mattered: an expected item and an item accepted into custody were different states. The account team objected to the effect on the schedule, while operations faced idle capacity that could not immediately be reassigned to another client's material.
The remaining box was found in stock retained at the client's site. It had appeared on the planned dispatch list but had not been sent with the other sixty-seven. The records were corrected to reflect the actual transfer, and intake resumed after five days. The pause cost ₹7 lakh in standby expense. We retained the reconciliation with the job record, so that a later reviewer would not have to infer the explanation from a corrected total alone.
This was a resolved receipt discrepancy, not a recovery of material from our custody and not evidence that every future dispatch would be accurate. I would improve the process by agreeing the dispatch confirmation point earlier and separating planned quantities from confirmed handovers in the intake pack. The cost of the pause was visible and avoidable in principle. Concealing the mismatch would have been easier that week, but would have left the company unable to give a reliable account of the source material it claimed to hold.
Refusing to infer which drawing revision was current
In a 12,000-record sample from an industrial drawing archive, metadata reviewers identified 720 items whose revision status could not be determined confidently from the supplied material. Some had similar identifiers with conflicting dates; others carried annotations that did not clearly establish approval. A production lead suggested using the latest visible date to complete the status field. That would have produced a tidy index, but it would also have turned an uncertain source into an apparently authoritative classification.
I asked the team to retain an explicit unresolved status and route the questions to the client's designated record owners. The client programme manager was concerned that exceptions would make the archive look unfinished at an internal review. Our commercial lead wanted to avoid expanding the work. We agreed on a defined clarification exercise, with ₹11 lakh additional operating expense and nine extra working days, rather than allow the metadata team to make an engineering judgement it had not been authorised or equipped to make.
The client's review resolved 540 of the 720 questions. The remaining 180 stayed visibly unresolved, with the relevant source references and the question requiring an answer. The other 11,280 items in the original sample had not required that clarification route. We kept those categories separate in the report. A resolved classification meant that the designated owner had provided the basis for the metadata entry; it did not mean that Thervani had certified the drawing or approved its use in an engineering task.
The resulting archive was less cosmetically complete than an index with every field filled, but its limits were more honest. I would establish the clarification capacity and response timetable with the client before a similar programme begins. Exceptions are not always a surprise; historical records often contain ambiguity. The operating failure would be to discover that ambiguity and then erase it from view because the schedule has no category for a question that only the owner can answer.
Testing retrieval before calling the index usable
A client's converted archive had the required metadata fields populated, yet users said it was difficult to find the intended record. The delivery team could demonstrate that a search returned results, which was narrower than showing that the right item appeared first for a practical query. I asked for a defined retrieval exercise with client users. In an initial set of 400 queries, 296 returned the intended item first. Similar identifiers and inconsistent naming conventions accounted for much of the difficulty observed.
The project manager believed the original scope had been met because field completion was high. The client expected a usable archive and saw the extra work as part of that promise. We reviewed the contract and agreed a bounded improvement exercise rather than leave either interpretation implicit. The company allocated ₹24 lakh to normalise 160,000 metadata entries and clarify relationships between related records. The work did not alter the underlying drawing content or decide the revision status of unresolved items.
In a later set of 400 queries designed to cover comparable tasks, 368 returned the intended item first. The query sets were not identical, so we reported the figures as two defined tests rather than a universal measure of search improvement. Thirty-two later queries still needed further navigation or review. The result did not establish that every possible user term would work, nor that a quickly retrieved record was suitable for an engineering decision. Those limitations remained part of the handover.
The case changed how I discuss acceptance with commercial and delivery colleagues. Populating fields is an activity; supporting a useful retrieval task is a service outcome that needs its own definition. I would agree that definition before conversion begins, including who supplies representative queries and what remains outside scope. A late usability discussion can be resolved, but it consumes money and trust that a clearer opening commitment could have preserved.