Chief Information Officer / India
When a successful component left the task unfinished
Decisions & consequences
01Replacing a nearest-dot recommendation with a road-context question
In 2022 we tested sixty controlled partner-selection scenarios. The application ranked workshops by straight-line distance, and I had supported that simple rule because it was easy to explain. Forty-nine rankings were supported by the operating review;eleven were not. A nearby point could be separated by a divided road or require an access route the calculation had not considered. The exercise did not concern a live emergency or prove that a partner was available or technically qualified for an assignment.
The applications lead preferred to keep distance ordering and add a general caution. He argued that experienced dispatch colleagues already understood the roads and that a more complicated recommendation would cost time and integration work. Arun wanted the relevant road context brought into the review rather than repeatedly reconstructed around an overconfident list. I supported a nine-week programme costing₹47 lakh. Technology worked with operating owners on the recommendation and the conditions that should remain explicit. It did not assume authority over safe routes or the decision to undertake a service.
Eight of the eleven unsupported scenarios were resolved through the revised context, while three remained held for operating judgement. The final account therefore showed fifty-seven supported scenarios and three exceptions in the original set of sixty. That was a test outcome, not a measured improvement in live response time or a blanket approval of every future recommendation. Partner availability, technical capability and the actual assignment still required the responsible owners' decisions.
The change gave up some of the original rule's simplicity and required more careful explanations when the application could not support a ranking. I acknowledged that I had approved the earlier shortcut. The useful correction was to identify what the calculation did and did not represent, rather than make the distance number more precise. I would involve dispatch colleagues in the recommendation's acceptance criteria before choosing the ranking method next time. A system can arrange options helpfully without implying that the first item has already answered every question relevant to a real journey.
Recovering the evidence removed by an image-size default
A 2023 review examined two hundred selected job attachments after the application's standard image compression. One hundred and forty retained the readability needed for the defined service record;sixty did not. The files opened successfully, and their smaller size reduced the burden of transfer and storage. Those technical properties had made the default look effective. They did not establish that a qualified colleague could still use the material for its intended purpose.
The mobile-platform sponsor wanted to keep the compression setting and ask users to photograph closer. Operating colleagues argued that this would transfer the consequences of a system choice to people whose circumstances varied. I supported a₹39 lakh programme to review the processing decision with technical, operating and privacy owners. We examined the evidence the legitimate service task required and the conditions under which it should be collected and retained. The review did not invite unlimited image capture or substitute a photograph for professional diagnosis.
Fifty of the sixty unreadable attachments could be recovered from retained originals. Ten needed a fresh request and remained pending. The resulting account contained 190 usable attachments and ten unresolved cases within the selected two hundred. We did not invent detail that processing had removed or count the intention to request another image as completed recovery. The technical owners still determined whether the material was sufficient for their particular decision. No repair approval or safety outcome is represented by the readability result.
The correction increased the attention and resources devoted to evidence quality instead of treating file size as the only measure of success. It also required a clearer account of when retaining an original served a proper purpose and when another request was appropriate. I would bring those requirements into the initial design review, rather than discover them after a convenient default had become normal practice. An application should help the service obtain usable evidence with appropriate limits. It should not make an attractive transfer result depend on removing the detail the receiving colleague needs.
Removing unrelated location history from the current partner's view
In 2024 we reviewed ninety selected completed-service cases. In sixty-eight, the assigned partner's view was appropriately scoped to the examined service. Twenty-two also exposed unrelated earlier location history. The broader view had been introduced as a convenience for understanding a vehicle's previous support context. The review showed that the convenience could extend beyond what the current assignment needed. It did not establish whether anybody had used that earlier history or justify an assumption of misconduct.
A partner-service manager wanted to retain the history for experienced workshops because it could make a conversation easier. Our privacy and contract owners required a clearer account of purpose and access. I supported a₹54 lakh programme to implement their decisions and test the resulting views. Technology was responsible for the implementation and its evidence. It did not decide that a useful-looking history was automatically lawful to share or select a universal retention period for every class of record.
Twenty of the twenty-two overbroad cases were corrected. Two remained outside the partner view while their review continued. The final account showed eighty-eight supported case views and two held cases within the original ninety. We retained the held status rather than declare the programme fully complete. Restricting the examined views was also distinct from proving that every route to the underlying data had been reviewed. Qualified owners continued the wider assessment appropriate to the actual service and obligations.
Some partners lost a convenience they had become accustomed to, and service colleagues needed a better way to explain what could appropriately be shared for the current task. That was a practical consequence of the decision, not a reason to leave the scope implicit. I would now ask for the relationship between task, purpose and information view before approving a history feature. A record may be useful to the enterprise without being necessary for every counterparty. The system should preserve that distinction rather than allow one broad view to define the organisation's position by default.
Taking a recovery exercise beyond the restored application
The first 2025 recovery exercise brought the application back within four hours. The technical team could sign in and see the restored records. That was a real result, but the exercise had not established whether the telephone path and working dependencies allowed the service team to use the application as intended. I wanted the executive account to retain that gap before the four-hour figure became a claim about recovery of the whole roadside-coordination service.
The infrastructure lead preferred to close the technical exercise and leave telephony testing to another supplier's routine assurance. Operations wanted a joined exercise because the service did not arrive in separate supplier reports. I supported₹62 lakh of recovery work and controlled testing with the relevant owners. The purpose was to examine a defined intake-to-partner-contact workflow, including the supported fallback where it could not be completed. We did not simulate a live emergency through public channels or issue instructions for an actual roadside intervention.
Twenty-eight controlled test calls were examined. Twenty-two completed the defined workflow, and six required manual fallback. Those six remained exceptions in the report. The test did not establish a universal recovery-time commitment, prove that every future interruption would behave similarly or measure a live response or safety outcome. The earlier four-hour application restoration retained its own narrower meaning alongside the later functional evidence.
The joined exercise required time from operating colleagues and suppliers whose individual components could already show favourable technical results. It also made the final account less absolute. I considered that preferable to discovering the missing dependency during a real disruption. My improvement would be to agree the minimum working service and its evidence before the next recovery proposal is funded. A restored component matters, but the enterprise needs to understand how that component reconnects with the people and other services on which its actual work depends.