Chief Information Officer / India
The service decision after the successful screen
- Technology colleagues
- 24
- Three leaders across the current CIO remit
- Retail shops
- 14
- A Jaipur-based craft-supplies network with an online catalogue
- Annual technology operations
- ₹4.1 crore
- ₹1.2 crore approved change authority recorded separately
Decisions & consequences
01Stopping the retry release after the response disappeared
In 2019 I was head of systems and had endorsed an order-application release plan whose acceptance account concentrated on ordinary successful completion. A later rehearsal used sixty synthetic orders and deliberately removed the response after an order had been stored. Eighteen of those orders acquired a second record when the caller retried. The test had not involved live customers or actual payment collection. It showed that a missing acknowledgement could be mistaken for a failed action and that the retry did not reliably preserve the original transaction's identity.
The delivery team wanted to keep the date by asking shop colleagues to check manually before repeating an uncertain request. Operations objected that the colleague might not have enough evidence to know whether the first order existed. I supported a release hold and acknowledged that the earlier test scope carried my approval. Engineers redesigned the handling of the intended repeat, while finance and operations helped define the reconciliation and exception account. We avoided a shortcut that would treat every similar-looking order as a duplicate, because a customer could legitimately request the same goods twice.
The ₹28 lakh, eight-week repair programme examined fourteen defined interruption scenarios. Ten met the acceptance conditions at the review point and four remained held for further work. The release stayed held for the affected flow until its required scenarios were resolved; the ten passing scenarios were not used as permission to ignore the other four. The sixty-order rehearsal and the fourteen-scenario review described different units and were reported separately.
I did not present the eighteen duplicate synthetic records as eighteen customers protected from an actual charge, or assign an invented financial saving to the correction. The decision changed the evidence required before release and made uncertain transaction states an operating concern. It also changed how I reviewed a successful demonstration. I wanted to know what the colleague would see when the expected response did not arrive, and who owned the unresolved case. The delay had a cost, but handing that uncertainty to the shop without a dependable method would not have made the service ready.
Following keyboard access beyond the first reachable page
In 2023 a review examined thirty defined tasks in the online buying journey using keyboard interaction in the agreed test environments. Twelve could not be completed as intended. A customer could enter the catalogue yet become blocked when selecting a colour variant, leaving a dialog or reaching a continuation control. The site looked complete during the pointer-based demonstrations most colleagues had seen. That visual impression did not establish that the same journey was available through another input method.
Commercial colleagues proposed keeping the features and adding a request for customers to telephone when they had difficulty. I did not treat assisted service as a substitute for repairing a journey the enterprise offered online. We retained an appropriate support route while funding work with design, engineering, qualified accessibility reviewers and participants using the relevant methods. The purpose was to understand the task and the failure, not to ask a participant to endorse the whole site after a brief demonstration.
The programme allocated ₹42 lakh over ten weeks. A repeat review of the same thirty tasks found twenty-four meeting the defined conditions and six still requiring work. Those figures described tasks, not people, and the result was not a certification that every page or assistive-technology combination was accessible. The six unresolved tasks remained in the release and service account with owners, while changes to repaired controls were brought into continuing acceptance review.
The decision gave up some feature work in the period so that the existing offer could be used more reliably. I explained that priority to the executive team without promising a revenue increase that we had not measured. The commercial owner retained responsibility for the proposition, and specialist reviewers retained their assessment role. My responsibility was to fund and organise the change, preserve the limits of the evidence and stop treating a polished pointer-based demonstration as a sufficient description of the customer's available route through the service.
Replacing the devices before physical failure became the only trigger
A 2025 review of ninety shop devices found fifty-four approaching the verified end of vendor security support, while thirty-six remained within their documented support period. Many of the fifty-four still performed their everyday functions. Shop colleagues reasonably asked why working equipment should be replaced when other operating needs were visible. The support review concerned the continuing software and security position, not evidence that all fifty-four had failed or that any specific device had been compromised.
Security and procurement colleagues checked the actual product and version information before we proposed a transition. I recommended a ₹1.1 crore, nine-month programme covering replacement, deployment preparation and controlled retirement. The executive team approved the commitment with shop operations involved in the schedule. We did not assume that buying a newer device would by itself make the service dependable: application compatibility, configuration, user preparation and the disposition of the old equipment all needed explicit ownership.
At the stated programme review, forty-two replacements had been installed and twelve remained pending procurement or deployment. The outstanding group retained documented interim restrictions and review dates under security and operating ownership. Those measures did not make unsupported technology risk-free, and the pending twelve were not included in a headline claiming the programme complete. Retired equipment followed the approved data-handling and disposal process; the account distinguished removal from service from verified completion of that process.
The thirty-six devices still in support stayed in the asset review rather than being described as permanently safe. Their status and future dates could change. I did not attach an invented incident-avoidance value to the replacement programme or claim that forty-two installations resolved every security risk in the shop estate. The decision made continuing support a visible part of investment planning, with unfinished deployment and retirement work retained in the account. It also required me to explain the cost in terms the operating team could assess, without relying on a vague appeal to modernisation.