Chief Product Officer / India
The useful step was the one that returned control to the customer
- Product employees
- 29
- Four heads; professional insurance decisions separately owned
- Annual product-function expense
- ₹9.6 crore
- Distinct from company premium and insurance commitments
- Service research and product responsibility
- 28 years
- Customer choice, accessible timing and reviewable changes
Decisions & consequences
012019: the optional extra had already been chosen by the interface
I approved a quotation design in which an optional extra was selected by default. In a fictional illustration, the base annual amount was ₹18,000 and the optional addition ₹2,400, producing ₹20,400 when selected; these invented research amounts do not describe an available policy. A thirty-six-person task exercise found that twenty-four participants treated the extra as part of the base proposition. Twelve recognised it as optional. The apparent completion path had concealed a decision the product should have made explicit.
The design team proposed keeping the default and enlarging the explanatory text. I withdrew that version and allocated ₹31 lakh over eight weeks to revise the choice and review experience. The new design asked the person to select or decline the extra and showed the corresponding total before continuation. Pricing and insurance owners retained the underlying proposition; product management changed how the choice was made, not the amount or coverage terms. The previous design did not go to wider release through my original approval.
In a repeat exercise with the same thirty-six participants, thirty-one could explain the optional status and the resulting total, while five still needed assistance. Familiarity with the earlier task limited what we could infer about improvement. We retained the remaining difficulty in the release account and continued refinement rather than claim every customer would now understand. No sales uplift, complaint reduction or recovered premium is invented from the research. The costs and unresolved evidence stayed beside the revised design.
The important correction was to my own definition of a good journey. I had treated an already-selected option as a convenient starting point. The research showed that the starting point could become the decision in the customer’s mind. A shorter route was not a sufficient reason to preserve that behaviour. I now ask which choices the interface has already made before the person reaches them, and whether the customer can understand the commercial consequence of choosing differently.
2023: a session ended before assisted users could finish
Thirty-two assisted-user sessions examined a defined application task. Participants used screen readers or needed additional time for motor interaction. Twenty completed the task within the original session; twelve encountered a timeout before completion. The proposed response was to shorten the form further. That might help some people, but it did not address how the time limit was communicated or how work could be recovered when a person needed longer. The review concerned a selected exercise, not a measured failure rate across the insurer’s customers.
I recommended that the product account include timing and recovery as requirements. We allocated ₹42 lakh over twelve weeks to a usable warning, a way to request more time and a protected route back to the saved draft. Security and accessibility specialists reviewed the design for the actual environment. We did not remove all session controls or make a universal compliance claim from one feature. The purpose was to avoid treating the loss of a person’s work as an acceptable side effect of a convenient default duration.
In a subsequent exercise with the same participants, twenty-seven completed the defined task and five still required an assisted route. The repeated participants and revised support conditions limited any causal or population inference. Those five cases remained visible, and the product continued to provide an appropriate assisted path while further improvements were assessed. Completing a test task did not establish insurance acceptance, and no change in issued policies or commercial return is claimed here.
The decision changed which part of the experience the team owned. A form is not usable merely because its fields can be read and filled in by someone who finishes quickly. The warning, extension and return to the draft also shape whether a person can act. I wanted release evidence to include the users for whom the journey still failed, rather than improve the headline by excluding them from the definition of completion.
2025: the removal action needed to identify the person it would affect
A household-application prototype used a compact list with a removal control beside each entry. In forty-eight scripted interaction scenarios, eleven led to the wrong household member being selected for removal from the unsubmitted draft; thirty-seven identified the intended person. Similar displayed names and navigation that depended too heavily on visual row position contributed to the problem. The exercise used synthetic household details. It did not remove anyone from an issued policy or establish a production incident affecting actual customers.
The team proposed making the row control more visually prominent. I asked for the action itself to be reviewable. We allocated ₹60 lakh over ten weeks to a clearer identity summary before removal and a recovery route within the draft. The review showed the selected person’s relevant identifying details and relationship in an accessible form, with the opportunity to return without changing the draft. Privacy and security colleagues assessed the display and recovery design. We did not add unnecessary sensitive information simply to make the confirmation longer.
After revision, forty-five of the forty-eight scripted scenarios identified and handled the intended person correctly. Three remained unresolved and were excluded from the proposed release scope pending further work. That was a bounded test result, not a claim that every possible household or assistive setup had been covered. Existing policy changes remained outside this draft-edit action and required the insurer’s own authorised process. The product account did not convert a draft removal into a cancellation of cover.
The useful improvement was a clearer relationship between the action and its subject. A familiar remove icon had been treated as self-explanatory even though the person it affected was not always clear. The additional review was justified by the consequence, not by a general preference for more confirmation screens. My responsibility was to preserve the customer’s ability to recognise and correct an unintended action before the product carried it forward as though it had been deliberate.