Closing an exception ticket at approval can discard the most useful part of the story: whether the alternative solved the problem and who had to keep it working.
Consider two illustrative requests for the same tool. One team needs a recovery behavior the shared service does not support. Another wants familiar commands for a workload the shared service already handles. A count of “two requests for tool X” merges a product gap with a preference. It cannot tell the platform owner what to build.
Keep a short follow-up record for each approved exception. Start with the unmet requirement in the consumer's words, then record the chosen path, the operating owner, and the result after real use. Include central help that remained necessary. A supposedly independent solution that repeatedly needs the platform team's intervention has not established the independence its approval assumed.
The review can then make a specific choice. A recurring recovery requirement may justify extending the shared guarantee. A discoverability problem may need a better entry page. A workload with materially different duties may remain outside the platform. Each response changes a different part of the product.
Give refusal the same precision. If the platform cannot accept the support burden, name that burden. If a consumer must retain common identity or incident controls, keep those controls in the exception. Departure does not require unlimited freedom, and standardization does not remove the need to explain a boundary.
Adoption measures can help, provided they answer the right question. Required usage establishes coverage of the mandate. To learn whether the consumer path works, inspect a completed task, the assistance it required, and any work performed outside the counted service. Where consumers have a genuine opportunity to choose again, record that choice separately.
Bring related exceptions and their follow-ups to the roadmap review. Ask which requirement recurs, whether the alternatives actually met it, and what their continuing operation cost in work. This can produce a smaller and better-defined investment than adding the most requested tool.
The exception process has paid for itself when the next consumer encounters a better guarantee, a clearer exclusion, or a workable transfer of responsibility. Approval alone tells you none of those things happened.