The problem.
An online grocery business had two workstreams that mattered to revenue. One fed its retail-media advertising. The other kept the online product catalogue correct and complete. Both ran through a shared backlog that nobody trusted. It was full of broad, ageing items, work that had drifted, and duplicates. The team couldn't plan from it, and leaders couldn't see progress. The instinct in the room was to push the delivery team harder. The data said the real problem was further upstream.
What we did.
We started with evidence, not opinion. A full export of the delivery data showed the team did deliver. The constraint sat at the front end. Work waited there to be refined and made ready to pull. It waited again at the sign-off gate before it could be called done. Making the engineers go faster wouldn't have moved the number that mattered.
So we fixed the layer the plan sits on. We restructured the backlog so the real advertising and product-data work was visible and correctly labelled. We caught a scoping error in a bulk edit before it could land. We set a simple, repeatable standard for tracking the plan of work, so senior stakeholders could see it and its progress. We designed a rule for retiring long-dead items, with a standing version so the clutter wouldn't build back up.
We anchored all of it in the client's own documented standards for refinement and prioritisation. We didn't import a new framework for the team to learn. The value wasn't a new method. It was making the one they already had actually hold.
In parallel, we got the fixes that mattered moving. The retail-media inventory and regional-data issue was lined up for release. The product-data problems had been living in chat threads and documents. We pulled them into the backlog, scoped them, and made them ready to commit.