Every presales organisation reaches a point where estimation becomes the bottleneck. It happens quietly. The people who can estimate well are the people you also need on delivery, and the spreadsheet that captures their judgement has quietly forked into six versions.
The instinct is to standardise the template. That helps less than expected, because the template was never the asset — the model underneath it was, and it lives in people’s heads.
Treating estimation as a product changes the questions. What are the actual complexity drivers? Which of them move effort more than ten percent? What ranges do we have real evidence for? Can a new joiner produce a defensible number in an afternoon, and can they explain how they got there?
The output should be a range, not a number, and it should come with its assumptions attached. A single figure with no visible basis is worse than no estimate: it survives into the contract, and delivery inherits it without knowing what it assumed.