Bargaining Over Estimates Shrinks the Buffer, Not the Work

When a number needs to come down, cut scope instead of negotiating the estimate.

“Three hours spent arguing a figure is not that question.”

Three hours of bargaining over engineers' estimates shrank the buffer for being wrong, while the actual work stayed the same size.

When a team sees its estimates get bargained down, it learns to pad them, so the numbers negotiated next are worse.

Pushing back is fair when an estimate looks padded or rests on unquestioned scope, but when a number must fall, cut scope and let the estimate follow.

Three hours of a recent call went into bargaining over estimates. The work behind those numbers stayed the same size.

I sat through one of these. A startup CEO joined a technical call and started bargaining over the engineers' estimates, one by one.

Three hours later, we had covered about 40% of the estimates. Working afterwards was hard; the feeling in the room was that we'd been had, the hours skimmed off us.

The CEO, as far as I could tell, was pleased with how far he had moved the numbers.

What shrank in that room was the buffer for being wrong — the one part of an estimate you actually miss when reality shows up.

And there's a second cost I think founders underweight. A team that watches its estimates get bargained learns, fairly quickly, that estimates here are opening positions.

So the next ones come in padded, and now the numbers you're negotiating are worse numbers.

The fair version of the other side: pushing back is sometimes right. Estimates can be padded, or built on scope nobody has questioned yet. If a number surprises you, asking why is your job.

But that's a question about the work. Three hours spent arguing a figure is not that question.

When a number genuinely needs to come down, I think the honest move is to cut scope. Remove work, and the estimate follows it down on its own. That's a real reduction.