Estimation Is Fake but Roadmaps Are Real: Living With Both

I have sat on both sides of this table. As an engineer I have given estimates I invented under pressure, and as a product owner I have taken those invented numbers and put them in a slide for leadership. The whole chain knows the numbers are soft. Everyone pretends otherwise. It is a strange little ritual.

Here is the thing I eventually made peace with: the estimate is fake but the conversation that produces it is not. When an engineer says "three weeks" what actually happened is they mentally walked the work, spotted the scary parts, and compressed all that into a number. The number is the least valuable output. The list of scary parts is the treasure, and most planning meetings throw it away.

So I changed what I ask for. Not "how long" but "what would make this take three times longer than you think." Answers to that question are specific. The vendor API has no sandbox. Nobody has touched that module since 2021. We need legal review. Those are risks you can actually work, and working them early is what makes dates hold.

For the roadmap itself I stopped publishing dates for anything more than a quarter out. Near work gets dates because the unknowns are mostly burned down. Far work gets sequencing only, this before that, and a rough size in t shirt terms. When leadership pushes for a date on far work, I give a range so wide it is honest, and I say what would narrow it. That lands better than you would expect, because most execs have been burned by confident dates before.

Estimates are a lossy compression of understanding. Keep the understanding, hold the number loosely, and never let someone else's slide deck turn your guess into a commitment you did not make.