---
title: "Noted on 15 Sep 2026"
author: Marc Roberts
date: 2026-09-15
url: https://www.marcroberts.info/notes/2026-09-15-onboarding-time-as-a-diligence-metric/
---

# Noted on 15 Sep 2026

There's one number I ask for that almost nobody has ready: how long between a new engineer's first day and their code running in production.

Most of what's in a technical data room is an assertion. The architecture diagram is what somebody believes is true. Test coverage is a real number that says nothing about whether the tests are any good. The deployment chart might be one engineer who knows the incantation. Onboarding time is different, because it's already happened, and nobody can tidy it up for the data room.

To get a new person to production inside a week, an awful lot has to be working at once: an environment that builds from a documented command, a test suite that passes on a machine that isn't yours, credentials that don't need a founder, a deploy that someone other than its author can run, and at least one person with enough slack in their week to answer questions. If any of those is broken it shows up in the number. I'll still go and look at each of them (it's the job), but the number has usually told me where to look.

Six weeks, on the other hand, tells you something the founders haven't priced in. If it takes six weeks to make an experienced engineer useful, the hiring plan in the model is wrong. Not the salaries, the dates.

And it matters more now than it used to. DORA's 2025 research, from nearly five thousand people in the industry, found 90% of developers using AI at work, and concluded that AI acts as an amplifier: it magnifies whatever system it lands in. Give it a codebase with a fast feedback loop and it compounds nicely. Give it one where nobody can get a clean build and you get more code, faster, going into a system that already couldn't tell you whether the last change worked. So onboarding time has stopped being an induction question and become a measure of how quickly an organisation can find out it's wrong.

I went looking for a public benchmark to tell you what good looks like. There isn't one, just vendor blogs quoting each other's figures with nothing underneath. Which is why I'm asking: if you've onboarded an engineer in the last year, how long was it from day one to their first change in production, and what (if anything) actually held it up?
