In January of last year, we pushed a Notion page live to the whole company. It listed every role at Velo, the band it sat in, the floor, the ceiling, and the multiplier we used to convert level to base. We wrote about it on a Wednesday afternoon in Slack. By Friday, three people had asked for a chat.
Eighteen months later, that page is still up. Some of what we expected happened. Most of what mattered did not. We wanted to write down what we published, what we held back, and one thing that broke.
What we put on the page
The Notion doc has six tabs. It contains everything a new hire or a curious teammate would ask us in a one to one:
- Level definitions for engineering, design, product, and go to market, with three example projects per level.
- Base salary bands for every level, in GBP and EUR, refreshed each January using Radford and our own market pulls.
- The formula that turns level into offer: floor plus a location factor, capped at the ceiling.
- Equity grants as a range of shares, plus a plain English note on our 409A and the strike price at the last round.
- The rubric we use in promotion committee, including the four criteria and how we score them.
- A change log at the bottom, so people can see when a band moved and why.
We linked it from the onboarding checklist in Linear and pinned it in the #comp channel. New hires get a Loom walkthrough on day two.
What we did not publish
We were honest with ourselves about the gaps. A few things stayed off the page on purpose:
- Individual salaries. We publish bands, not names. Two teammates asked for a fully open sheet in month four. We said no, and wrote a short doc explaining why.
- Bonus targets for the two people on commission. Their plans are bespoke, and putting a range on them would have misled the rest of the team.
- The one off retention grants we handed out during a rough patch last summer. Three people got them. We told each of them the same thing: this is not part of the standard rubric.
- Founder salaries for the first year. We added those in month nine, after a Datadog outage weekend where the on call engineer asked, in a Slack DM, whether the founders were still on a discount. It was a fair question and we should have answered it sooner.
We are still uneasy about the retention grants. If you publish a rubric and then quietly step outside it, you owe people an explanation. The next time we do one, it will show up in the change log with a category, even if the individual amounts do not.
The hire that fell through
In April, we were close to signing a senior backend engineer out of Berlin. She had spent an hour on a Figma review with our design lead and shipped a small GitHub PR against a starter repo. The team was ready.
Our offer was the top of the L5 band. She had a competing offer that was twenty two percent higher, from a US company hiring in Europe. She sent us a screenshot of the other letter. We said no to the counter, because raising for one hire would have blown a hole through the band we had just published.
She took the other job. It was the right call and it stung. If our page had been private, we would have found a way to say yes and moved on. Publishing the numbers meant we could not privately break our own rules without someone finding out on their first day. That constraint cost us a good hire.
The point of writing something down is that it becomes harder to walk away from. That cuts both ways.
The quiet compression
The thing we did not predict was what happened to the bands themselves.
Before publishing, our L4 engineering range spanned roughly forty percent floor to ceiling. Managers used the full width. Two people at the same level, hired six months apart, could sit twelve thousand pounds apart based on how the negotiation went.
Once the numbers were visible, the top of the band became a promise. Anyone below the midpoint asked, reasonably, why. Over three review cycles, the width of that band shrank to about eighteen percent. Not because we changed the policy on paper, but because managers stopped defending the low end and stopped offering the high end without a clear promotion signal.
The average went up. The variance went down. Our comp spend for the L4 cohort grew by nine percent year over year, and we lost the flexibility we used to have to reward a rare skill without moving someone a level.
We are not sorry about it. It made every one to one about pay shorter, and it turned promotion committee, which meets the second Thursday of each month, into an argument about evidence rather than an argument about numbers. Two people got promoted last cycle who would have been talked out of it under the old system, because the manager could no longer say, quietly, that they were already paid like the next level.
What we would do again
If we were starting over, we would publish sooner and narrower. Wider bands look generous on paper. They also give managers room to make decisions they cannot defend in daylight. We would rather have the argument up front.
The Notion page is still open in a tab on our screens most Fridays. It has not fixed hiring. It has made the conversation we have about pay a smaller, calmer thing. Given where we started, that is enough.