Promotion Criteria Framework Template

Start with the levels, since every case will be judged against them. For each one, write down what the person owns, how much guidance they need and what others trust them to do.

Write each line so someone who has never worked with the person could check it. "Strong communicator" fails that test. "Reviews others' work in the area by default" passes, because anyone can see who reviews what.

Share the levels with everyone they apply to, well before any case depends on them. Once they're settled, each level can carry its own salary band.

The bar is sustained work at the next level, shown by evidence. Dropbox's public engineering framework calls this "evidence, not faith"¹: "We don't promote someone because we hope they can handle it." It also says no set of checkboxes guarantees a promotion.

Doing the current job well is not the same as being ready for the next one. In a study of sales staff at 214 firms, researchers Alan Benson, Danielle Li and Kelly Shue found top sellers were more likely to be promoted into management. Yet each doubling of a new manager's past sales was associated with a 7.5% fall² in the sales of each person they went on to manage.

Leave tenure out of the criteria. Time at a level helps you spot who to look at, but nobody should move up because months have passed. Nobody should wait once the evidence is there, either.

A process that only reviews submitted cases can't see the people nobody nominated. Lara Hogan, who coaches managers in tech, notes that the names we raise for opportunities are naturally the people we're closest to³. They are also the colleagues whose work we know best, and who look like us.

Being passed over has a cost. In a 2022 Pew Research Center survey, 63% of US workers who quit a job in 2021⁴ named a lack of opportunities for advancement as a reason. A time-in-level check at the start of each cycle shows who may have been missed:

1. 1

   Count months at level

   For each person, the months since they reached their current level, by promotion or hire.

2. 2

   Find each level's median

   Work it out level by level, because normal time grows with seniority. Dropbox's engineering framework gives individual contributors a typical time in level⁵ of 1.5 to 2 years at L1 but 2 to 4 at L3.

3. 3

   Flag the long stays

   The template flags anyone at or above 1.5 times their own level's median. A flag means look closer, not promote.

4. 4

   Remove anyone already handled

   Take out anyone with an open case, since they're in process. A rating below Meets in the last performance review calls for a performance conversation instead.

5. 5

   Ask why for each name left

   Their manager says which is true: the evidence isn't there yet, or nobody has built the case.

In the template's sample team of 22 engineers, the medians are 15, 21.5 and 27 months at L3, L4 and L5. Five people pass the 1.5 times mark. One already has a case in committee, so four people, 18% of the team, are overlooked.

Some levels are meant to be a long-term home. Dropbox expects L4 to be a career level for many engineers⁵, so a long stay there is normal. At a level like that, treat a flag as a talk about growth, not a missed promotion.

A promotion case is the written argument for one person. Its job is to let a stranger decide, so use the same four parts every time:

1. 1

   Evidence at the target level

   Specific instances, each tied to a line in the next level's definition: what happened and what it shows.

2. 2

   The gap

   Anything in the definition not yet shown. A case can go forward with a gap, as long as it is named.

3. 3

   Rating context

   If the latest rating is Meets rather than Exceeds, say so and explain why the case still stands.

4. 4

   Sponsor and reviewers

   The manager backing the case, plus peers who saw the work and can vouch for it.

Collect evidence as the work happens, not the week before the committee meets. Results tracked in an OKR template make strong evidence, since the goal and the outcome are already written down.

Have a small committee decide, rather than each manager alone, so one bar applies across teams. Members reread the target level's definition before each case, then judge every line of evidence against it.

Tenure and a manager's enthusiasm aren't evidence. A case with three years at level and thin evidence is weaker than one with a single year and a clear example of the target scope.

Put the overlooked list on the same agenda as the submitted cases, so a missing case is as visible as a rejected one. Log each decision and its reason in the case tracker, and tell every candidate what the committee found.

Open this template with Edit with AI to skip the arithmetic. River works out each level's median from your team list, checks every flag against open cases and ratings, and drafts cases for the people your committee puts forward.