The first time I ran refresh grants, I gave everyone on the team the same percentage bump and called it fair. Eighteen months later my best engineer was interviewing elsewhere and my least engaged hire had a fresh four-year vest sitting untouched. Treating equity refreshes like a cost-of-living raise was the mistake. A refresh is a retention tool aimed at a specific risk: the people whose original grant is about to fully vest, whose leaving would actually hurt, and who have options elsewhere. Spread it evenly and you dilute yourself to retain people who were never going to leave, while under-investing in the ones who were.
Here's the checklist I use now to decide who gets a refresh, who waits, and who doesn't get one at all.
Start with the vesting cliff calendar, not the org chart
Pull every employee's original grant date and vesting schedule and sort by time-to-fully-vested, not by tenure or title. The point of a refresh is to close the gap before someone's equity incentive hits zero. Anyone with 12-18 months left on their original vest isn't a refresh candidate yet, regardless of how good they are. Anyone inside that window is the actual pool you're deciding within. This single filter usually cuts a 20-person team down to 4-6 real candidates for this cycle, which is the point: a refresh program that touches everyone every year isn't a refresh program, it's a bonus program with extra paperwork.
Score against three criteria, not one
Once you have your vesting-window pool, run each person against three questions.
Performance. Is this person's output directly tied to a metric you'd miss if they left — revenue closed, product shipped, incidents prevented? A general "is a good employee" bar isn't specific enough; you need a bar tied to something you can point to in a board deck.
Role criticality. Would replacing this person cost you more than six months of ramp time, tribal knowledge, or a client relationship? A senior engineer who owns your billing system is more critical than an equally good generalist you could backfill with a contractor in three weeks.
Flight risk. Has this person been recruited, mentioned a competing offer, or shown the classic disengagement signs — declining 1:1 energy, skipping optional syncs, updated LinkedIn? You won't always have hard evidence here, but managers usually know before HR does. Ask them directly: "if this person got an offer tomorrow, would they take it?"
Someone who scores high on all three is an obvious yes. Someone who scores high on one and low on the other two probably isn't a refresh candidate this cycle — they're a conversation about the other two things equity can't fix (comp mismatch, role frustration, or a manager problem).
Set a dilution ceiling before you look at names
Before running the scoring, decide what percentage of the company you're willing to allocate to refreshes this cycle — most seed-to-Series-B companies I've seen land between 0.5% and 2% of fully diluted shares per year, but the number matters less than picking it before you know who's on the list. If you score names first and set the budget after, you'll rationalize a bigger pool than you meant to grant, because it's easier to justify one more "yes" in isolation than to hold a hard line against a specific person's face. The ceiling is what keeps a refresh program from quietly becoming an annual re-grant to your entire team.
Size by scarcity, not by seniority
Within your ceiling, size grants by how scarce the person's combination of skill and context is, not by their title. A staff engineer who's the only person who understands your data pipeline should get a bigger refresh than a VP whose function has three qualified internal backups. Most companies size a refresh at 25-50% of the original new-hire grant, vesting on its own four-year clock stacked on top of what's left of the original — but the percentage inside that range should track scarcity, not org chart position. This is also where you should sanity-check against your dilution math from the prior grant round, since refreshes compound faster than founders expect once you're two or three cycles in.
Decide the cadence and write it down
Pick annual or biennial and stick to it, tied to your performance review cycle so it doesn't turn into a series of one-off negotiations triggered by whoever complains loudest. An unwritten, ad hoc refresh policy is how you end up with two people at the same level, same tenure, same output, and wildly different equity outcomes because one of them asked and the other didn't. Put the criteria above in a one-page internal doc, even informally. It won't stop every hard conversation, but it turns "why did they get more than me" into "here's the framework, and here's where you scored on it" — which is a very different conversation to have.
The teams that get this right don't run a bigger refresh program. They run a narrower one, aimed precisely at the handful of people whose vesting cliff, criticality, and flight risk actually overlap — and they say no, clearly and on the record, to everyone else.