HR metrics for startups, without the vanity numbers

A small team does not need an HR dashboard with forty tiles. It needs a handful of numbers it reads honestly and acts on. The useful set is short: headcount and growth, attrition and retention, time to hire, and offer acceptance. Everything else is either a version of these or a distraction from them.

This guide covers each one, how to read it without fooling yourself, and why some of the metrics vendors sell, particularly individual risk scores, do more harm than good.

Headcount and growth

Start with the simplest fact: how many people work here, in what capacity, and how that is changing. Split it by employees and contractors, because the two behave differently and blending them hides the picture.

Read it over time rather than as a snapshot. A team that went from eight to twenty in a year has a different set of problems from one that has sat at fifteen for eighteen months, even though both employ a similar number. Growth rate tells you whether your onboarding, your management structure, and your policies are about to be tested. If you are adding people faster than you are building the structure to hold them, the number is a warning, not a trophy.

Attrition and retention

Attrition is the share of people who leave over a period. Retention is the mirror: the share who stay. You do not need both, but you do need to define the one you use precisely, including whether you count only regretted departures or every exit, and whether the denominator is your starting headcount or an average.

At small scale, the rate is noisy. One person leaving a team of ten is ten percent, and a single resignation can swing the number in a way that means nothing about the health of the team. So resist reading the percentage in isolation. The honest version of this metric is qualitative: for each person who left, why did they go, was it something you could have changed, and is there a pattern across several departures. A rising rate driven by one bad manager is a fixable problem. The same rate driven by pay that has fallen behind the market is a different one. The number points you at the question; it does not answer it.

Time to hire

Time to hire is the elapsed time from opening a role to the candidate accepting. It matters because a slow process loses good people to faster competitors and leaves the team carrying the gap.

Measure it from a consistent start point, usually the day the role is approved and opened, to the day the offer is accepted. If you measure from first interview instead, you hide the delay that often does the most damage, which is the time a role sits unfilled before anyone starts working it.

Read it alongside quality, not on its own. Driving time to hire down by lowering the bar is not a win, it is a cost you pay later in attrition and management time. The useful reading is: where in the process does time actually go, and which of those steps is slow for reasons you can fix.

Offer acceptance

Offer acceptance is the share of offers that candidates say yes to. It is a quiet but revealing number. A high rate means your offers are well pitched and your process leaves candidates keen. A falling rate means something is off: the pay is below market, the process took too long and the candidate cooled or took another job, or the role was sold as one thing and turned out to be another.

Because you make relatively few offers early on, treat this as a prompt to ask each declined candidate why, rather than as a statistic to average. The reasons are worth more than the rate. Two people in a row declining over compensation is data you can act on this week. Acting on it usually means looking at the bands rather than at one offer, which is where compa-ratio, a person’s pay against the midpoint of their range, starts to earn a place alongside the four numbers above.

Why vanity metrics and risk scores are a trap

Two failure modes are worth naming, because both are easy to fall into and both waste the attention a small team cannot spare.

The first is vanity metrics: numbers that look like insight but do not change a decision. Total training hours logged, average tenure quoted without context, an engagement score with no attached action. If a metric goes up or down and you would do nothing differently, stop tracking it. It is costing you attention and buying you nothing.

The second is more corrosive: individual risk scores. Some tools claim to predict which named employees are about to quit, or rank people by a flight-risk number. At small headcount these predictions are statistically weak, built on too little data to mean much. Worse, they change how you treat people the moment you start believing them, and they poison trust the day an employee discovers a machine has been quietly scoring their loyalty. Capstan deliberately does not do this. No model in the product evaluates individuals or assigns attrition risk scores, and the Advanced Analytics module reports on aggregates only. It runs under the identity of whoever is reading it, so a report can only narrow what that person could already see rather than widen it, and small groups are suppressed rather than reported. The way to understand whether someone is unhappy is to talk to them, which also happens to be the only way to do anything about it.

One practical note on where these numbers come from. Everything in the list above except offer acceptance is answerable from the core, which includes reporting and a report builder on every plan, and scheduled delivery of a report is available on every plan too rather than being held behind a tier. Deeper cuts, and the aggregate views that cross modules, sit in Advanced Analytics, which is a paid module like any other. Time to hire and offer acceptance need the recruitment pipeline to be in the system in the first place, which is the real prerequisite rather than the reporting tool.

Reading the set together

None of these five numbers means much alone. Headcount growth stresses your structure; attrition tells you whether the structure is holding; time to hire and offer acceptance tell you whether you can refill what you lose. Read them together, monthly, and treat each one as a question rather than a verdict. That is a habit a team of any size can keep, and it beats a dashboard nobody trusts.

If you are still setting up the system that produces these numbers cleanly, the HR software guide covers what to look for. And since much of your attrition signal comes out of how you manage and review people, the performance review guide is the natural next read.