Most onboarding improvement efforts start with a survey. A link goes out the afternoon training ends, the average comes back at 4.2 out of 5, and the number lands in a report. It is not a bad number. It simply does not tell you what to change next quarter.
Onboarding is not a satisfaction problem. It is a ramp-up problem. What matters is when a new hire can work without supervision, where they get stuck, and why some of them leave within three months. This piece covers four metrics you can realistically collect inside 90 days, each with a definition, a formula, a collection method, and the trap that comes with it.
What Satisfaction Surveys Do Not Tell You
A satisfaction score does say something about the quality of the training experience. You learn whether sessions dragged and whether the material made sense. The problem is that the score barely connects to work outcomes. A warm mentor and well-built slides score high without making anyone capable of shipping a first change alone.
The format has structural limits too. People answer right after training, before they know what was missing in real work. New hires are being evaluated, not evaluating, so low scores are rare. Sample sizes are small: in a team that hires five people a quarter, the gap between 4.2 and 3.9 means nothing.
So keep the survey, but demote it. The primary metrics have to come from the work itself.
Four Metrics You Can Actually Collect
One rule when picking metrics: use something that is already being recorded somewhere. Any metric that needs a brand-new collection process tends to die within two months.
Time-to-Productivity
The number of working days until a new hire completes a defined reference task without supervision. Pick one reference task per role and the math stays simple. For a developer it might be the first commit merged without major rework, for a publisher the first page marked up alone, for a PMO the first weekly report they wrote that shipped as-is. The formula is the reference-task completion date minus the start date, and at the org level you track the median across a quarter of hires.
Collection comes from tools you already run: git history, the issue tracker, and the request system all carry dates. The trap is that reference-task difficulty drifts from person to person. One new hire gets a copy fix in week one, the next gets the payment module. If you do not define the reference task including its difficulty, this metric ends up measuring assignment luck. That is also why the median beats the average here.
Time to First Contribution
If Time-to-Productivity measures independence, first contribution measures the step before it: the first moment any real output of theirs entered the team, regardless of size. A typo fix, a set of meeting notes, one test case — all of it counts. The formula is the date the first output landed minus the start date.
This one is useful because it exposes entry friction directly. If first contribution lands on day nine, the cause is usually not the person. Something is stuck in account provisioning, local setup, access permissions, or first-task assignment. The trap is gaming the number by inventing token work. A meaningless first commit improves the metric and does nothing for ramp-up.
30/60/90 Checkpoint Completion
Write down, in advance, the state a new hire should reach at day 30, 60, and 90, then check it at those points. Day 30 might be knowing the tools and processes and knowing who to ask. Day 60 is handling routine work alone. Day 90 is making a call in an exception and finding the right people for it. The formula is items met divided by total items.
Managers collect this in the regular one-on-one. The trap is writing items as feelings instead of behaviors. There is no way to verify that someone feels comfortable with the team culture. Write observable sentences instead, such as being able to explain the escalation order when an incident hits. Keep it under five items per role or it will not survive the quarter.
90-Day Early Attrition
People who leave within 90 days divided by hires in the same window. It is the slowest onboarding metric and the heaviest one, because it burns recruiting spend and team time together.
In smaller organizations a single departure swings the ratio hard, so read it annually rather than quarterly, and treat the reason coding from exit conversations as more informative than the rate itself. The trap is charging the whole number to the onboarding program. Early exits mix in expectation gaps from hiring, the manager relationship, and compensation. Split the causes or you will fix the wrong thing.
A Realistic Way to Collect the Data
Trying to launch all four at once usually fails. Stage it instead.
- Quarter 1: record start dates and first-contribution dates only. One spreadsheet is enough.
- Quarter 2: define the reference task per role and start measuring Time-to-Productivity.
- Quarter 3: write five 30/60/90 items per role and attach them to one-on-ones.
- Quarter 4: with a year of data, look at how early attrition relates to the earlier metrics.
To keep measurement from becoming the work, review the numbers once a quarter and change exactly one thing each time.
Data That Game-Based Onboarding Leaves Behind
Scenario-based onboarding gives you part of the above with no extra collection, because playing is already a record.
In a quest structure like First Run, each session leaves behind accuracy per role quest, wrong answers that repeat within the same type, time spent on judgment questions, retry counts, and the areas nobody finished. That data is for instructional design, not performance review. If five of six new hires picked the same option in the change-request quest, that is not individual error. It is a signal that the procedure is written ambiguously.
One caveat worth stating plainly: a quest score is not a proxy for job performance. Judging well inside a game and judging well on a live project are correlated at best, never equivalent. Use quest results as a diagnostic for what to reinforce, and judge readiness with the work metrics above.
Measurement Checklist
- Is there one defined reference task per role?
- Are start date, first-contribution date, and reference-task date recorded in one place?
- Are the 30/60/90 items written as observable behaviors?
- Are you reading medians instead of averages?
- Are early-exit reasons split into hiring, manager, compensation, and onboarding?
- Do you pick exactly one improvement per quarter?
- Is the satisfaction survey a supporting metric rather than the headline?
The point of measuring onboarding is not a scorecard. It is checking the same few places on the same cadence to see where people get stuck. Four metrics and one checklist are enough for most organizations to know what to fix next.
---