How to measure team performance

Most advice on measuring team performance is a list of metrics: throughput, cycle time, customer satisfaction, utilisation. Every one of them is useful, and a team can hit all of them for two quarters while it burns out its best people, stops telling the manager bad news and loses the ability to work together. The numbers look fine right up until they fall off a cliff.

The fix is not more metrics. It is measuring the right layers, so that the early warning shows up in the numbers before the damage shows up in the results.

The three layers of team performance

The most useful definition of a performing team still comes from J. Richard Hackman, the Harvard researcher who spent decades studying teams. In Leading Teams (2002) he judged a team on three criteria: its output meets the standards of the people who receive it; the way it works builds the team's capability to work together in future; and the experience contributes to the learning and wellbeing of the people in it. A team that delivers this quarter by consuming its own capacity is not performing. It is borrowing. This fuller standard is what researchers call team effectiveness.

Translated into something a manager can track, that gives three layers:

Results tell you about last quarter. Health tells you about next quarter. Measure both or you will be surprised by the second.

Team performance metrics by team type

Results and delivery measures depend on what the team exists to do. The principle is the same everywhere: one measure the team's customers would recognise, one for speed and one for quality, so that nobody can improve one by quietly sacrificing the other.

The DORA measures come from Google's DevOps Research and Assessment programme, written up in Accelerate (Forsgren, Humble and Kim, 2018). They are a good model for every team, not only engineers, because they pair speed with stability by design. For learning teams, the Kirkpatrick model does the same job: it stops you counting attendance and calling it impact.

How to measure team health, the layer most teams skip

Team health is usually measured with an annual engagement survey, which arrives too late and asks about feelings rather than behaviour. Health is better read from things people do, because behaviour is observable, countable and harder to answer politely. Six that predict trouble early:

1. Commitments kept. Of the things people promised in last week's meeting, what share were done by the date? This is the plainest measure of trust in a team.
2. Help asked early. When work slips, did the person flag it while there was time to help, or on the day it was due?
3. Feedback that changes work. Can the team name three pieces of work this month that visibly changed because someone gave feedback on them?
4. Disagreement in the room. Are objections raised in the meeting, or in private messages afterwards? Count them for two weeks.
5. Decisions with owners. What share of decisions leave the meeting with a named owner and a date?
6. How early bad news travels. How long between a problem appearing and the manager hearing about it? Days are fine. Weeks are a warning.

These are hard for a manager to judge alone, because the manager is the person people are most careful around. So pair the counts with a short survey that the whole team answers separately. Our free team performance check does exactly this: eighteen questions on observable behaviour across six dimensions, scored on the spot, with a written report. Have every member take it on their own, then put the scores side by side.

If the scores point at trust, the narrower psychological safety check is the next one to run, because a team that cannot disagree safely will not report the problems your delivery metrics depend on.

Build a one-page team scorecard

A scorecard is only useful if it changes a decision, so keep it to one page and six or seven measures. To build one:

1. Name the team's customers. Write down who receives the team's work and what they need from it. Internal teams have customers too.
2. Pick one or two results measures those customers would recognise as success.
3. Pick two delivery measures: one for speed, one for quality. Never one without the other.
4. Pick two health measures: one you count (commitments kept, say) and one you survey.
5. Record a baseline before you change anything. Without it, every later number is an opinion.
6. Set the rhythm: delivery weekly, results monthly, health quarterly.
7. Retire any measure that has not changed a decision in two quarters.

Here is an illustrative scorecard for a nine-person customer success team. The figures are invented to show the shape, not a benchmark:

Notice what the scorecard catches. Resolution got faster, but reopened cases rose. Without the quality measure beside it, the speed gain would have been celebrated and the cost would have arrived as churn two quarters later.

Five ways team measurement goes wrong

1. Measuring activity instead of outcomes

Hours logged, emails sent and tickets touched measure effort, and effort is not the point. A team that solves a customer's problem in one call scores worse on activity than one that needs five.

2. Turning a measure into a target

Goodhart's law, usually paraphrased as when a measure becomes a target, it ceases to be a good measure. Reward a support team for closing tickets fast and tickets get closed fast, solved or not. The defence is the pairing above: every speed measure travels with a quality measure.

3. Comparing teams with each other

Velocity, story points and most productivity numbers are calibrated inside one team. Comparing them across teams invites inflation and punishes the team doing the hardest work. Compare each team with its own baseline.

4. Adding up individual scores

Team performance is not the sum of individual performance. A team of strong individuals can underperform badly at the joins: handovers, decisions nobody owns, conflict nobody names. Lencioni's five dysfunctions are all join problems, and none of them shows up in an individual review.

5. Surveillance dressed up as measurement

Keystroke counts and active-time trackers measure presence. They also tell the team they are not trusted, which damages the health layer you most need to protect. Anything you would be uncomfortable showing the team should not be on the scorecard.

Read the gap between the manager and the team

When a manager and a team answer the same health questions separately, the two sets of scores rarely match, and in our experience running team checks the manager's is usually the higher one. The size of that gap is often the most useful number on the page, because it measures how much of the team's reality is reaching the person making decisions.

A worked example: a manager rates people here get specific feedback within a week at 4 out of 5. The team averages 2.6. Neither number is wrong. The manager gives feedback, but in a form the team does not recognise as specific. That gap turns a vague complaint into a concrete fix, and the SBI feedback model is usually where the fix starts.

How often to measure

Expect the pattern to shift as the team matures. A newly formed team will look weak on health while it is still in the early Tuckman stages; that is normal. A team that has worked together for two years and still scores low on health has a problem that more targets will not fix. The characteristics of high-performing teams show what healthy looks like in an ordinary week.

Tour De Force runs this as live, experiential training for teams worldwide, online and in person. Talk to us, or play The Weekly Challenge from our gamified learning apps to see the method in ten minutes.