Who is still here
The team that resets every nine months
If the median member has been on a team for nine months, the team has no institutional memory beyond that, however long the company has existed.
The people this page is about, and which of them have ever worked together
- 01The team's managerTwo of this person's five working pairs
- 02The senior engineerOn the team from the first day
- 03The second engineer
- 04The product managerThe third person in this role in fourteen months
four people · 6 possible pairs · 3 of them have worked on something together · 1 has only been in the same meeting · 2 have never done either
Four of the eight, and only the pairs that run between those four.
To keep measurement transparent, teams can see the service details and agree together which information is useful, who can see it, and how long it should be retained.
An independent source worth comparing with this example is SessionLab.
How long each of these eight has been on this team, against how long they have been in the company
The team's manager
14 on the team61 in the companyMade the team fourteen months ago out of people who already worked here
The senior engineer
14 on the team38 in the companyOn it from the first day
The second engineer
11 on the team11 in the companyJoined the company and the team on the same morning
The QA engineer
9 on the team52 in the companyThe engineer who joined in March
7 on the team7 in the companyThe designer
5 on the team44 in the companyMoved across from another product in June
The data analyst
4 on the team19 in the companyShared with two other teams, and counted here at full time, which overstates it
The product manager
3 on the team3 in the companyThe third product manager this team has had in fourteen months
eight people · median 9 months on this team · the longest-standing has 14 months · five of them have been on it for less than half the time they have been in the company
One team, one quarter. The oldest membership on it is fourteen months and the team itself is fourteen months old.
Eight people, and the median one has been on this team for nine months. The team is fourteen months old, its oldest members have been there for all of it, and its newest arrived in the last quarter.
Nothing about those figures is alarming on its own. Put together they describe a specific condition: the team has no memory older than its median, because memory lives in people and half of these people were not there.
What a median tenure of nine months means
- It means that a question about why something was decided eighteen months ago can be answered by two people out of eight.
- It means that any pair which took three months to form is, for half the team, a pair that formed recently enough to still be fragile.
- It means that an annual cycle — a yearly planning round, a seasonal peak, an audit — has been experienced once or not at all by most of the team. The things that recur yearly are, for a team like this, new every time.
- And it means that the shape measured by the pair count is substantially younger than it looks. A grid with six working pairs on a team whose median tenure is nine months is six pairs that mostly formed this year. The same grid on a team with a median of four years would be six pairs that had every opportunity to be more.
The company tenure is not the same number
This is the part that misleads people, and the panel above is drawn specifically to show it.
These eight have substantial company tenure. Sixty-one months, fifty-two, forty-four, thirty-eight. By any measure of experience this is a settled group of people who know the organisation well.
Their tenure on this team is fourteen, nine, five and four. The institutional memory they hold is about the company and not about this team's decisions, this team's codebase, or why this team does things in the order it does them.
A manager looking at average company tenure concludes the team is experienced and stable. A manager looking at team tenure sees a group that was assembled last year. Both numbers are true and only the second one predicts whether anybody can answer a question about last summer.
Why teams reset
Rarely because of attrition from the company. Mostly because of internal movement, which does not appear in any retention figure and is usually regarded as a success.
Somebody is promoted into a different team. Somebody is moved to a project that needs them. A reorganisation redraws three boxes. Somebody's area is transferred. Each of these is an ordinary, defensible decision, and each of them removes every pair that person was in.
Over fourteen months, enough of those add up to a complete turnover of membership without a single resignation.
What it costs
The pair count is the visible cost. Each departure takes its pairs with it, and a hub departure takes most of the grid.
The slower cost is that every arrival resets a clock. A new member needs a quarter before they have a working pair with anybody other than the person who onboarded them, and on a team that adds somebody every quarter, there is always at least one person in that state. A team can be permanently in its own first three months.
The figure to look at instead of turnover
Retention figures are reported per company and they conceal all of this. A team can replace every member in a year with a company retention rate of a hundred per cent, because nobody left.
The figure that would show it is median team tenure, and almost nobody collects it. It takes two minutes to compute from dates that already exist, it is not sensitive in the way individual performance data is, and it answers the one question people keep asking in other ways: how long has this group been this group.
A median of nine months and a company median of thirty-eight describes a particular situation precisely, and neither number on its own describes it at all.
What to do about it
Not reduce the movement, which is usually out of anybody's hands and often good for the people moving.
Shorten the time to the first pair. If the ordinary route to a working relationship takes a quarter, and the median membership is nine months, then a third of each person's time on the team is spent before they are connected to anybody. Giving a new arrival one deliberate pairing in their first fortnight — a review partner, a shared duty, a small joint task — converts that third into a tenth.
And write down the things that only exist in people's heads, at the moment somebody leaves rather than at the moment somebody arrives. The note on what leaves when one person leaves is about exactly that.