People Stay 29 Months, Not 19. And Remote Work Is Not the Question We Thought.

The tenure figure in the board pack leaves out everyone who has not left yet, which is 57 percent of the company. Remote work turns out to halve resignations for eighteen months and then double them, which is a management problem with a date on it rather than a policy question.

To Chief People Officer

From People Analytics

Re Retention reporting, and the remote-work decision

Where this comes from

Chapter Chapter 203 · Survival Analysis: Time to Event

Part Part XXXIII · Capstone Projects: Specialized Methods

Dataset capstone-survival-analysis-time-to-event.xlsx

Notebook View the analysis

Recommendation

Bottom line

Keep the remote-work policy and build something specific at the eighteen-month mark. Change the retention figure we report from 19 months to 29, and stop reporting mean tenure of leavers entirely. Platform is the best-retaining team in the company, not the worst.

The number in the board pack is measuring the wrong thing

We report mean tenure of people who left: 19.1 months. That figure excludes 1,733 employees who are still here, and it always will, because anyone who stays a long time has not left yet and so never enters the average.

Using the standard method for this, which counts a current employee as evidence that people last at least as long as they have so far, median tenure is 29.0 months. We have been understating retention by about a third for as long as we have been reporting it.

A survival curve falling from 100 percent at hire to about 30 percent at six years, with a shaded confidence band. A dashed red vertical line marks the 19 months currently reported and a dotted line marks the true median of 29 months, with a horizontal reference line at 50 percent. — Figure 1. The share of hires still with us at each month, with the figure we currently report marked.

	Still with us
At 12 months	84%
At 24 months	58%
At 36 months	41%
At 48 months	36%

These are the numbers to plan headcount against.

Platform is not in trouble

Platform shows the shortest tenure in the company, 6.9 months against roughly nineteen everywhere else, and it has been raised twice as a concern. Platform was created eighteen months ago. Nobody in it can have long tenure, and only 5.3 percent of its people have resigned at all.

On a like-for-like basis

93 percent of Platform's hires are still here at twelve months, the best of any team. The worst is Customer Support at 76 percent. Any duration statistic will make a young team look like a failing one, and we should stop comparing teams on one.

Remote work: the answer changed three times

This is the finding worth the most attention, because the two obvious analyses both say remote work makes no difference and both are wrong in the same way.

Two survival curves. The remote curve sits clearly above the onsite curve for the first eighteen months, the two cross at around twenty months, and the remote curve then falls below the onsite one and stays there, ending about twelve points lower at four years. — Figure 2. Remote and onsite retention over the first six years. The lines cross at about eighteen months.

Analysis	What it says	What it would have us do
Average tenure of leavers	remote stay 3.8 months longer	expand remote work
Standard statistical test	no difference, p = 0.17	ignore the policy
Allowing the effect to change over time	resignations halve for 18 months, then double	keep it and manage the 18-month mark

Three defensible-looking analyses of the same data, three different recommendations.

The middle answer is the one we would normally have accepted, and it is an average of a halving and a doubling. Remote employees resign at half the onsite rate for their first eighteen months, and at twice the onsite rate afterwards.

What we are asking for

Three things

One. Retire mean-tenure-of-leavers from every report and replace it with retention at 12 and 24 months.
Two. Design an intervention for remote employees approaching eighteen months. That is where the risk sits and it is a date we can act on in advance.
Three. Close the Platform review and open one on Customer Support, where 24 percent of hires are gone inside a year.

← Back to Chapter 203 📓 Companion notebook 📊 Download dataset

From Statistics, Data Science and AI: A Visual Handbook by John Fisher. Every statistic, table, and figure in this report is reproduced by the companion notebook.