People Centricity case study

The Leavers the Talent Review Said Would Stay

4years of HR records, leaver by leaver
13variables tested for every leaver
5questions: who leaves, why, where to, at what cost, who stays
3talent lists crossed with who actually left

The situation

The question behind the question.

Across the market, resignations were rising after the pandemic, and this bank was building an employee value proposition to hold on to its people. Before deciding what to offer, HR wanted to know who it was losing. The usual turnover report gives a rate and a list of exit reasons. It rarely says whether the people leaving were the ones the bank could least afford to lose, or whether anyone had seen them coming.

As it was asked

“Tell us why our people are leaving.”

As we reframed it

Find out who leaves, why and where to, test each pattern before it counts, and check whether the bank saw its most valuable leavers coming.

How we worked

How the work unfolded.

01

Ask 5 questions

We started by turning a rate into 5 questions: who leaves, why, where to, at what cost, and who stays. 4 years of records from the HR system were the base, with 13 variables for every person, from tenure and generation to performance, talent status and country, plus the reasons leavers gave and where they went next. The fifth question mattered as much as the first: what keeps the people who stay is what a retention plan must not break.

02

Test before telling

Then the tests. Each segment’s share of leavers was set against its share of the workforce, and its own rate read year on year, so a large group could not pass for a risky one. Every apparent pattern then went through a statistical test, and a regression held the other factors constant, so tenure was not mistaken for age, or country for business mix. Several intuitive suspects dropped out. Gaps in the data were stated, not smoothed.

03

Check the talent review

Then the question most turnover reports skip: were the leavers the people the bank could least afford to lose, and had anyone seen them coming? Regret was defined before the fact: a leaver counted as a regretted loss only if the talent review had already placed them on one of the bank’s 3 talent lists, from succession to high-impact roles. Each was then checked against the flight-risk rating given before they left. The answer is the headline below.

04

Price the excess

Then the price. Some turnover is healthy, so only the excess over the pre-pandemic rate was costed: counting every leaver overstates the case and loses finance at the first slide. Each excess leaver was costed at 100 to 150% of salary, covering the vacancy, the hire and the months before a newcomer is fully productive, an industry rule stated as such. Projected over 3 years, it gave a range, not a point. The findings went straight into the employee value proposition, where careers became a priority.

The instrument

3 checks before a pattern counts

Share and rate. A group’s share of leavers against its share of the workforce, and its own leaving rate year on year: large is not the same as risky.

Test, then control. A statistical test for each pattern, then a regression that holds the other factors constant, so one cause is not mistaken for another.

Back-test the ratings. Every leaver checked against the flight-risk rating given before they left, year after year, so the ratings learn from who actually went.

A flight-risk rating nobody checks against who actually left is an opinion, not a measure.

What moved

What changed, and what didn't.

4 in 10high-impact leavers had been rated at low risk of leaving
Evidence Decision taken Adopted in practice Measured impact
What moved

4 in 10 high-impact leavers had been rated at low risk before they left. The findings became the evidence base of the employee value proposition, which put careers first, and we recommended 3 retention moves: recalibrate how flight risk is rated in talent reviews, hold stay conversations with key talent at risk, and record a reason and a talent status for every leaver.

What didn't move

The analysis read what leavers said on exit, not what stayers felt. Up to a third of leavers had no documented talent status, and the cost rested on an industry rule, not the bank’s own salaries. The 3 retention moves were recommended, not yet in place, and had no named owner when we handed over.

What we'd say to a peer

Before you ask why people leave, ask whether you saw them going. Define regret before the fact, then check every leaver against the flight-risk rating given before they left. If your most valuable leavers were rated safe, the problem is not only the offer, it is how risk is judged. A rating nobody back-tests is an opinion.

Let's talk Discover People Centricity Read more cases
Method note:

Own approach in 4 steps; 4 years of HR records, 13 variables; statistical tests and regression; flight-risk back-test; costed excess turnover.

WeAreHuman@Work

One piece of evidence on sustainable work, every Sunday. Free, and one click to leave.

Get the weekly edition