Attrition is up 18% — do you have a retention problem? Regretted vs non-regretted attrition

1. Before you start

Attrition (or turnover) is the share of your people who leave over a period — usually a year. If 90 people leave a company that averaged 500 staff, the headline attrition rate is 90 ÷ 500 = 18%. That single number is what lands on a board slide, and it is almost always the wrong thing to react to. The reason is that not every departure is a loss worth worrying about.

Regretted attrition is the departures you did not want: the strong performer who quit for a competitor, the engineer you had just finished training. Non-regretted attrition is the departures you were fine with or even chose: someone managed out for weak performance, a role eliminated in a restructure, a planned retirement. Split the same 90 leavers — say 30 regretted and 60 non-regretted — and the number that should keep you up at night is not 18%. It is the regretted rate, 30 ÷ 500 = 6%.

Three honest statements before you start:

This is a Decide course. You read a situation, learn the idea, and make a call. You do not write or run any code.
The company below, Brightloom, and the company in the final scenario, Rowan Health, are composite — invented from ordinary, realistic figures so the arithmetic is clean. No number is a claim about any real employer.
This is not a certification. It proves, to you, that you can defend a call on whether a rise in turnover is a real retention problem worth spending on.

You need only arithmetic. The hard part is judgment: deciding which leavers actually count as a loss, and refusing to panic at a headline that hides the signal.

2. The Situation

Brightloom is a B2B SaaS company of about 500 people, and its year-end turnover came in at 18% — up from 12% the year before. The board has seen the slide, the word “crisis” is being used, and the founder wants to approve a company-wide retention package: a pay review for everyone, new perks, a re-engagement campaign, roughly $600,000 a year. The head of people has gone quiet, because she has split the leavers the way the board slide never did, and the split tells a very different story about where — and whether — the money should go. You have to decide whether Brightloom has a retention problem worth spending against, and if so, where.

3. What you’ll be able to do

After this course you will be able to:

Split a year’s departures into regretted and non-regretted, and say why the headline turnover rate on its own cannot tell you whether you have a problem.
Compute the regretted attrition rate and read it against the headline rate, so you can tell a real loss of talent from a wave of exits you were fine to see.
Read the regretted rate by team and by tenure, and find the pocket where the loss is concentrated instead of spreading a budget evenly across a company that is mostly fine.
Decide whether a retention spend is worth it — comparing what the regretted losses cost against what a targeted program would cost and save — and name the one number that flips the call.

4. Prerequisites & time box

Prerequisites: arithmetic, and the everyday idea of turnover (people leaving a job over a period). No HR background and no spreadsheet are assumed; every term is defined again as it comes up. New to the Decide hall? Read its short how-to-read page first.

Time box: about 22 minutes of reading (measured), plus your own thinking time on the call in section 7. That is under the 25-minute cap for a concept course.

Difficulty: 4 / 8. This is a manager’s decision, not a specialist’s. There is only light arithmetic — a couple of rates and a segment breakdown — but the trap is real: the obvious reading of the headline number (“attrition jumped, panic and spend”) is usually the wrong one, and you have to hold two ideas at once (which leavers count as a loss, and where that loss is concentrated). If the leavers were already sorted for you and you only had one number to read, this would be a 2 or 3; the judgment about which departures matter and where to point the money is what puts it at 4.

Free-tier honesty: no signups, no paid tools, no special hardware. You read and decide in the browser.

5. The case & where the numbers come from

Brightloom is a composite SaaS company: a mid-size software business assembled from the ordinary dynamics any people leader would recognise — a mixed year of voluntary quits, a small restructure, a handful of retirements, and one team quietly losing its best engineers. The figures are chosen for clean arithmetic and are not drawn from or claimed about any real company. The definitions — turnover, regretted vs non-regretted attrition, retention — are standard and cited in section 11. Every figure in the table below is an in-course assumption; every later number is computed from these.

Item	Figure
Average headcount over the year	500
Total departures in the year	90
Headline turnover rate (90 ÷ 500)	18%
Departures that were non-regretted	60
Departures that were regretted	30
Loaded cost to replace one regretted engineer	$60,000

The 60 non-regretted departures break down as: 20 people managed out for sustained weak performance; 30 roles eliminated in a planned restructure of a product line the company is winding down; and 10 retirements. None of these is a talent loss the company wanted to prevent. The cost figure — $60,000 to recruit, onboard and bring one replacement engineer back to full productivity — is an in-course assumption in the mid-range of what a knowledge worker’s replacement typically costs; it is used only to size the decision, not cited as a fact about any firm.

6. The Concepts

Regretted vs non-regretted attrition

Turnover counts every exit the same way, and that is exactly its flaw. Retention is the company’s ability to keep the people it wants to keep — so the only exits that measure a retention problem are the ones you regretted. Sort every leaver into two piles:

Regretted — you wanted this person to stay. A strong performer who resigned for another offer, an engineer you invested a year in training, a manager a team relied on. Losing them costs real money to replace and real time to recover.
Non-regretted — you were fine to see this exit, or you chose it. Someone managed out for weak performance, a role eliminated in a restructure, a planned retirement. These still show up in turnover, but they are not evidence that you are failing to keep talent.

You classify with the facts you already have: performance ratings, whether the exit was voluntary or involuntary, and what the exit interview said. Of Brightloom’s 90 leavers, 60 fall in the non-regretted pile (20 managed out, 30 eliminated roles, 10 retirements) and 30 are regretted. That single split is the whole game: an 18% headline made of mostly non-regretted exits is a completely different situation from an 18% made of your best people walking out. The board reacted to the 18%. Your job starts with the 30.

The regretted attrition rate

Once the piles are sorted, put each on the same base — average headcount — to get a rate you can compare year to year and team to team:

Regretted attrition rate = regretted departures ÷ average headcount = 30 ÷ 500 = 6.0%.
Non-regretted attrition rate = 60 ÷ 500 = 12.0%.
These add back to the 18% headline: 6% + 12% = 18%.

Read that out loud: two-thirds of Brightloom’s turnover — the 12% — is exits the company was fine with. The regretted rate, the part that actually measures a retention problem, is 6%, and a 6% regretted rate is unremarkable for a SaaS company. The headline jumped from 12% to 18% mostly because the restructure eliminated 30 roles this year, a one-off. Reacting to the 18% with a company-wide pay review would be spending $600,000 to fix a number that is only partly, and mostly not, about retention. The headline rate tells you how many people left; the regretted rate tells you whether that is a problem.

Reading the rate by team and tenure

A 6% company-wide regretted rate can still hide a serious local fire. The next move is to break the regretted departures down — by team and by tenure — because a retention problem is almost never evenly spread. Here are Brightloom’s 30 regretted leavers by team:

Team	Headcount	Regretted leavers	Regretted rate
Engineering	150	18	12.0%
Sales	120	6	5.0%
Customer Success	80	3	3.8%
G&A and other	150	3	2.0%
Company	500	30	6.0%

Engineering’s regretted rate is 12% — double the company rate and more than three times every other team. More than half of all the regretted losses (18 of 30) came from one function. The retention problem, if there is one, is not “Brightloom”; it is “Brightloom Engineering.” Now split those regretted leavers by tenure:

Under 1 year: 9 leavers.
1 to 2 years: 15 leavers.
Over 2 years: 6 leavers.

Half of the regretted losses hit in the 1-to-2-year band — people who had just finished ramping and become fully productive, which is the most expensive moment to lose someone. That is a signal about why they leave (something turns sour right after ramp) and a pointer to where a fix would pay off most. The lesson: a calm company average can sit on top of a team that is bleeding, so always read the regretted rate segmented, never as one number.

(An interactive calculator sits here — enter the headcount, the total departures, how many were non-regretted, and a single team’s headcount and regretted leavers, and it returns the headline turnover, the company regretted and non-regretted rates, the team’s regretted rate, and how many times the company rate that team is running at — with the call on whether the loss is concentrated there. Change the non-regretted count or the team’s leavers and watch the decision flip.)

Is the problem worth spending on

A concentrated regretted rate tells you where the problem is; it does not yet tell you the spend is worth it. That is a money question, so size it. Engineering lost 18 regretted people at a loaded replacement cost of $60,000 each: 18 × $60,000 = $1,080,000 a year walking out of one function. Against that, weigh a targeted fix, not the company-wide package: a $150,000 retention program for Engineering — better onboarding through the ramp, a compensation correction for the 1-to-2-year band, a manager change — that the head of people expects to cut Engineering’s regretted rate from 12% to about 7%.

Cutting 12% to 7% on 150 engineers is 7.5 fewer regretted departures a year (0.05 × 150). At $60,000 each that is $450,000 saved for a $150,000 spend — a net gain of $300,000, and the program pays for itself if it prevents just three departures. Compare that to the founder’s $600,000 company-wide plan, most of which would land on teams whose regretted rate is already 2 to 5% — money spent to keep people who were not leaving. The call is not “is attrition up?” It is “where is the regretted loss concentrated, what does it cost, and does a targeted fix save more than it costs?” For Brightloom the answer is: skip the broad package, fund the Engineering program.

7. Your Call

You have seen how the regretted/non-regretted split, the regretted rate, segmenting it, and the worth-it test decide Brightloom’s call. Now a different one lands on your desk.

Rowan Health is a composite regional hospital network — a very different world from a SaaS company — and its nursing turnover just came in at 20%: 160 of an average 800 nurses left last year. Of those 160, 112 were non-regretted: 40 retirements, 32 managed out, and 40 travel nurses whose fixed-term contracts simply ended as planned. The other 48 were regretted. The regretted losses concentrate in two units — the Med-Surg ward (500 nurses, 20 regretted) and the ICU (100 nurses, 16 regretted) — with the remaining dozen scattered a few at a time across smaller units. One of those smaller units is an outpatient clinic Rowan is winding down: its turnover looks alarming in percentage terms, but its exits are people leaving a unit that is closing, not talent the network is losing (you will weigh it in section 10). You have a single retention budget that can fund only one unit this year — you cannot split it — so rather than a company-wide go/no-go, your job is to decide which unit gets it, and to be ready to say what would change the answer.

How this differs from the taught case (the transfer): this is a different sector and company (a hospital network, not a SaaS firm), the numbers are different so the arithmetic must be redone rather than recalled, it asks a different kind of decision (pick which of two units to fund, not a single company-wide go/no-go), and it adds a constraint the Brightloom case never had — one budget that covers only one unit, so you must rank, not just diagnose. The core concept is the same: split regretted from non-regretted, read the regretted rate by segment, and spend only where the real loss is.

8. Self-check

Before you write the memo, make sure you can say each of these in one line:

What is the difference between a regretted and a non-regretted departure, and why does only one of them measure a retention problem?
The headline turnover rate and the regretted attrition rate can move in opposite directions — how, and which one should drive a spending decision?
A calm company-wide regretted rate can still hide a serious problem — what do you segment by to find it, and what number would flip your call on where to spend?

If any is fuzzy, reread section 6 — the regretted/non-regretted split, the regretted rate, reading it by segment, and the worth-it test are the whole course.

9. Stretch

Push the decision further on your own:

Back at Brightloom: the restructure that eliminated 30 roles is over, so next year the non-regretted rate should fall. If the 30 eliminated roles do not recur and nothing else changes, what does the headline turnover rate become — and what does that tell you about reacting to this year’s 18%?
The Engineering fix is expected to cut the regretted rate from 12% to 7%, saving $450,000 for $150,000. At what reduction (how many points of regretted rate avoided) does the program merely break even against its $150,000 cost — and what does that threshold tell you about how confident you need to be in the fix?
The genuinely hard one: a rival argues that some regretted attrition is healthy — that a company with 0% regretted attrition is overpaying or hoarding underused talent. Write the one paragraph you would give your board on whether there is a regretted rate that is too low, and what evidence would tell a healthy churn from a real leak.

10. Ship it — your decision memo

Write a one-page memo to Rowan Health’s leadership. State the call (fund the ICU: its regretted rate is 16%, nearly three times the 6% company rate, while Med-Surg’s is 4% — below the company rate). Show the two-line arithmetic (regretted = 160 − 112 = 48, so the company rate is 48 ÷ 800 = 6%; ICU 16 ÷ 100 = 16% versus Med-Surg 20 ÷ 500 = 4%). Name what you rejected and why — splitting the budget across a unit that is already below the company rate, chasing the unit with the larger raw number of leavers, and pouring retention money into a clinic whose 30% turnover is entirely non-regretted. Name the one number that would change your mind (a reclassification showing the ICU’s departures were mostly non-regretted, which would drop its real regretted rate). Keep it to a single page a busy executive grasps in two minutes. This memo is your own argued claim — something you can defend in a room, not a credential.

11. Sources

Brightloom and Rowan Health, and every figure attached to them, are composite — invented from ordinary, realistic figures for clean teaching arithmetic, not drawn from or claimed about any real employer. The concept definitions used to reason about them are standard; references below.

Concept / claim	Source (publisher)	URL	Accessed
Employee retention and the cost of turnover	Wikipedia — Employee retention	https://en.wikipedia.org/wiki/Employee_retention	2026-07-20
Turnover/attrition as a rate (churn concept)	Wikipedia — Churn rate	https://en.wikipedia.org/wiki/Churn_rate	2026-07-20
Exit interviews as the basis for classifying a departure	Wikipedia — Exit interview	https://en.wikipedia.org/wiki/Exit_interview	2026-07-20
Layoffs / eliminated roles as involuntary, non-regretted exits	Wikipedia — Layoff	https://en.wikipedia.org/wiki/Layoff	2026-07-20
Performance ratings used to tell a regretted loss from a managed-out exit	Wikipedia — Performance appraisal	https://en.wikipedia.org/wiki/Performance_appraisal	2026-07-20

Next up

Finished this call? Continue the People & HR track:

Who gets the raise? Splitting a fixed merit pool across your team → · Browse all courses