A Series B founder called me in February with a hiring problem. He had eight AEs the previous January. By December he had five, and two of the five were new. Three people had walked, and his read was simple: our OTE is not competitive, we need to pay more.
I asked him to send a CRM export before we talked about money. Deal records with owners and created dates, stage history, account assignments, activity logs, going back two years. It took his ops person half a day.
What came back was not a pay story. Two of the three people who left had inherited account lists that had already been worked. On one of those lists, 41% of the named accounts carried a closed-lost deal from the prior 18 months. Both reps had watched their own new pipeline creation fall by more than half over five months before they resigned. Neither ever raised it in a one-to-one. The team pipeline number looked fine the whole time, because the two reps who did have fresh territory were covering the gap.
He was about to spend an extra 180,000 euros a year on OTE to fix a territory design problem.
The number almost nobody puts in the plan
Sales turnover is high everywhere, and B2B software is worse than average. A survey of 939 B2B companies covering Q2 2025 through Q1 2026 puts annual turnover at 45% for SDRs, 30% for AEs, 28% for sales managers and 25% for CSMs, with B2B SaaS running above the cross-industry sales figure at 38% overall (Optifai). The Bridge Group's AE benchmark work has median AE attrition around 32%, roughly 20 points voluntary and 12 involuntary, against average tenure of about 2.2 years (The Bridge Group).
Set that against the all-industry turnover baseline of roughly 13% and you get the shape of the problem. Sales is not a normal function. You should plan for losing about a third of the team every year, and most plans do not.
The replacement cost is the part that makes CFOs sit up. Current estimates put a departure at 115,000 to 150,000 dollars fully loaded, and the breakdown is worth reading line by line: 15,000 to 25,000 in recruiting, 20,000 to 40,000 in onboarding, 25,000 to 50,000 in pipeline lost during a 45 to 60 day vacancy, 22,000 to 38,000 in ramp productivity you pay for but do not receive, plus manager hiring time (Gangly). For an enterprise AE on a six-figure OTE with a nine-month ramp, it goes past 200,000.
Notice that only two of those five lines are recruiting costs. The rest is revenue that never happened.
Turnover is a capacity number, not an HR number
Here is the framing problem. Turnover lives in the HR deck. It shows up once a year next to engagement scores and gets discussed for four minutes. Meanwhile the sales plan is built on headcount, and headcount is treated as a fixed input.
That is how you end up with a plan that is wrong by a third before the year starts.
Run the arithmetic on a team of nine AEs with a 400,000 quota each. On paper that is 3.6 million. Now apply reality. Historical attainment is 68%, so a fully ramped rep delivers about 272,000. You will lose three of the nine at 32% attrition. Each vacancy runs about six months before a replacement is producing, counting search and ramp. That is roughly 1.5 rep-years of capacity gone, or about 400,000 of delivered revenue, before anyone underperforms.
The same math belongs in your capacity model, where effective capacity is heads multiplied by ramped percentage multiplied by expected attainment multiplied by one minus turnover. Skip the last term and the plan inflates by exactly the amount you are about to lose.
Turnover is the only capacity input your plan treats as zero.
You discount attainment. You discount ramp. Then you assume the nine people you drew in January are the nine you have in December, which has not been true at any company I have worked with.
Your CRM knew before the resignation email
This is the part that gets missed. Exit interviews are late and polite. People give you the version that keeps the reference intact, which is usually comp or a vague better opportunity. The honest version was visible in your systems months earlier, and it was measurable.
Five signals, all available from data you already have.
Pipeline creation trend by owner. Not pipeline value, which lags and hides behind one large deal. Count of new opportunities created per rep per month, on a three-month rolling average. When a rep checks out mentally, self-sourced pipeline creation is the first thing that drops, typically four to six months before they resign. They still work the deals they have. They stop starting new ones.
Activity concentration. Calls and emails clustering into two or three accounts while the rest of the book goes quiet is a rep protecting their last chance at a commission cheque. It reads as focus on a dashboard. It is usually resignation in slow motion.
Territory yield. Closed-won value divided by the number of accounts assigned, by rep, over the last four quarters. Then the same for closed-lost coverage: what share of a rep's named accounts already carry a lost deal from the prior 18 months. If one rep's territory yields half of what another's does, you have found a fairness problem that no coaching plan will fix. Details in sales territory planning.
Attainment spread. If 30% of the team is at quota and the rest are at 40%, the quota is not a stretch target, it is a fiction, and the bottom two thirds know it. Healthy is 60 to 80% of reps landing near target. Anything lower and your quota model is the retention risk.
Manager contact. One-to-one completion rate by manager, pulled from calendar data. One analysis found teams holding 85% or more of scheduled weekly one-to-ones saw meaningfully lower turnover (Gangly). A manager carrying nine reports holds fewer of them, and the reps who get skipped are usually the quiet middle performers you can least afford to lose.
The four causes that show up in the data
Published research on why reps leave is reasonably consistent, and it does not lead with pay. In one 2026 compilation, admin overload and burnout accounted for 35% of SDR departures, no clear career path 28%, quota and comp disputes 18%, and manager quality 14% (Gangly). That ordering matches what I see on engagements, with one adjustment: territory quality belongs in the top group and rarely gets its own line because most companies do not measure it.
Admin load
Roundups of Salesforce's State of Sales data put actual selling time at roughly 30 to 40% of a rep's week (Salesmotion), with CRM data entry alone eating something like 20 to 30% (AskElephant). Nobody takes a commission job to fill in fields.
This one is squarely a RevOps fix and it is the cheapest win available. Auto-log calls and emails. Derive close date and stage from activity where you can rather than asking. Kill the custom properties nobody reports on, which at most companies is a third of them. Push enrichment into the record automatically instead of having reps look things up. I have taken eight hours a week off a rep's calendar doing nothing more exciting than deleting required fields and wiring up process automation. If reps are refusing to update the CRM at all, that is a different symptom of the same disease.
Territory design
Somebody has to get the worst list. The question is whether you know who, and whether the quota reflects it. Run yield per assigned account by rep for the last four quarters. If the spread is more than about 2x, you are effectively paying two different jobs the same money and calling the difference performance.
Fix the assignment first, then the quota. Rebalancing a book mid-year is disruptive and worth it. Telling a rep with a picked-over patch to work harder is how you lose them in month six.
Quota credibility
A quota nobody hits is not motivating, it is a signal that management is not serious. Build quota up from capacity and historical attainment rather than down from the board number, and the attainment spread tightens. When the number has to come down from the board, say so out loud and adjust the accelerators, because reps can handle a hard year. What they cannot handle is being told a fiction with a straight face.
Manager span and coaching
Six to eight direct reports is the point where one-to-ones start getting cancelled. Past that, coaching becomes deal inspection and deal inspection becomes a forecast call, and the rep who needed help never gets any. Sales coaching that consists of asking about close dates is not coaching.
Build the early warning report
None of this needs a new tool. It needs one dashboard, refreshed monthly, that nobody is currently building.
Two hours of HubSpot or Salesforce report building. The discipline is in reading it every month and treating a flag as a scheduling trigger rather than a performance note.
One practical warning. Do not hand managers a "flight risk score" that ranks people. It gets treated as a verdict, it changes how managers talk to the flagged rep, and it becomes self-fulfilling. Surface the underlying numbers with context and let the manager form the view.
What to fix first
Rank causes by how many people they affect, not by how loud the last person to leave was.
Admin load first, because it is cheap, it applies to everyone, and it is entirely within your control. Territory second, because the fairness problem compounds every quarter you leave it. Quota third, on the annual cycle, since mid-year changes damage trust more than they help. Manager span last in sequence but not in importance, because fixing it usually means hiring, and hiring takes a quarter.
Compensation goes at the bottom of this list, which surprises people. Raise OTE across a team that is leaving because of territory and admin and you will spend the money and keep the turnover. Comp design matters enormously for whether your top performers stay. It is rarely what is driving your middle performers out.
Want to know what your CRM already says about this?
We pull the territory yield, attainment spread and pipeline creation trends from your own data and show you which reps are at risk and why. Free 30-minute audit, no tooling changes required.
Book an audit →Frequently asked questions
What is a good sales turnover rate for a B2B company?
Below 20% annually for AEs is strong, 20 to 30% is normal, and above 35% means something structural is broken. SDR roles run higher by design because the job is often a stepping stone, so 35 to 45% there is not automatically a problem, provided people are leaving upward into AE seats rather than out of the building. Track the destination, not only the rate.
How do I calculate the true cost of losing a rep?
Add five numbers: recruiting spend, onboarding and training cost, the revenue that would have closed during the vacancy, the productivity you pay for during ramp but do not receive, and manager time spent hiring. Published estimates land between 115,000 and 150,000 dollars for a mid-market AE. Rather than borrowing that figure, use your own: take your average delivered revenue per ramped rep, multiply by the fraction of a year lost to vacancy plus ramp, and add recruiting costs. For most teams the answer is bigger than the benchmark.
Should I count turnover in my annual sales plan?
Yes, as a direct multiplier on capacity. Effective capacity is heads times ramped percentage times expected attainment times one minus turnover. On a team of nine at 30% attrition, that last term removes roughly 400,000 in delivered revenue before anyone misses a number. Plan the backfills before the departures, not after.
Is pay really the main reason sales reps quit?
Usually not. Published research puts admin overload and burnout at the top, career path second, and comp disputes third (Gangly). Pay is what people say in exit interviews because it is the least awkward answer. Look at whether the reps who left had worse territories, lower attainment against the same quota, or fewer one-to-ones than the people who stayed. That comparison usually settles the question.
How early can I see a departure coming in CRM data?
Four to six months, in my experience, using self-sourced pipeline creation as the primary signal. New opportunity count per rep drops well before performance on existing deals does, and well before anyone says anything. A three-month rolling average against each rep's own trailing six-month baseline catches it. Team-level numbers will not, because a strong rep covers the gap for a quarter or two.
The short version
Roughly a third of your sales team will leave this year. That is normal and it is not going away. What is optional is being surprised by it, paying six figures per exit, and fixing the wrong cause afterwards.
Put turnover in the capacity plan as a number, not a footnote. Build the monthly report that shows pipeline creation, territory yield and attainment spread per rep. Then fix admin load and territory design before you touch OTE, because those two affect everyone on the team and cost far less than a blanket raise.
If you want a second pair of eyes on the data, get in touch. We do this inside CRM and RevOps engagements and it is usually the first thing we find.