CHURN IS DEAD
The Expansion Recession Is Here. Stop Calling It a Retention Problem.
11 min read · Revenue
Archive note: This issue predates the evidence ledger introduced in August 2026. Treat uncited benchmarks and examples as editorial analysis, not independently verified findings.
Median private SaaS NRR drifted from around 105% in 2021 to roughly 101% in 2024. The 15th annual SaaS Industry Survey, the one SaaS Capital and ChurnZero put out in early June, finally said the quiet part with data attached: most of that four-point slide happened while gross retention barely moved.
Read that again, because it changes who's responsible.
Gross retention holding flat means customers aren't leaving. Net retention falling means they've stopped buying more. Those are two completely different failures, and the second one is not yours.
Yet NRR sits on the goals slide of nearly every VP of CS I know. They put it there proudly. It was the number that finally got Customer Success into the board deck, the proof that the function drives revenue and not just goodwill. And now that same number is sliding for reasons no QBR can touch, and the person who volunteered to own it is about to get asked why.
This is an expansion recession dressed up as a retention conversation. SaaS Capital's own read is blunt: upsell, cross-sell, and price increases have all stalled. You Mon Tsang at ChurnZero called AI "the second punch" on top of the macro shock, with CIOs now openly asking whether they even need to renew at the same level. None of that is a value-realization problem. It's a pricing-power and roadmap problem.
And if you accept a raw NRR target this planning cycle, you have signed up to be blamed for it.
The broken state: one number doing three jobs
Here is what the falling NRR actually looks like inside a real enterprise book.
You have accounts that adopted everything you sold them. Usage is healthy, the champion is engaged, the renewal is locked. They love you. They are also done. There is no new module, no new use case, no new tier that gives them a reason to spend another dollar this year. You did your job perfectly and the account is flat.
You have accounts where the only growth on the page came from a contractual price escalator that finally got enforced after years of being waived when inflation was near zero. The book grew. You ran no expansion play. The clause did the work.
And you have a thin slice of accounts where something genuinely new landed. A second team picked up the product for a use case nobody had scoped at signing. Real net-new value, real net-new spend.
All three of those roll up into one blended NRR number. Leadership reads that number as a single signal of how well CS is driving growth. When it goes up, you get credit for all of it, including the price clause you didn't negotiate. When it goes down, you inherit the whole drop, including the expansion the roadmap never gave you anything to sell.
That is the trap. One number is doing three jobs, and you only control one of them.
The fixed state: you bring three numbers, not one
Now picture the version where you've already pulled it apart.
When the planning conversation comes, you don't walk in defending a blended 104% that drifted to 101%. You walk in with the 104% already decomposed into the part CS controls, the part that's contractual mechanics, and the genuine new-value slice. You point at the genuine slice and say: that is the number you can hold me to. The rest is pricing strategy and roadmap, and those report to people who aren't me.
That is the whole game. Not deflection. Decomposition. You are not refusing accountability. You are refusing to be accountable for three different things under one label.
The framework that gets you there has four layers.
The Expansion Ceiling
The Expansion Ceiling is the maximum NRR your accounts can reach given the pricing architecture and product surface you actually have to sell. It is set above your head. CSM effort moves you toward the ceiling. It does not raise the ceiling. Confusing those two is how good CS leaders end up owning a board-level failure.
Underneath the ceiling, every dollar of net retention falls into one of four layers.
1. Floor (GRR). This is the churn you genuinely control. Adoption, value delivery, champion health, the saves you make when an account wobbles. When gross retention holds flat while net retention falls, your Floor is doing its job. This is the evidence that the problem is upstairs, not in your book.
2. The Lever Gap. This is the expansion that would require a new use case, a new SKU, or a price point the roadmap and pricing model don't yet offer. It's the whitespace you can see and cannot sell, because the product to fill it doesn't exist or the packaging won't let you. The Lever Gap is not your failure. It is the gap between what the account could buy and what the company built. Naming it is how you stop running expansion plays into a wall.
3. Mechanical Expansion. Contractual escalators, seat true-ups, usage overages. Real revenue. It hits the bank. But it is administrative enforcement of terms that already existed, not strategic growth you manufactured. SaaS Capital is right that compounding a waived 3 to 5% escalator is meaningful money that needs no new sales cycle. Just be honest about what it is. It's accounts-receivable discipline wearing a growth costume.
4. Earned Expansion. Genuine new-value upsell. A use case you actually unlocked, a second team you actually landed, a problem the customer didn't know they had until you showed them. This is the only NRR you should be measured on. It is also, in almost every book I've seen, the smallest of the four slices.
That smallest slice has a name. Earned NRR: the portion of net retention that came from a real new use case, stripped of escalators, true-ups, and price hikes.
Earned NRR is the number you should be fighting to put on your goals slide. The blended one is the number you should be fighting to take off it.
Show me the arithmetic
A concept you can't compute is a slogan. So let's actually do the subtraction on a realistic book.
Say your blended NRR last year was 104%. That four points of net expansion above 100 is what everyone is celebrating. Pull it apart.
Start with the price escalators. You go into billing and you find that 2.4 points of that came from contractual annual increases, most of which were boilerplate clauses your predecessor negotiated and that finally got applied. That's Mechanical. You didn't sell it. The contract did.
Next, the true-ups. Another 1.1 points came from seat reconciliations and usage overages that triggered automatically because customers grew their own headcount or volume. That's also Mechanical. The customer's growth drove it, not your expansion motion.
Now you're at 3.5 of your 4 points accounted for, and not one dollar of it required a new use case. What's left is 0.5 points of genuine net-new value: the accounts where a second team adopted, where you landed a module nobody scoped at signing, where you created spend that wasn't latent in the contract.
Your blended NRR is 104%. Your Earned NRR is 100.5%.
That is the number leadership has been crediting to your expansion playbook. Half a point. When the blended figure slides from 104% to 101% next year, almost all of that drop will come from escalators that hit their cap and true-ups that flattened as customers stopped growing seats. Your Earned NRR may not have moved at all. But unless you've done this arithmetic, you will eat the entire three-point decline as a CS performance miss.
The numbers above are illustrative, not from any account I work. The point is the method. When you run it on your own book, the proportions will shock you, because almost nobody has separated the price clause from the win.
Where the data actually lives
The most common objection I get to this is that the data is too tangled to separate. It isn't. Every dollar of expansion has a transaction behind it, and the transaction tells you which layer it belongs to.
For Mechanical from escalators: pull every account where the year-over-year increase maps to a contractual percentage in the order form. If the bump equals the escalator clause, it's Mechanical. Your billing system and your CLM have this. It's a join, not a judgment call.
For Mechanical from true-ups: pull every line where the quantity changed but the rate and the SKU did not. More seats of the same product at the same price is a true-up. Same usage tier, higher consumption, is an overage. Your billing exports tag these.
For Earned: this is the residual, but verify it, don't infer it. Pull every expansion where a new SKU appeared on the account that wasn't there before, or where a new cost center or business unit shows up as the buyer. A new product code or a new internal buyer is the signature of a genuinely new use case. That's your Earned line.
Anything left over that you can't classify, classify conservatively. Put it in Mechanical, not Earned. The discipline is to never give yourself credit for a dollar you can't trace to a use case you unlocked.
For Floor: that's your GRR, computed the normal way, dollars retained over dollars up for renewal before any expansion. If it's flat year over year, that's your headline. It's the proof that the function is healthy and the ceiling is the problem.
This is a one-day exercise for a competent CS Ops analyst. The reason it rarely gets done isn't difficulty. It's that the blended number flatters everyone, and nobody volunteers to make their own contribution look smaller. Until the year the blended number turns down, and suddenly the decomposition is the only thing standing between you and a target you can't hit.
What the discourse gets wrong about expansion
The prevailing advice when NRR dips is to expand harder. Map the whitespace. Multi-thread the champions. Run tighter value-realization QBRs. Build the expansion muscle.
All fine in a book that has room above the ceiling. Useless in a book that's hit it.
The Lever Gap is the part the discourse refuses to name, because naming it admits that the constraint is the product and the price, not the CSM. When an account has adopted everything you sell, at the tier that fits them, and there is no new module on the roadmap that solves a problem they actually have, no amount of whitespace mapping conjures revenue. You can multi-thread to the CEO. There is still nothing new to put in front of them.
This is structural. Per-seat SaaS in a flat-headcount economy expands by adding people who aren't being added. Usage and outcome models keep expanding because consumption keeps rising. If your company is on per-seat pricing and your customers have stopped hiring, your expansion ceiling dropped, and not one thing in your control caused it. That's a pricing architecture decision made years ago in a room you weren't in.
Gainsight, under Chuck Ganapathi, has rebranded the whole category around "Retention-as-a-Service," invoking Sam Altman's line about the "fast fashion era of SaaS" where outcomes are the only moat. He's right that outcomes are the moat. But outcomes are a product-and-pricing achievement, not a CSM workflow. You can automate every health check and QBR draft on earth and it won't add a SKU to the roadmap that gives a maxed-out account a reason to buy.
The move, before someone makes it for you
Here is what to do this quarter.
1. Decompose last year's NRR into the four layers. Floor, Lever Gap, Mechanical, Earned. Use the data sources above. Compute your Earned NRR as a hard number. Do this before any target gets set, not after you've missed one.
2. Quantify the Lever Gap explicitly. For your top accounts, list the expansion you can see and cannot sell because the product or the pricing doesn't support it. Total it. That number is your evidence that the constraint lives upstream. Hand it to product and pricing as a demand signal, not as an apology.
3. Walk into the target meeting with three numbers, not one. Show the blended figure, then show the decomposition. Say plainly: hold me to Earned NRR, hold the contract to Mechanical, and hold product and pricing to the Lever Gap. Refuse to let one label carry three owners.
4. Separate Mechanical enforcement from strategic expansion in how your team is measured. Enforcing a waived escalator is worth doing. It's real money. But it is operations, not growth, and a CSM who's good at it should be credited for discipline, not mistaken for a closer who'll save you when Earned expansion stalls.
5. Stop volunteering for the blended number. The instinct that got CS into the board deck was to own the biggest revenue figure available. The instinct that keeps you there is to own the one you can actually move.
I sat in a planning conversation not long ago where a CRO opened with the blended NRR slide and the line every CS leader has heard: "We need this number back up." The CS leader in the room had done the decomposition. They put the four-layer breakdown on the screen. Earned NRR was a fraction of the blended figure, the Floor was flat and healthy, and the Lever Gap, quantified, was larger than the entire net-expansion line.
The room went quiet, then the CFO asked the only question that mattered: "So how much of this is actually a pricing decision we haven't made?"
That is the question you want asked in your meeting. It only gets asked if you bring the breakdown that forces it. Bring the blended number and you'll spend the next year explaining a decline you couldn't have prevented. Bring the decomposition and you turn an interrogation into a strategy conversation about the ceiling, which is the only conversation that actually raises it.
The expansion recession is real and it isn't ending soon. The CS leaders who survive it won't be the ones who expanded hardest. They'll be the ones who knew exactly which slice of the number was theirs, and said so out loud before anyone could decide it for them.
By Kuber Sethi · All issues · Subscribe