CHURN IS DEAD
You Don't Own NRR. You Rent It From Pricing.
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.
The 60-Second Version
Most of your NRR is set by your pricing model before a single CSM opens their laptop, and agentic AI is now eating the seats that model depends on.
The Inherited Number: split your portfolio's NRR into the part fixed by pricing architecture and the part your team can actually influence, then use the gap to earn a voice in packaging.
This week, run the decomposition on ten accounts and find out how much of your target is unwinnable by execution alone.
Thirteen points.
That is the gap the Benchmarkit *2026 SaaS Performance Metrics* report puts between the median net revenue retention of usage-based SaaS companies (around 108%) and seat-based ones (around 95%). Same discipline. Same buyers. Same economy. A structural thirteen-point spread that widens every year, set entirely by how the contract is shaped.
No CSM has ever closed that gap with a better QBR.
And yet every planning cycle, a VP of CS walks into a room and gets handed an NRR target as if it were a measure of how good the team is. As if the number were a scoreboard for outreach, adoption plays, and relationship depth. It is not. Most of your NRR was decided by people you have never met, in a pricing meeting you were never invited to, using a model you have no authority to change.
You don't own NRR. You rent it. And the landlord is Pricing.
The industry got the ownership question backwards
The accepted wisdom is clean: CS owns NRR because CS owns the relationship. Retention is the output of trust, adoption, and proactive risk management, so if the number is soft, the playbook is soft.
That logic held when seats were a decent proxy for value. More users meant more entrenchment meant more expansion. The pricing model and the value model pointed the same direction, so nobody had to separate them.
Agentic AI just severed that link.
When a customer deploys AI agents that collapse the team using your product, seat count drops while satisfaction stays high. The value the customer receives can rise as the seats fall. Your contract was built to capture headcount, and headcount is exactly what your customer is now automating away.
The cleanest public evidence sits in customer service, where the seat-to-work link snapped first. Klarna said publicly in 2024 that its AI assistant was doing the work of roughly 700 full-time agents. Salesforce ran its own support organisation from around 9,000 people toward roughly 5,000 as Agentforce absorbed tier-one and tier-two volume, and its CEO has spoken openly about not backfilling those roles. Intercom rebuilt its entire pricing around its Fin resolution bot because per-seat support licensing stopped mapping to the value being delivered.
Those are service seats. But the same physics is now moving up the stack into the knowledge-worker tools that enterprise CS sells. The seat was always a billing convenience standing in for value. AI just exposed the stand-in.
The number you were handed was already spent
Here is the part nobody puts on a slide, so let me do the arithmetic the discourse keeps ordering you to do without ever showing you.
Take a book of 40 enterprise accounts, all seat-based. Suppose your gross retention floor, before any expansion, sits at 88% for the portfolio. That 88% is not a CS outcome. It is a function of switching cost, contract structure, and how deeply the product is wired into the customer's stack. You inherited it the day the account signed.
Now layer expansion. In a healthy seat-based book you might add 9 to 10 points of expansion through natural seat growth as the customer's own headcount grows. Call it 9. That lands you around 97%. Notice: most of that expansion is the customer hiring, not you selling. It happens whether your QBRs are brilliant or cancelled.
So before your team does anything skilful, roughly 88 points are architecture and another handful are the customer's own org chart. The genuinely CS-influenceable slice, the part that moves because a human noticed risk early, ran a save, drove a real cross-sell, might be 4 to 6 points on either side of that number.
That is the inherited number: the portion of your NRR fixed by pricing architecture and customer headcount before execution enters the room. In a seat-based book it is frequently 90-plus points of a 95-point result. Your team is being measured on a metric it barely moves.
Now introduce the seat cannibal. One account of the forty deploys agents and cuts 30% of its seats while raving about the product. On a book that size, that single event can pull two or three points off your portfolio NRR. There is no adoption play that reverses it, because adoption was never the problem. The customer adopted so completely they automated the users.
Run that math across your real book. Most orgs have never done it. When they do, the target and the pricing model turn out to have been fighting each other the whole time.
The 1.2-human warning
On Lenny's Podcast in early 2026, Jason Lemkin described taking his go-to-market org from roughly eight or nine humans down to "1.2 humans plus 20 AI agents" with the same business performance, his AI officer spending about a fifth of her time orchestrating the agents.
Most people heard that as a headcount story. It is a pricing story.
If that ratio becomes the norm on your customers' side, the seats that fed your expansion are gone. Not because your customer is unhappy. Because your customer did to their org what Lemkin did to his. The workload didn't disappear. It moved from people you licensed to agents you don't. Your contract has no line item for the outcome, only for the vanished headcount.
Lemkin's own numbers carry the qualifier that matters: he still needs 1.2 humans orchestrating. The work compresses, it doesn't evaporate. That is the shape of the trap. The value is still being delivered. Your pricing model just stopped being able to see it.
The Inherited Number
Here is the framework. It does one thing: it stops you defending a number you don't control and points your energy at the lever that actually moves it.
1. The Architecture Line
Decompose your portfolio NRR into two buckets: architecture-fixed and CS-influenceable.
Architecture-fixed is everything determined before you act. Gross retention floor set by switching cost. Baseline expansion driven by the customer's own headcount growth. The pricing model's structural ceiling.
CS-influenceable is what changes because a human intervened. The save on an account that was genuinely going to leave. The cross-sell that wouldn't have happened without you. The seat expansion you actively drove rather than passively received.
Draw the line for each account and total it. You will land on a figure like: of our 95% target, roughly 90 points are architecture and 5 are us. Now you have a real conversation with finance instead of a guilt trip. You are no longer arguing about effort. You are showing them the ceiling.
2. The Seat Cannibal
Flag every account contracting on healthy signals. High usage, strong sponsor, positive sentiment, shrinking seats.
These are not churn risks. They are proof that satisfaction and seat count have decoupled. When you tag them separately, two things happen. Leadership stops reading them as CS failures. And you build the evidence file for the pricing argument, because a seat cannibal is the single clearest demonstration that your model is mispriced against the value delivered.
When you find a seat cannibal, you don't run a retention play. You run a repricing conversation. More on the script in a moment.
3. The Outcome Ledger
Stop instrumenting product usage as your value story. Start instrumenting the business result the customer actually bought.
Not logins, not feature adoption. Hours saved. Tickets deflected. Revenue influenced. Cycle time cut. The number the customer's own economic buyer would recognise as the reason the tool exists.
This is the hardest of the four and the most valuable. It is the asset that wins renewals today regardless of pricing, and it is the evidence base finance needs before it will consider a model with a higher retention ceiling. Consumption leaders retain above 120% precisely because their product bills against the outcome and expands without a renewal negotiation. You are building the ledger that makes that shift arguable.
4. The Pricing Seat
Turn the ledger into the artifact that earns CS a voice in packaging.
The real NRR lever does not live in your outreach cadence. It lives in the pricing meeting. McKinsey's work on SaaS pricing found best-in-class packaging correlates with roughly 16 points of higher NRR, more than almost any operational CS improvement can deliver. That is the room you were never invited to. The Outcome Ledger is your invitation.
Walk in with per-account evidence that value is decoupling from seats, and you are no longer the person who missed a target. You are the person holding the data that determines whether the company's next pricing model has a ceiling of 95% or 108%.
What you actually say to procurement
The framework fails at exactly one point if I leave it abstract: the moment a seat cannibal's procurement team tells you they're cutting 30% of licences and you have to open your mouth.
Don't defend the seats. You will lose, and you should lose, because the seats really are surplus. Reframe the unit.
You say something close to this: "You're right that you need fewer seats, and honestly your team should. What's changed is that the value you get from us stopped being about how many people log in. Last quarter this platform deflected 4,000 tickets and saved your ops team roughly 2,100 hours, and that number went up while your seat count went down. I don't want to bill you for chairs nobody sits in. I want to structure this so you pay for the work getting done, which is growing. Let me bring you a packaging option that tracks the outcome instead of the headcount, so your cost follows your value instead of your org chart."
That sentence does three things. It concedes the seat cut, which builds trust. It moves the negotiation onto the ledger, where your numbers are rising. And it hands the customer a reason to want a consumption or outcome model, which is the model with the higher ceiling. You just turned a contraction into the opening argument for the pricing change your own company needs to survive.
You cannot always close that in the room. But you can carry it back, and now you're carrying data, not an excuse.
The audit, on Monday
Don't boil the ocean. Ten accounts.
1. Draw the Architecture Line on ten accounts. For each, estimate the gross retention floor and the customer-headcount-driven expansion. Everything else is you. Total it. Get your inherited number.
2. Tag every seat cannibal in that ten. Any account contracting while healthy. Note the seat delta and the reason. If even one shows up, you have your first exhibit.
3. Pick one outcome metric and start the ledger. One number the economic buyer cares about. Hours, tickets, revenue. Instrument it for those ten accounts this quarter. Imperfect data beats no data.
4. Book fifteen minutes with whoever owns pricing. Not to pitch. To show the Architecture Line and ask one question: how much of the NRR target do we believe execution can actually move, and how much is the model? Watch the room go quiet.
The room you're actually fighting for
The reason this argument matters is not vindication. It is leverage.
When you can prove that most of your number is inherited, two doors open. Finance stops setting targets that assume CS is the whole retention engine. And you earn standing in the one conversation that raises the ceiling instead of scraping the floor.
The CSM who spends 2026 running better adoption plays on a seat-based book is optimising the four points they influence while the ninety-one they inherited erode underneath them. The CSM who spends 2026 building the Outcome Ledger and walking it into the pricing meeting is changing the architecture itself.
One of those is renting. The other is buying a stake in the building.
The question I'd put to every VP of CS heading into planning is simple, and it will make the room deeply uncomfortable: of the number you just agreed to, how much can your team actually move, and who in this company is accountable for the rest?
If nobody can answer, you didn't get a target. You got a rent bill.
By Kuber Sethi · All issues · Subscribe