Someone at the board meeting said "we need NRR at 110 next year," and now it is your number. You own expansion, or you run the customer success team that is supposed to deliver it, and you have a book of several hundred accounts. The goal is clear. What you should do on Monday morning to move it is not.
The reason is structural. Net revenue retention is an outcome metric. It is the sum of hundreds of small account-level decisions: which customer to raise a second product with, which renewal to worry about, which champion left, which division just got a budget. Nobody can act on NRR directly, the same way nobody can act on "revenue." You can only act on the accounts underneath it, one at a time, and most teams make those account decisions without a reason attached to any of them.
This article takes NRR apart into the decisions that move it, names which of those decisions are really research problems, and gives you a test you can run on your own book this week.
What NRR is actually made of
The formula is familiar: take the recurring revenue your existing customers paid at the start of the period, add expansion, subtract contraction and churn, and divide by the starting figure. New logos are excluded on purpose. NRR asks one question: if you stopped selling to anyone new, would the book you already have grow or shrink?
That formula hides four different movements, and each one is moved by a different decision made by a different person. Treating them as one number is how an NRR strategy turns into a slogan.
| Movement | The decision that moves it | Who usually makes it | Is it a research problem? |
|---|---|---|---|
| Churn | Which accounts need an intervention before the renewal, and what kind | CSM, account manager | Partly. Usage data shows risk inside the product. A new executive or a reorg at the customer shows up only outside it. |
| Contraction | Whether a seat or usage drop is a budget cut, a team change, or a product problem | CSM, finance | Partly. The cause of the drop usually sits in public news about the customer, not in your dashboard. |
| Upsell | Which accounts are ready for more of what they already buy | Account manager | Mostly no. Usage and seat data answer most of it. |
| Cross-sell | Which accounts should hear about a second product, which product, and why this quarter | Account manager, strategic AE | Yes, almost entirely. Nothing in your own systems tells you a customer just started needing a product they do not own. |
Upsell is the movement most teams are equipped for, because the evidence lives in the product. Cross-sell is the movement most teams are not equipped for, because the evidence lives in the customer's world. If you have not separated the two before, the difference is laid out in cross-sell vs upsell.
Where the benchmarks actually sit
Before setting a target, it helps to know what normal looks like, because the "120% or bust" figure that circulates in board decks is a top-performer number, not a median.
- The 2026 Aleph and Benchmarkit SaaS benchmarks, built from full-year 2025 data across 342 SaaS and AI-native companies, put median NRR at 102% and the top quartile at 110%. Median gross revenue retention in the same cohort was 84%.
- The same dataset shows a split by pricing model: usage-based companies posted a median NRR of 108% against 98% for seat-based companies.
- SaaS Capital's 2026 benchmarks, drawn from more than 1,000 private B2B SaaS companies, report a median NRR of 103% and a median GRR of 91% for bootstrapped companies between $3M and $20M in ARR.
- High Alpha's 2025 SaaS benchmarks, from more than 800 survey respondents, found that companies above $50 million in ARR generate roughly 60% of new ARR from existing customers, and that beyond roughly $20 million in ARR expansion becomes the dominant growth engine.
Read those together and a pattern shows up. Gross retention can never exceed 100%, and in these cohorts the median sits between 84% and 91%. So every point of NRR above gross retention is expansion: how much the retained customers buy on top of what they already had. In the Aleph and Benchmarkit cohort, the gap between an 84% median GRR and a 102% median NRR suggests roughly 18 points of the typical company's retention number comes from expansion, not from keeping customers. On a $10M book, the eight points between a 102% median and a 110% top quartile is $800,000 of revenue that comes from accounts you already own.
That is the part of NRR this article is about. Retention work matters, and most teams already have a health score, a renewal calendar, and a playbook for it. Expansion is where most books leave money unclaimed, and it is the part where the missing input is a reason.
The account decisions that move NRR
Strip the strategy decks away and an expansion motion comes down to four decisions, repeated for every account in the book, every quarter.
- Which accounts to raise expansion with this quarter. You cannot have a real expansion conversation with every account at once, and the customer will not welcome one every quarter either. Someone has to choose.
- Which product to lead with. If you sell more than one product, the second product is not the same for every customer. A customer that just opened a new region needs something different from one that just replaced its CFO.
- Who to talk to. Your champion bought product one. The buyer for product two may sit in a different function and may never have heard of you.
- Why now. The expansion pitch that lands is the one attached to something that changed at the customer. Without that, the conversation is "have you considered our other product," which the customer has heard at every QBR since they signed.
Most teams answer the first three from inside their own systems: account size, product usage, contract dates, the CRM's contact records. The fourth has no internal source at all. It lives in hiring pages, press releases, earnings calls, leadership changes, funding announcements, and acquisition news. And the fourth is the one that decides whether the first three were right.
This is why a whitespace matrix on its own does not move NRR. It shows you that account 214 owns product A and not product B. That is true for dozens of accounts in your book. It does not tell you which of those dozens has a reason to buy product B this quarter. The empty cell is the start of the question, not the answer. The account whitespace analysis guide covers how to build the matrix; this section is about what to do after it is built.
The reason test: run it on your own book
Here is the method. It needs no software and you can start it today with a spreadsheet and a browser.
An account is worth an expansion conversation when you can finish three sentences about it:
- What changed. Something specific happened at the customer that connects to a product they do not own yet. A new VP of Revenue Operations. A second office in a new country. A job post for the function your second product serves. An acquisition that doubled their headcount.
- Where you saw it. A link. The press release, the job post, the LinkedIn announcement, the earnings transcript. If you cannot point at it, you are guessing, and the customer will hear that you are guessing.
- When it happened. A date. A reorg from two years ago is history. A new leader who started six weeks ago is still deciding what to change, which makes it the window.
No reason, no expansion call. The account goes in the "not yet" pile and gets checked again next quarter. That rule feels harsh until you notice what it replaces: the quarterly ritual of sending the same cross-sell email to every account with an empty cell, and reading the silence as "they are not interested."
A worked example
Take an illustrative book: you sell a sales engagement product and a newer analytics product, and you have 180 customers on the first one only.
Without the test, the expansion plan is "pitch analytics to all 180." With the test, an account manager spends the first week of the quarter checking each account for a reason. Most of the 180 have none this quarter. A minority have something: a customer that posted three RevOps analyst roles in a month (what changed: they are building the function analytics serves; where: their careers page; when: the last 30 days). A customer whose new CRO came from a company that runs on a competitor's analytics tool. A customer that announced an acquisition and now has two sales teams to report on.
Those accounts get a conversation that opens on the change, not on the product. The rest get nothing this quarter, and nobody burns their relationship credit on a pitch the customer has no reason to hear. The hit rate on the short list will be very different from the hit rate on the 180, and that difference is what shows up in NRR a few quarters later.
The categories of change worth looking for are the same trigger events that matter in new business. The full list, and how to tell a real one from noise, is in trigger events in sales.
Why most NRR strategies stall
If the reason test is this simple, why do expansion motions still stall? Three failure modes come up again and again.
The health score only looks inward
A health score built from logins, feature adoption, and support tickets is a good churn radar. It is a poor expansion radar, because the events that create a need for a second product mostly happen outside your product. A customer can be perfectly healthy on product one while hiring the exact team that needs product two, and the health score will never say so.
The research does not scale to the book
The reason test takes a rep somewhere between a few minutes and half an hour per account, depending on how public the customer is. That is fine for a strategic AE with twelve accounts. It breaks for a CSM with 150. So the test gets run on the top ten accounts, which were probably going to expand anyway, and the long tail, where most of the unclaimed expansion sits, gets the generic email. The share of wallet analysis post makes the same point from the other direction: the gap is usually widest in the accounts nobody has time to look at.
The reason never gets written down
Even when a rep does find a reason, it lives in their head or in a Slack message. It does not travel to the CSM who owns the relationship, or to next quarter's review. The account gets researched again from scratch next time, or not at all. A reason that is not attached to the account, with its source and its date, does not compound.
Turning it into a quarterly NRR strategy
An NRR strategy that a team can execute has a cadence, a rule, and a measure. A workable version looks like this.
- Split the target. Take the NRR goal and split it into a gross retention goal and an expansion goal. They are owned differently and they fail differently. Put a name on each.
- Build the whitespace view once. Which customers own which products. This is the universe of possible cross-sell conversations, not the plan.
- Run the reason test every quarter, on every account in the whitespace. Not the top twenty. Every account with an empty cell gets checked for what changed, where, and when.
- Sort into three piles. Pitch now (a reason, a product it points to, and a contact who owns that problem). Not yet (no reason this quarter; recheck next quarter). Skip (the account is at risk, or the reason points away from you, such as a new leader who is bringing a competitor with them).
- Open every expansion conversation on the change. The first sentence is the thing that happened at the customer, with the source. The product comes second.
- Measure reason-to-conversation rate, not emails sent. How many "pitch now" accounts turned into a real conversation, and how many of those turned into pipeline. That number tells you whether your reasons are good. Emails sent tells you only that someone was busy.
If your team already runs a land-and-expand motion, this slots into the expand half of it. The land and expand strategy and customer expansion playbook posts cover the motion around it.
Where the research labour goes
Everything above works by hand. The limit is the one named earlier: the reason test does not scale to a book of several hundred accounts when each check costs a rep real time, so it gets run on the accounts that needed it least.
That labour is what PitchSmart does. You give it the products you sell and the list of accounts you already own, and it researches every account against each product. For each one it returns a plan: pitch, not yet, or skip; which of your products to raise; the buying signals behind that call, each with its source; who to contact; and an opener built on the strongest verified fact. It is allowed to say not yet, and on most accounts in most quarters it should. The point is not more expansion emails. It is knowing which accounts in your book have a reason this quarter, and being able to show the customer where you saw it.
Whether you do it by hand or not, the rule is the same one. Your book is not the problem. The account list you already own has the expansion in it. What is missing is the reason to call each account now, and a net revenue retention strategy that does not produce those reasons is a target without a plan.
