What is ServiceNow net revenue retention in 2026?

What net revenue retention actually measures at ServiceNow, and why the number is contested

Net revenue retention is a cohort metric. You take the set of customers that existed twelve months ago, measure what that same set pays you today, and divide. Expansion, downgrades, and churn all land in the numerator; new logos never do. A company at 120% NRR grew its existing book by a fifth without signing a single new customer. That is the number investors care about because it isolates the compounding engine from the sales engine.

ServiceNow does not report it. What ServiceNow reports is a subscription renewal rate, and the difference between those two metrics is the single most misunderstood thing about the company's financials. Renewal rate answers a narrower question: of the annual contract value that came up for renewal this period, what percentage actually renewed? ServiceNow has posted approximately 98% on that measure for a very long stretch of quarters, and management routinely frames it as the most durable metric in enterprise software. It is a genuinely impressive number. It is also nearly silent on expansion.

Here is the arithmetic that makes the point. A 98% renewal rate is compatible with an NRR of 98% — every renewing customer renews at exactly flat, nobody upgrades, nobody adds a module. That same 98% renewal rate is equally compatible with an NRR of 135%, if every renewing customer signs a larger contract. The renewal rate constrains only the downside. It tells you the floor is solid; it tells you nothing about the ceiling. Any analysis that treats "98% renewal rate" as though it were "98% NRR" has the story exactly backwards, because it reads a strong signal as a weak one.

So where does the 115–120% figure come from? It comes from triangulation, not disclosure. ServiceNow publishes a set of adjacent data points that, taken together, let an analyst reconstruct expansion with reasonable confidence. The most useful of these is the customer-count progression across annual contract value bands. ServiceNow discloses how many customers sit above $1 million, $5 million, $10 million, and $20 million in ACV, quarter after quarter. When a customer crosses from the $1M band into the $5M band, that migration is expansion you can see and size. It cannot be a new logo, because new logos do not land at $5 million on day one. Aggregate enough of those migrations and you have a lower bound on expansion inside the installed base.

The second input is current remaining performance obligations — cRPO, the contracted revenue expected to be recognized in the next twelve months. cRPO growth net of estimated new-logo contribution gives you a second read on how the existing book is behaving. The third is management commentary, which has consistently characterized the large majority of net new ACV as coming from existing customers rather than new ones. Layer those three together and the implied dollar-NRR clusters in the high teens above 100%. Different sell-side models land at slightly different points inside that band because they make different assumptions about new-logo ACV, but they converge on the same neighborhood.

For a RevOps practitioner, the lesson generalizes well past ServiceNow. Most enterprise software companies with seat-and-module pricing face the same reporting problem. Consumption vendors — Snowflake, Datadog, the usage-metered cohort — can compute NRR cleanly because revenue is a continuous meter and last year's cohort revenue is a straightforward query. A company selling seats and SKUs sees expansion arrive in discrete renewal events, often as a *different* SKU replacing an old one, which makes cohort math genuinely messy rather than merely inconvenient. That is a real reporting constraint, not evasion. It is also why the consumption names routinely post NRR above 130% in good years and give it back fast in bad ones, while subscription names post a narrower, steadier band. The metric shapes are different because the revenue models are different.

The step-by-step process for reconstructing ServiceNow's NRR yourself

If you want to build this estimate rather than borrow it, the work is tractable. It takes a few hours per quarter once the spreadsheet exists, and the discipline of building it teaches you more about the business than reading someone else's number ever will.

Step one: pull the disclosed cohort data. From the quarterly investor materials and the 10-Q, extract customer counts by ACV band, total and current RPO, subscription revenue, and the stated renewal rate. Record them in constant currency where ServiceNow provides it, because FX swings will otherwise show up as fake expansion or fake contraction. Keep at least eight quarters of history so you can see trend rather than noise.

Step two: separate new logos from the installed base. This is the hardest judgment call and the one that drives most of the variance between models. Total customer count growth gives you gross new logos, but you need an ACV assumption for them. Land deals at ServiceNow's enterprise motion typically start well below the $1 million band, so a reasonable model assumes new logos contribute a modest share of net new ACV and grow into the bands over subsequent years. Management's repeated framing — that the strong majority of net new ACV comes from existing customers — is the sanity check on whatever assumption you pick.

Step three: size the band migrations. Count how many customers moved up a band quarter over quarter. Assign each migration a conservative ACV delta — a customer crossing from the $1M band into the $5M band added somewhere between the band floors, and using the low end keeps your estimate honest. Sum those deltas. This is your visible expansion floor. It systematically understates true expansion because it misses every customer who grew from $1.2M to $1.9M without crossing a band boundary, which is a large population.

Step four: model the invisible expansion. Customers who expand within a band do not show up in the migration count at all. The standard approach applies an expansion rate to the non-migrating population derived from the observed migration rate — if a meaningful share of band-boundary customers crossed, the intra-band population is expanding at a broadly similar clip. This is the softest part of the model and where you should widen your error bars rather than pretend to precision.

Step five: subtract contraction and churn. The disclosed renewal rate gives you the aggregate contraction and churn drag directly. At roughly 98%, that is about two points of drag against gross expansion. Note that renewal rate is measured on the renewing book, not the total book, so applying it as a flat two-point haircut is an approximation — but it is a defensible one given what is disclosed.

Step six: blend and weight by ACV, not customer count. This step is where most amateur models go wrong. The long tail of small customers is numerous but contributes little revenue. A weighted blend that treats a $150,000 customer the same as a $12 million customer will drag your estimate toward the tail's behavior and understate the real number badly. Weight everything by dollars.

Step seven: reconcile against cRPO. Your implied NRR, multiplied against the prior-year base and adjusted for new logos, should roughly reproduce observed cRPO growth. If it does not, one of your assumptions is wrong — usually the new-logo ACV assumption. Iterate until the two views agree within a point or two.

Expansion levers, contraction drags, and the ranges each one moves

The reason ServiceNow's implied NRR sits where it does is that several distinct expansion motions run simultaneously against a small set of drags. Understanding the levers individually matters more than the blended number, because the levers are what a RevOps team at any comparable company would actually operate.

Pro and Pro Plus tier uplift. ServiceNow's packaging ladder — Standard, Pro, Enterprise, and the Pro Plus tier that bundles the Now Assist generative AI capabilities — creates a per-seat price escalator that fires at renewal rather than continuously. Pro Plus carries a meaningful list-price premium over Pro. Even partial attach into a renewing Pro base produces expansion without a single new seat being sold. This is the cleanest, most mechanical expansion vector the company has, and it is the one most directly under sales' control. The trade-off is that a price-driven uplift is a one-time step function per customer, not a compounding line — once a customer is on Pro Plus, that lever is spent for them.

AI agent attach. Agentic capabilities layered onto existing workflows add both fulfiller seats and, where priced by usage, a consumption component. This is the vector management has emphasized most in recent commentary, and it is strategically the most interesting because it is the first ServiceNow motion with genuinely open-ended expansion characteristics. Whether it settles into seat-style attach — predictable, capped — or behaves like a true consumption meter is the single biggest swing factor in the long-run NRR ceiling. Seat-style attach supports the current band. Consumption-style behavior could push the number higher and make it more volatile.

Cross-sell into adjacent product lines. ITSM is the beachhead; HR service delivery, customer service management, integrated risk management, security operations, and the creator/App Engine workflows are the expansion surface. A customer running ITSM across twenty thousand employees who adds HRSD has roughly doubled its platform footprint without changing vendors. These motions have long runways precisely because attach rates in the newer lines remain well below the ITSM base. For RevOps teams, this is the textbook land-and-expand pattern and the reason platform vendors trade at premium multiples relative to point solutions.

Departmental spread through low-code. When finance, legal, or facilities build workflows on the platform, they add fulfiller seats and platform SKUs that were never part of the original IT purchase. This is the quietest expansion vector and the most durable, because it embeds the platform in processes the original buyer does not own.

Against those levers, the drags are real but bounded:

Mid-market pricing friction. Customers in the low-hundreds-of-thousands ACV range have the least negotiating leverage to absorb a tier uplift and the most incentive to push back. Flat renewals in this segment dilute the blended average measurably even though they represent a small share of dollars.

Competitive bundling at the low end. Broad productivity suites that bundle workflow and automation capability into agreements a customer already holds create discount pressure at the bottom of the market. This is rarely a churn event at large accounts — the switching cost of a deeply embedded workflow platform is enormous — but it compresses expansion at the edges.

Public-sector budget pacing. Government contract cycles do not follow commercial renewal logic. Spend resets and appropriation timing can push expansion into later quarters, creating lumpiness that looks like weakness in any single period and disappears on a trailing-twelve-month view.

Currency. Reported expansion on European and Asian cohorts moves with the dollar. Constant-currency figures strip this out, and any NRR model that mixes reported and constant-currency inputs will produce noise it then mistakes for signal.

Seat rationalization. In segments where customer headcount is flat or shrinking — technology being the obvious example — fulfiller seat counts come up flat at renewal, partially offsetting tier uplift. This is the mechanism by which a seat-priced vendor inherits its customers' employment trends, and it is a structural reason seat-based NRR runs below consumption-based NRR across a cycle.

Rough cohort shape, stated as estimates rather than facts: the largest accounts, those above $1 million in ACV, carry the highest expansion rates and near-zero churn — this cohort does the heavy lifting and drives the headline. The mid-market band sits meaningfully lower, healthy but pricing-sensitive. The long tail runs closest to flat, which is where competitive bundling pressure concentrates. Because the blend is weighted by dollars, the top cohort dominates, which is exactly why the blended estimate lands well above what a customer-count-weighted average would suggest.

Where teams get this wrong

The errors here are consistent enough to enumerate, and most of them show up in RevOps dashboards well beyond ServiceNow.

Quoting renewal rate as NRR. This is the dominant error and it runs in the pessimistic direction. Someone reads "98% subscription renewal rate," writes "ServiceNow NRR: 98%," and concludes the company has no expansion engine. It is a category mistake: one metric bounds the downside, the other measures net movement. If you see a precise ServiceNow NRR quoted anywhere without a caveat about it being modeled, treat the entire source as unreliable, because the author did not check the underlying disclosure.

Comparing across revenue models without adjustment. Holding a subscription vendor's NRR against a consumption vendor's is not a like-for-like comparison. Consumption NRR is high and volatile — it inflates when customers scale usage and collapses when they optimize spend. Subscription NRR is lower and steadier because expansion arrives in discrete negotiated events. Judging a subscription business as underperforming because it does not hit consumption-vendor numbers is like judging a bond as underperforming an equity: different instrument, different distribution.

Weighting the blend by customer count. Covered above, but worth repeating because it is so common in internally built dashboards. Most companies have far more small customers than large ones, and small customers behave worse. An unweighted average tells you about your median customer, not your revenue. Both views are useful; only one belongs in a metric labeled *net revenue retention*.

Mixing currencies mid-model. Pulling revenue in reported dollars and cohort counts from constant-currency disclosures produces a number that moves for reasons having nothing to do with customer behavior. Pick one basis and hold it.

Reading a single quarter as trend. Enterprise renewals are seasonal and lumpy. Large contracts land in specific quarters; public-sector cycles bunch around fiscal year ends. NRR is a trailing-twelve-month metric for exactly this reason, and any quarter-over-quarter read is mostly noise.

Confusing gross retention with net. Gross revenue retention caps at 100% and measures pure leakage — churn and downgrades only, no expansion credit. Net can exceed 100%. A business with 92% GRR and 118% NRR has an expansion engine papering over a real churn problem, and the blended number hides it. Both belong on the dashboard; the gap between them is often the most diagnostic number either produces.

Ignoring cohort maturity. A cohort that landed eighteen months ago expands faster than one that landed six years ago, simply because the six-year cohort has already bought most of what it is going to buy. Companies growing new-logo count quickly get an NRR tailwind from cohort mix that decays as they mature. Reading that tailwind as durable improvement in expansion motion is a forecasting error that shows up two years later.

Treating a modeled number as a disclosed one in a board deck. If you carry an estimate into a decision-making document without its provenance and error bars, you have laundered an assumption into a fact. Label it.

Decision framework: what to do with this number depending on why you're asking

The right use of an implied NRR figure depends entirely on the question behind the question. Three distinct audiences ask this, and they need different things from the answer.

If you are evaluating ServiceNow as an investment, the modeled 115–120% band is directionally useful and precisely useless. Do not build a model that is sensitive to the difference between 116% and 119%, because the disclosure does not support that resolution. What you can lean on is the durability of the renewal rate — a sustained move below the high-90s would be the first genuine structural warning — and the spread between cRPO growth and subscription revenue growth, which is the cleanest forward-bookings signal available. Watch band migrations quarterly; they are the highest-signal expansion disclosure the company makes.

If you are benchmarking your own SaaS company against it, first confirm you are computing the same metric. Compute NRR on a trailing-twelve-month cohort basis, dollar-weighted, excluding new logos, in constant currency. Then ask whether your revenue model resembles ServiceNow's. If you sell seats and modules to enterprises, the high-110s is an excellent target and the low-110s is respectable. If you sell consumption, you should be running higher in growth periods and you should expect the number to be less stable. If you sell to SMBs, a number near or slightly above 100% may be entirely healthy for your model — SMB churn is structurally higher and no amount of expansion motion fully offsets it.

If you are a RevOps leader building the expansion motion, the interesting content is not the blended figure at all — it is the lever decomposition. Pull your own book apart into price uplift, seat growth, module cross-sell, and consumption, and size each independently. Most teams discover that one lever is carrying the number and the others are dormant, which is both a risk (single point of failure) and an opportunity (three untouched motions). That decomposition is the actual deliverable; the blended NRR is just the scoreboard.

One more framing worth carrying: pair whatever NRR you compute with gross retention and with the net revenue retention of your newest cohorts specifically. The blended figure is a lagging average dominated by mature customers. The newest cohorts tell you whether the expansion motion you are running *today* works, and that is the number that predicts next year.

Related questions

Does ServiceNow report NRR in its SEC filings?

No. ServiceNow's 10-K and 10-Q filings disclose subscription revenue, remaining performance obligations, and customer counts by ACV band. They do not include a dollar-based net revenue retention figure. The company's headline retention disclosure is the subscription renewal rate.

Is a 98% renewal rate good?

Exceptionally. Most enterprise software companies run materially lower. Sustained high-90s renewal means near-zero involuntary loss of contracted value, which indicates deep operational embedding. It says nothing about expansion, though — renewal rate and net revenue retention answer different questions entirely.

Why do consumption vendors report higher NRR?

Usage-metered revenue expands continuously as customers scale workloads, so cohort revenue compounds between measurement points. Seat-and-module vendors capture expansion only at discrete renewal negotiations. The higher consumption figures come with more volatility — they fall hard when customers optimize spend.

What is the difference between gross and net revenue retention?

Gross retention counts only churn and downgrades and cannot exceed 100%. Net retention adds expansion and can. The gap between them isolates how much expansion is covering leakage — often the most diagnostic pair of numbers on a RevOps dashboard.

Which ServiceNow disclosure best signals expansion?

Customer-count migration between ACV bands. A customer crossing from the $1 million band into the $5 million band cannot be a new logo, so the migration is unambiguous installed-base expansion that you can count and roughly size from public materials each quarter.

FAQ

What is ServiceNow's net revenue retention in 2026?

There is no official figure. ServiceNow does not publish dollar-based NRR. Analyst models that reconstruct cohort expansion from disclosed ACV-band customer counts, cRPO growth, and management commentary imply something in the neighborhood of 115–120% for the period. Any source quoting a precise number without labeling it as modeled is misrepresenting the underlying disclosure, and you should discount it accordingly.

Why doesn't ServiceNow just publish the number?

Partly because its revenue model does not compute cleanly into cohort NRR — expansion arrives as new SKUs at renewal rather than as metered usage — and partly because the renewal rate is a stronger, simpler story for a company whose pitch is durability. Publishing a new metric also creates a new obligation to keep publishing it through weaker periods.

Can I use the modeled 115–120% figure in a board deck?

You can, if you label it as an analyst estimate rather than a company disclosure and state the range instead of a point. Carrying it in unlabeled turns an assumption into an apparent fact, and the first person who checks the filings will find no such number, which costs you more credibility than the data point was worth.

How does this compare to other enterprise platform vendors?

A blended figure in the high 110s sits comfortably in the upper tier for enterprise subscription software, above the broad SaaS median and below what consumption-metered vendors post in expansion cycles. The comparison is only meaningful against companies with similar pricing architecture — seats and modules versus usage meters produce structurally different distributions.

What would make ServiceNow's NRR go up or down from here?

Up: broader AI agent attach behaving like a consumption meter rather than a seat add-on, faster cross-sell of newer product lines into the ITSM base, and continued upward migration through the ACV bands. Down: sustained mid-market pushback on tier pricing, flat customer headcount suppressing seat growth, competitive bundling compressing the long tail, and currency drag on international cohorts.

How should a RevOps team apply any of this internally?

Compute your own NRR on a trailing-twelve-month, dollar-weighted, new-logo-excluded basis, then decompose it into price uplift, seat growth, cross-sell, and consumption. Pair it with gross retention and with the NRR of your newest cohorts. The decomposition tells you which expansion motion is actually working; the blended figure only tells you the score.

Sources

• https://investors.servicenow.com/
• https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001373715&type=10-K
• https://www.servicenow.com/products/pricing.html
• https://www.gartner.com/en/information-technology
• https://www.forrester.com/research/
• https://www.mckinsey.com/capabilities/growth-marketing-and-sales/our-insights
• https://www.bvp.com/atlas/state-of-the-cloud
• https://openviewpartners.com/blog/
• https://www.fasb.org/asc606

Related on PULSE

• [What is NRR (Net Revenue Retention) and what is a healthy benchmark?](/knowledge/q12690)
• [Why is net revenue retention compressing and how do you fix it?](/knowledge/q12975)
• [How do you operationalize net revenue retention?](/knowledge/q12864)
• [What is the Net Revenue Retention benchmark for B2B SaaS?](/knowledge/q12021)
• [How do you structure RevOps OKRs that tie to net revenue retention?](/knowledge/q10436)
• [Should I hire a fractional CRO if my net revenue retention is below 100 percent?](/knowledge/q16082)