Net revenue retention: the one number investors obsess over

In 2021, a growth-stage SaaS company told investors it was growing 80% year over year. The investors passed. The company had a net revenue retention of 72% — which meant that for every dollar it had at the start of the year, only 72 cents remained after churn and downgrades, before any new logo was counted. All that 80% growth was being spent just keeping the bucket from draining. The business was a treadmill, not a flywheel.

NRR is the number that shows you which one you have.

What NRR actually measures

Net revenue retention (NRR) — sometimes called net dollar retention (NDR), the terms are interchangeable — measures how the annual recurring revenue (ARR, the annualized value of active subscription contracts) from a fixed cohort of existing customers at the start of a period changes by the end of it.

The key word is existing. You freeze a cohort — say, everyone who was a paying customer on January 1 — and you track what happens to their revenue over the next twelve months. Expansions and upgrades add to it. Churn (customers who cancel entirely) and downgrades (customers who drop to cheaper plans) subtract from it. New customers acquired during the year are not counted at all.

The formula is straightforward:

NRR = (Starting ARR + Expansions - Downgrades - Churn) / Starting ARR

Express it as a percentage. If you started with 1,000,000 in ARR from that cohort, added 250,000 in expansions and upgrades, and lost 150,000 to churn and downgrades, your NRR is:

(1,000,000 + 250,000 - 150,000) / 1,000,000 = 1,100,000 / 1,000,000 = 110%

110% means the existing customer base, left alone, grew by 10% over the year. You did not need a single new logo to produce that growth.

The leaky-bucket metaphor, made precise

The canonical mental model for a SaaS business is a leaky bucket: you pour new customers in at the top, and some fraction leak out the bottom every month through churn. Growth happens when you pour faster than you leak. That model is fine as far as it goes, but it misses an important dimension — the size of the drops.

NRR captures whether the drops that stay in the bucket are getting larger or smaller over time. A business with 80% NRR has a bucket where the drops that survive are actually shrinking — every year the existing base contracts by 20%, and you are in a race against contraction. A business with 110% NRR has a bucket where the surviving drops expand — each year the existing base grows by 10% on its own, and new customer acquisition piles on top of an already-rising floor.

The compounding effect is severe. At 80% NRR, a cohort worth 1M ARR is worth 410,000 after five years. At 110% NRR, that same cohort is worth 1,610,000. The growth rate on new business is the same in both cases; the underlying engine is completely different.

Why 100% is not just a round number — it is a regime change

Below 100% NRR, your business has a structural problem that sales cannot fix. It can only paper over it. Every new dollar you bring in starts its clock, and before long it joins the base that is already shrinking. The implication: to grow ARR 20% year over year with 80% NRR, you need to acquire 40% of your current ARR in new logos every single year. That is an exhausting, expensive machine.

Above 100% NRR, the arithmetic inverts. Existing revenue expands. New customer acquisition is additive to a growing base rather than compensatory to a declining one. A company at 110% NRR that also signs new business is effectively getting compounding growth from two directions simultaneously.

This is why investors treat NRR as a quality signal rather than a growth signal. A company growing 30% year over year with 90% NRR is telling you its growth depends almost entirely on relentless new sales. A company growing 30% with 115% NRR is telling you its customers love the product more over time, expand into it, and stay. Those are two completely different businesses wearing the same growth-rate number.

What drives NRR above 100%

High NRR is not something you engineer in the pricing spreadsheet. It comes from product architecture and customer success in roughly equal parts.

Expansion revenue — the numerator advantage — comes in three flavors: seat expansion (users adding colleagues), usage-based expansion (workloads growing into the platform), and tier upgrades (customers outgrowing a plan and moving up). The best SaaS businesses have at least two of these three running simultaneously. Data platforms like Snowflake and Databricks benefit from all three: more users, more compute, higher tiers. That is how they have historically reported NRR above 130%.

Churn is the denominator problem. And churn is almost always a product problem wearing a sales costume. Customers who leave in year one usually found that the product did not fit their workflow. Customers who leave in year two usually found a competitor that fit better. The fix is not a better renewal playbook — it is identifying which customers are under-adopting the product before they decide to leave, then helping them get value faster.

The best NRR programs treat expansion and retention as a single motion: the same customer success team that drives adoption is also identifying natural moments to introduce a higher tier, an add-on module, or additional seats. That sequencing matters. An expansion conversation with a customer who is already getting value is easy. The same conversation with a customer who has barely logged in is tone-deaf and accelerates churn.

The cohort lens reveals what annual snapshots hide

There is a subtlety in NRR that trips up first-time analysts: it is a cohort metric, not a calendar-year metric. When someone says “our NRR is 110%”, they mean that the group of customers who existed at a specific point in time generated 10% more revenue from that base twelve months later.

This is different from comparing total ARR at two points in time, which conflates new customer acquisition with existing-customer behavior. A company that tripled its sales team and signed hundreds of new logos while simultaneously experiencing 85% NRR would show excellent total ARR growth. The NRR number exposes the silent rot that the top-line hides.

The cohort framing also reveals a timing artifact worth knowing: NRR can look healthy on aggregate while hiding deteriorating cohorts. A 2022-vintage cohort might be at 120% NRR today. A 2024-vintage cohort, from a period when the company was growing fast and accepting less-fit customers, might be at 85%. Average them together and you see something that looks fine. Build the cohort waterfall — each row is a vintage, each column is months since acquisition — and the problem cohort sticks out like a sore thumb.

The 120% benchmark and what it implies

In enterprise SaaS — software sold on annual or multi-year contracts to companies rather than individuals — 100% is table stakes, 110% is good, and 120%+ is elite. These are not arbitrary thresholds. They reflect the practical reality that a business at 120% NRR needs to add only a modest stream of new customers to grow aggressively, because its existing base is already compounding at 20% annually.

The math becomes dramatic over a five-year horizon. A business at 120% NRR that stops signing new customers entirely would still have an ARR base 2.5 times larger than it started with after five years. That scenario never happens in practice, but the underlying compounding is real and it explains why investors will accept a lower current growth rate from a business with exceptional NRR: the durability is already priced into the structure.

Consumer SaaS and product-led growth businesses tend to have lower NRR benchmarks — 100 to 110% is considered strong — because individual users churn more frequently and expansion per seat is limited. The metric is most informative in businesses where the unit of sale is a company, not a person, because company needs grow over time in a way that individual needs often do not.

How to read NRR when you are not the CFO

If you are a product manager, an analyst, or an engineer, NRR is not just a metric you hand to finance. It is a diagnostic for the product.

A sudden drop in NRR often traces back to a specific cohort of customers who were sold on a use case the product does not actually serve well. That is a product scoping problem. A sustained NRR above 110% usually means the product has natural upsell surface — there is something customers want more of as they get comfortable with it. That shapes what you build next: lean into whatever is driving expansion rather than spreading effort evenly across the roadmap.

NRR also sets the floor on your customer success team’s headcount model. If a single customer success manager handles 2M ARR in book of business, and the goal is 110% NRR, you can model exactly how many expansion conversations and at-risk interventions that manager needs to drive per quarter. The metric becomes a staffing input, not just an investor slide.

The one number, actually explained

Net revenue retention is a deceptively simple formula. The numerator is what your existing customers are worth today. The denominator is what they were worth a year ago. The ratio tells you whether your product compounds or decays in the hands of the people who already chose it.

Everything else in SaaS — growth rate, payback period, lifetime value, sales efficiency — is downstream of this number. A great sales team working against bad NRR is running in sand. A mediocre sales team working with great NRR is compounding quietly while the market watches the one with the bigger growth-rate headline.

Investors obsess over it because it is one of the few metrics that cannot be faked with short-term tactics. You can gin up a growth rate with aggressive discounting. You cannot gin up NRR. The customers either stayed and spent more, or they did not. That is the signal.