Net Revenue Retention in 2026: The KPI That Miro’s $1.35B Exit Just Redefined

Miro sold for $1.35 billion this week. Three years ago, it was valued at $17.5 billion. That’s a 92% collapse — and it’s not a story about a bad product. Miro still has millions of users. It’s a story about one metric quietly failing while every vanity number looked fine: Net Revenue Retention (NRR).

On September 22, 2026, as OpenAI drops GPT-6 Astra and repositions the entire AI SaaS stack overnight, the Miro correction is the more important story for anyone running a subscription business. GPT-6 changes what’s possible. NRR determines whether you survive long enough to use it.

What Net Revenue Retention Actually Measures

NRR tracks the revenue you retain and expand from your existing customer base over a trailing 12-month period — including upgrades, downgrades, and churn — but excluding revenue from new customer acquisition. The formula:

NRR = (Starting ARR + Expansion − Contraction − Churn) / Starting ARR × 100

An NRR of 100% means you’re treading water: every dollar lost to churn or downgrades is exactly offset by expansion. Above 100% means your existing book is growing even before a single new logo closes. Below 100% means your core business is shrinking, and new sales are just filling a leaking bucket.

This is the metric investors actually price businesses on in 2026 — not top-line ARR growth, not user counts, not total funding raised. Miro’s collapse is the proof point every CFO now points to in board decks.

The 2026 Benchmark Reality

Benchmarks have shifted hard since the 2021 zero-rate era, when “growth at all costs” masked weak retention economics. Here’s where the market actually sits today:

• World-class (top decile): 130%+ NRR — typically infrastructure and platform SaaS with deep workflow lock-in. Instinct’s fresh $1B raise at a $10B valuation reflects exactly this: infrastructure-layer AI still commands premium multiples because switching costs are structurally high.
• Strong / venture-fundable: 110%–125% NRR — the range most Series B/C SaaS companies need to hit to raise at a step-up valuation in 2026’s tighter capital environment.
• Median / survivable: 100%–108% NRR — flat-to-modest growth from the existing base. This is now the “0% grower” cohort that’s become an unexpected career refuge for seasoned B2B execs, according to recent hiring trend commentary — stable, unglamorous, but not collapsing.
• Danger zone: Below 95% NRR — this is where collaboration-tool-era darlings like Miro landed. COVID-era virality drove top-of-funnel growth, but when hybrid work normalized and enterprise buyers consolidated tool stacks, expansion dried up and contraction accelerated. A company can look fine on ARR growth for 18 months while NRR quietly bleeds underneath it — until the exit multiple reflects the truth all at once.

The uncomfortable lesson from the Miro deal: valuation multiples don’t compress gradually when NRR erodes. They snap. A 92% valuation haircut in three years isn’t a market correction — it’s a metric catching up to a business that investors had been pricing on hope instead of retention math.

Why NRR Is the Only Metric Surviving the SaaS Reset

Every hype cycle produces its own vanity metric. In 2021 it was ARR growth rate. In 2023 it was “AI-native” positioning. In 2026, with GPT-6 Astra collapsing capability gaps across the entire AI tooling stack overnight, product differentiation is evaporating faster than ever — which means retention, not acquisition, is the only durable moat left.

Consider what’s happening in parallel today: Microsoft’s MAI-Transcribe-2 just undercut OpenAI and Google on speech AI pricing, which means any SaaS product with a bolted-on transcription feature is about to face margin compression or a forced vendor switch. Companies with weak NRR — customers who were only ever loosely attached via a single feature — will see that instability show up in churn within two quarters. Companies with strong workflow lock-in will absorb the pricing shift and keep expanding.

NRR is essentially a forward indicator of moat durability. TechCrunch’s latest VC sentiment survey confirms investors are now explicitly refusing to fund “generic AI wrappers” — the exact profile of company that posts great logo-count growth but can’t show expansion revenue from existing accounts. NRR is how due diligence teams catch that pattern before the term sheet, and how public/late-stage markets catch it at exit, as Miro’s buyers clearly did.

The Levers That Actually Move NRR

NRR doesn’t improve through better marketing. It improves through four structural levers, and most underperforming SaaS companies are only pulling one or two of them.

1. Land-and-expand product architecture

Companies with 120%+ NRR almost universally sell a product with a natural expansion path built into the architecture — more seats, more usage tiers, more modules. Fresha’s model is the clean 2026 example: it just hit a $1 billion valuation on an $80 million KKR-backed raise by bundling booking and payments into one marketplace. That bundling isn’t just a monetization trick — it’s an NRR engine, because every new payment volume processed is expansion revenue from an already-onboarded customer, not a new acquisition cost.

2. Usage-based or consumption pricing layered on subscription

Pure seat-based pricing caps your expansion ceiling — you can only sell as many seats as a customer has employees. Consumption-based layers (API calls, transactions, storage, AI compute) let revenue expand as usage grows, independent of headcount. This is why infrastructure AI plays like Instinct can post such strong retention numbers even in a compressing application layer.

3. Time-to-value compression in

📊 Track These Metrics

SEMrush gives you the competitive intelligence data behind these benchmarks. See exactly where you stand vs. competitors in real time.

Try SEMrush Free →