Net Revenue Retention by Software Category: A Reader's Guide

Net revenue retention (NRR) is the single most-quoted number in software earnings calls, and the single most-abused. The headline figure — say, 115% — tells you the average existing customer is spending 15% more than a year ago, after accounting for churn and downgrades. That sounds great until you realize a 115% NRR can mean three completely different things depending on the category, the pricing model, and how the company defines "existing customer."

Here's a framework for reading NRR by software type, what numbers should make you nod, and what numbers should make you check the footnotes.

What NRR Actually Measures (and Hides)

NRR compares this period's revenue from a cohort of customers to what that same cohort paid a year ago. Above 100% means net expansion beat net contraction. Below 100% means the existing book is shrinking.

Three things NRR does not tell you:

New logo growth. A company with 130% NRR and zero new customers is in slow decline.
Where the expansion came from. Seat growth, usage growth, price hikes, and cross-sell all roll into one number. They have very different durability.
The denominator games. Some companies report NRR only for customers above a revenue threshold (e.g. >$100k ARR). That excludes the SMB churn that usually drags the average down. Always check the cohort definition.

With that in mind, here's how to calibrate by category.

NRR Benchmarks by Software Category

Infrastructure & usage-based platforms (Snowflake, Datadog, MongoDB, Cloudflare). Healthy is 115-130%. Best-in-class historically printed 150%+, but those numbers came from an era when cloud budgets were uncapped. The honest expectation today is 115-125%. Below 110% in this category is a warning — it means customers are actively optimizing spend rather than letting consumption grow with their business. Above 140% on a sustained basis is increasingly rare and worth scrutinizing for one-time migration tailwinds.

Horizontal application SaaS (CRM, HR, collaboration — Salesforce, Workday, Atlassian, Asana). Healthy is 105-115%. The expansion engine here is seats and edition upgrades, both of which are bounded by headcount and feature ceilings. A mature horizontal app printing 120%+ is either riding a real product cycle or pulling forward via aggressive price increases. Below 100% is a clear warning that the install base is shrinking or downgrading.

Vertical SaaS (Veeva, Tyler, Procore, Toast). Healthy is 108-118%. Vertical players have smaller addressable seat counts per customer but high attach rates on adjacent modules. The expansion comes from selling the second and third product, not seat growth. Watch for modules-per-customer disclosure — that's the real signal underneath the NRR print.

Cybersecurity (CrowdStrike, Zscaler, Palo Alto's NGS, SentinelOne). Healthy is 115-125%. Security buyers consolidate vendors and add modules in waves. NRR here can look misleadingly strong during platform-consolidation cycles and then plateau hard. A multi-year NRR trend matters more than any single quarter.

Developer tools and API-first products (Twilio, Stripe-adjacent, GitLab). Healthy is highly variable — anywhere from 105% to 130%. These are usage-based and tied to the customer's own product growth. When tech customers cut, dev-tools NRR cuts with them, often violently.

Warning Signs and Fantasy Numbers

A few patterns that should trigger scrutiny regardless of category:

NRR rising while gross margin falls. Expansion is being bought via discount-heavy upsell or services-loaded deals.
NRR holding flat while new ARR collapses. The base is masking front-end weakness. Eventually the base ages out.
Cohort-restricted NRR only. If the company stopped disclosing all-customer NRR and only reports the ">$100k" cohort, that's a tell.
Price-driven NRR. Companies that raised list prices meaningfully in recent years got a one-time NRR lift. That tailwind doesn't repeat.
NRR above 140% in a mature category. Possible, but the burden of proof is high. Look for the specific product cycle driving it and ask when it normalizes.

The fantasy number is any NRR figure quoted without the denominator definition. "We had 127% NRR" means nothing until you know whether that's all customers, all paying customers, customers above a threshold, or only customers who renewed.

Putting It Together: A Quick Read Checklist

When a software company reports, run NRR through this filter:

Category benchmark. Is the number in the healthy band for its type?
Trend, not level. Is it stable, decelerating, or re-accelerating? Three-quarter direction matters more than one print.
Cohort definition. Did it change? Did the threshold move?
Decomposition. Has management disclosed seat vs. usage vs. price vs. cross-sell? If not, ask why.
Cross-check against gross retention. Net minus gross gives you the pure expansion contribution. If gross retention is sliding, NRR is doing the work of two metrics.

What to Watch Next

Pull the last 8 quarters of NRR for any software name you own and chart the trend — direction beats level every time.
Check whether the company discloses gross retention separately. If not, you're flying with one instrument.
For usage-based names, cross-reference NRR with customer count growth in the >$1M ARR cohort. Both rising = real. NRR rising, large-customer count flat = optimization pause.
Re-read the most recent 10-K footnote on how NRR is calculated. Definitions drift, and the drift is rarely in the investor's favor.

For more on how to calibrate earnings metrics by category, see Bookings vs. Billings vs. Revenue: A Software Investor's Framework, which breaks down how to parse different demand signals. And for a broader framework on reading the MD&A and 10-K disclosures where NRR definitions live, check Reading the MD&A: What Management Explains vs. Skips.