I want to talk about a specific way benchmarking quietly misleads people, because it’s one of those things that looks completely reasonable in the boardroom and falls apart the moment you actually dig into where the comparison data came from. 

Here’s the pattern. Someone pulls a benchmark deck. Your NRR is at 105%, median for your sector is 112%, so the obvious conclusion is you’re underperforming and need a retention initiative. Sounds sensible. Gets nodded through. Retention project gets greenlit. Six months and a fair amount of budget later, NRR hasn’t moved much, and nobody’s quite sure why. 

The problem usually isn’t the retention work. It’s the benchmark itself. 

Median figures in most benchmarking datasets are built from whoever happened to submit data, or whoever the data provider could scrape pricing pages and public filings from. That’s a very different population from “companies structurally similar to yours.” 

A company with a genuinely different customer concentration, contract length, or price point can post a very different NRR number for reasons that have nothing to do with how well they retain customers and everything to do with how their book is built. Comparing yourself to that median isn’t benchmarking. It’s noise dressed up as insight. 

The tell is usually in the confidence of the number rather than the number itself. If someone hands you a single median figure with no distribution, no segment breakdown, and no visibility into company size or contract structure behind it, that number is decoration, not evidence. 

Good benchmarking data comes with a shape, not a point. You want to know what the spread looks like for companies genuinely similar to yours on size, motion, and contract structure, not the entire category lumped together. 

There’s a second version of this trap that’s even more common and slightly more dangerous, because it survives scrutiny better. It’s when the benchmark is genuinely well-constructed, genuinely apples to apples, and the target you’re being measured against is still wrong for you. 

Median is a description of the middle of a distribution. It is not automatically the right ambition for your specific business. A company two years post-carve-out with a legacy pricing structure has no business chasing the same NRR target as a company three years into a land-and-expand motion with modern packaging. Both might be “at benchmark” for their sector and be in completely different positions of health. 

What I’d actually do with a benchmark number before it drives a board decision: ask who’s in the comparison set, ask what the distribution looks like rather than just the median, and ask whether the target implied by that median makes sense given your specific contract structure, pricing model, and customer concentration. 

If the answer to any of those is “we’re not sure,” treat the benchmark as a conversation starter, not a target. 

This is exactly the kind of thing worth stress-testing properly rather than eyeballing off a slide, which is a big part of why we built the benchmarking work the way we did inside this community. 

A number without its context isn’t a benchmark. It’s just a number that sounds like one.