Most benchmarking decks get built the same way. 

Someone pulls a dataset, filters by broad sector, and produces a chart with your company as a dot against a cloud of “comparable” businesses. It looks rigorous. It usually isn’t. 

The issue is what “comparable” actually means in most of these datasets. Sector tags are blunt instruments. “B2B SaaS” covers a company selling £50-a-month subscriptions to small businesses and a company selling seven-figure annual contracts to enterprise procurement teams. Both get the same tag. Neither one’s metrics mean anything next to the other’s, but they end up in the same benchmark cloud because nobody built the dataset with contract structure or customer size in mind. 

The second problem is survivorship. A lot of benchmarking data is built from whichever companies were willing to share numbers, or whichever companies are large enough and public enough to have numbers available to scrape. That’s a specific population (generally healthier, generally further along, generally more visible) and it skews every median upward in ways that make an otherwise perfectly healthy company look like it’s underperforming. 

What you actually want, and what almost nobody builds properly, is a benchmark that segments on the things that structurally drive the metric you’re looking at. For NRR, that’s contract length, customer concentration, and expansion motion. For CAC payback, it’s channel mix and average contract value. For gross margin, it’s whether the business carries meaningful services revenue alongside software. Compare on sector alone and you’re comparing apples to a fruit bowl. 

The practical test before you trust any benchmark: ask what’s actually driving the segmentation. If the answer is “industry” and nothing else, treat the number as a rough orientation, not a target. If the segmentation includes structural variables like contract length, customer size, and revenue mix, you’re looking at something closer to a genuine comparison, and it’s worth building your board narrative around it. 

This is exactly the distinction we built the benchmarking tool around; segment first, then compare, rather than compare first and hope the segment was close enough.