The Five Most Misused Benchmarks in Founder Decks
1. DAU/MAU ratios from broad “mobile app” reports. Gaming, social, and consumer utility apps systematically inflate this metric relative to B2B tools. The correct comparator population is B2B workflow tools with mandatory daily usage patterns.
2. Churn rates from “SaaS” benchmarks. In the decks we review, SaaS churn commonly varies by as much as 15× between SMB and enterprise segments. A benchmark that does not segment by ACV is not a useful comparator for either segment.
3. CAC from “startup” surveys. CAC is stage-dependent, channel-dependent, and geography-dependent. A pooled startup CAC average from a survey that includes both Stripe and a 3-person B2B vertical SaaS is meaningless for your model.
4. NPS scores from consumer benchmarks. Consumer and B2B NPS are not on the same scale. A 40 NPS is exceptional in B2B and mediocre in consumer. Citing the wrong baseline turns a strength into a liability.
5. TAM figures from analyst reports that aggregate adjacent markets. Market sizing reports frequently combine markets that are not actually accessible to your product. The definition of “Total Addressable Market” in an analyst report is often total global spend across a broad sector, not the serviceable market for your specific product at your price point.
Original claim: "Our 42% DAU/MAU beats the 35% industry average for mobile apps."
What the methodology section actually says: The "35% industry average" pools gaming, social, and consumer utility apps — categories with structurally higher forced-open rates than a B2B scheduling tool.
Corrected claim: "Our 42% DAU/MAU compares to a 28% median among B2B workflow tools with daily-use cases (SaaS Benchmark Report, category-matched sample, n=340), placing us in the top quartile for our actual category — not just ahead of an unrelated one."
Why it matters: The corrected version is a smaller edge on paper (42% vs. 28%, not 42% vs. 35%) but survives the methodology-section question an analyst will ask — the uncorrected version doesn't.
Figures above are illustrative, built to show the correction pattern — not a real founder's deck or a real published benchmark report.
How to Find the Right Benchmark
The correct approach starts with the methodology section, not the headline figure. Before citing any benchmark, answer four questions:
Who was surveyed? What is the sample size, and how were participants recruited? A survey of 40 founders is not a benchmark. A panel study of 3,000 products over 12 months with defined selection criteria is.
What product category is included? Does the report define its product population in a way that includes your exact type of product? If it covers “B2B SaaS” without segmenting by vertical, ACV, or deployment model, it may not be representative of your position.
What year is the data from? The publication date of the report and the data collection date are different. Check when the underlying data was collected, not when the PDF was published.
What are the confidence intervals? Headline figures in benchmark reports often obscure wide distributions. A reported “median of 35%” DAU/MAU might have a 10th–90th percentile range of 12%–68%. Knowing where you sit in the distribution matters more than the median.
Get the Friday Notes dispatch
Intelligence on pitch deck verification, AI governance, and what we’ve shipped. Every Friday at 09:00 GMT.
Recency: The Other Problem Nobody Talks About
Category mismatch gets most of the attention. Recency misalignment is just as dangerous and less discussed.
Markets move. A SaaS churn benchmark from 2022 was published into a market with different competitive dynamics, different interest rate expectations, and different buyer behaviour than the market in 2026. In AI infrastructure, the entire landscape changed between Q1 2023 and Q3 2024. Any benchmark from before that period is measuring a different market.
18 months is the maximum defensible recency window for fast-moving markets. For AI, developer tooling, and fintech compliance, the window may be shorter. If a benchmark is older than this, either replace it with a more recent source or acknowledge the gap explicitly and explain why the older data is still directionally relevant.
A Framework for Defensible Benchmarks
A defensible benchmark satisfies five criteria: (1) published by a research firm or institutional source with a defined methodology; (2) based on a sample that includes your exact product sub-type; (3) collected within 18 months of your raise date; (4) cited from the primary document, not a secondary summary; and (5) used to represent a metric your product actually measures in the same way the research defines it.
- Always read the methodology section before citing a benchmark’s headline figure
- Confirm the sample population matches your exact product sub-type, not just the parent category
- 18 months is the maximum defensible recency window for fast-moving markets
- Cite primary sources; secondary summaries often misquote or recontextualise the original data
- Know where you sit in the distribution — the median is less useful than your percentile
In 64% of decks reviewed in Q1 2026, the benchmark used for DAU/MAU or retention was sourced from a product category that did not match the founder’s product. In most cases, the mismatched benchmark made the metric look better than the correct comparator would have. This gap is what investors notice when they check.
❓ Common Questions
What is a benchmark category mismatch?
How do I find the right benchmark for my startup?
How old is too old for a benchmark data source?
Put this into practice: Get your benchmarks verified before an investor checks them →
