Customer discovery statistics 2026
Decision support from the customer signals you already have.
UpSight connects signals across conversations, emails, surveys, and research, then brings back the patterns and source evidence you can verify before you decide.
Start with the customer evidence you already have.
Decision support starts with evidence you can verify. These are sourced figures on why startups fail, how to tell you have product-market fit, and how many customer interviews are enough. Each figure links to the study it came from, with the publication date, the sample, and the exact sentence we relied on.
By Rick Moy · Last reviewed 2026-09-30 · Next review by 2026-12-31
Why startups fail, and how to tell you have product-market fit
Two sources: an analysis of startup post-mortems, and a survey benchmark for product-market fit.
Poor product-market fit was cited in 43% of the 385 startup failures CB Insights analyzed, ahead of bad timing at 29% and unsustainable unit economics at 19%. Startups could cite more than one reason, so the shares add up to more than 100%.
“poor product-market fit (43%), bad timing (29%), and unsustainable unit economics (19%)”
“The analysis includes 385 companies for which failure reasons could be identified.”
“Since many startups cited multiple reasons for their failure, the chart below exceeds 100%.”
- Source
- CB Insights, The top reasons startups fail
- Published
- 2026-03-05
- Period measured
- Not stated for the chart. The page describes CB Insights' post-mortem dataset as covering startup failures since 2023.
- Sample
- 385 companies for which failure reasons could be identified
- Read it with care
- Capital running out is the most common final cause; CB Insights treats these as the underlying reasons behind it. The page does not say how the 385 were selected.
Two-thirds of the product-market-fit failures CB Insights analyzed were early-stage companies that never found a market, and 20 Series B and later companies also cited poor product-market fit as a primary cause.
“Two-thirds of product-market fit (PMF) failures were early-stage companies that never found a market.”
“But 20 Series B+ companies also cited poor PMF as a primary cause.”
- Source
- CB Insights, The top reasons startups fail
- Published
- 2026-03-05
- Period measured
- Not stated
- Sample
- The product-market-fit failures within the same analysis of 385 companies
- Read it with care
- The page gives the later-stage count (20) but not the total number of product-market-fit failures.
After benchmarking nearly a hundred startups with his customer development survey, Sean Ellis found that companies that struggled to find growth almost always had less than 40% of users answer “very disappointed” when asked how they would feel if they could no longer use the product, while companies with strong traction almost always exceeded 40%.
“After benchmarking nearly a hundred startups with his customer development survey”
“Ellis found that the magic number was 40%.”
“Companies that struggled to find growth almost always had less than 40% of users respond "very disappointed," whereas companies with strong traction almost always exceeded that threshold.”
- Source
- First Round Review, How Superhuman Built an Engine to Find Product Market Fit
- Published
- 2018-11-13
- Period measured
- Not stated
- Sample
- Nearly a hundred startups, per the article
- Read it with care
- This is Sean Ellis's benchmark as reported in an article by Superhuman's founder. Ellis's underlying dataset is not published there. It is a rule of thumb, not a guarantee.
How many customer interviews are enough
Two peer-reviewed studies that measured when new themes stop appearing as interviews accumulate. Both come from health research, not product discovery.
In a study of sixty in-depth interviews, data saturation, the point at which no new themes appeared, occurred within the first twelve interviews, and basic elements for metathemes were present as early as six interviews.
“Using data from a study involving sixty in-depth interviews with women in two West African countries”
“they found that saturation occurred within the first twelve interviews, although basic elements for metathemes were present as early as six interviews”
- Source
- Field Methods (Guest, Bunce and Johnson), How Many Interviews Are Enough?
- Published
- 2006-02
- Period measured
- Not stated in the abstract
- Sample
- Sixty in-depth interviews with women in two West African countries
- Read it with care
- A health-research study of one population, not product discovery. It shows when themes stopped appearing in that data set, not a rule for every interview program.
In 25 in-depth interviews, code saturation, where the range of thematic issues was identified, was reached at nine interviews, while 16 to 24 interviews were needed to reach meaning saturation, a richly textured understanding of the issues.
“Examining 25 in-depth interviews”
“we found that code saturation was reached at nine interviews, whereby the range of thematic issues was identified”
“16 to 24 interviews were needed to reach meaning saturation where we developed a richly textured understanding of issues”
- Source
- Qualitative Health Research (Hennink, Kaiser and Marconi), Code Saturation Versus Meaning Saturation
- Published
- 2016-09-26
- Period measured
- Not stated in the abstract
- Sample
- 25 in-depth interviews
- Read it with care
- A health-research study. The gap between the two numbers is the useful part: hearing every theme takes fewer interviews than understanding each one.
How this page is compiled
- Primary sources only. A figure goes on this page only if we read it on the publisher’s or journal’s own page. Figures that appear only on aggregator or vendor blogs with no stated method are left out, including ones that are widely repeated.
- The sentence is printed. The quoted sentence sits next to each number, so you can check what the source said rather than what we summarized.
- Unknowns are stated. Where a source does not give the period or the sample, we write “not stated” rather than fill the gap.
- Journal abstracts. Journals block automated readers, so for the two peer-reviewed studies we confirmed the quoted abstract text through the Crossref record and link the DOI.
- Reviewed every quarter. Each review re-reads every quote on its source, replaces outdated figures, and adds a line to the log below.
UpSight’s own research will get its own section, with its method and sample size, once there is enough data to publish it responsibly.
Change log
- 2026-09-30. First publication: five figures from four sources. Every quoted sentence was re-read on the source page that day.
Put the numbers to work
Start with the method in our guides, then run it on conversations you already have.
Start with the evidence behind your next decision.
Upload it. Find the signal your team can act on.