# Product-Market Fit: Complete Guide 2026 | Enrich Labs > Measure product-market fit in 2026: Andreessen's definition, Ellis 40% test, Superhuman 22% to 58%, CB Insights 43% PMF failures, and a 90-day SaaS plan. _Source: https://www.enrichlabs.ai/blog/product-market-fit-complete-guide-2026_ --- #TLDR **43% of identified VC-backed shutdowns since 2023 cite poor product-market fit as a root cause**, even though "ran out of capital" is how 70% of those stories end ([CB Insights](https://www.cbinsights.com/research/report/startup-failure-reasons-top/)). Fit is a market that pulls plus a product that satisfies it. Marc Andreessen called it the only thing that matters before scale ([pmarchive](https://pmarchive.com/guide_to_startups_part4.html)). Sean Ellis's leading indicator is one question: how would you feel if you could no longer use this product? After nearly 100 startups, companies that grew almost always cleared **40% "very disappointed"** ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). Superhuman started at 22%, ran an engine around high-expectation customers, and hit **58% in three quarters** ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). Do not confuse this with a [go-to-market strategy](https://www.enrichlabs.ai/blog/go-to-market-strategy-complete-guide-2026), [product-led growth](https://www.enrichlabs.ai/blog/product-led-growth-complete-guide-2026), or [demand generation](https://www.enrichlabs.ai/blog/demand-generation-complete-guide-2026). Those spend money after users already pull. * * * ## What is product-market fit? Product/market fit, in Andreessen's June 2007 post, means being in a good market with a product that can satisfy that market. Market size is the number and growth rate of customers for that product. Product quality is how impressive the product is to one user who actually uses it. Those are different variables. A masterpiece for BeOS, Amiga, OS/2, or NeXT still dies if the market is empty ([pmarchive](https://pmarchive.com/guide_to_startups_part4.html)). He credits Andy Rachleff with the law: - **Great team, lousy market:** market wins. - **Lousy team, great market:** market wins. - **Great team, great market:** something special happens. The corollary: the only thing that matters is getting to product/market fit ([pmarchive](https://pmarchive.com/guide_to_startups_part4.html)). Andreessen Horowitz later restated the same line as the early-stage north star ([a16z](https://a16z.com/product-user-fit-comes-before-product-market-fit/)). You feel the absence. Customers do not quite get value. Word of mouth does not spread. Usage grows slowly. Press is blah. Sales cycles drag and deals die. You feel the presence when customers buy as fast as you can ship, usage grows as fast as you add servers, cash piles up, you hire support as fast as you can, and reporters call because they already heard ([pmarchive](https://pmarchive.com/guide_to_startups_part4.html)). Those signals lag. By the time investment bankers stake out your house, you already have fit. Pre-launch teams need a number they can move, which is why Ellis's survey exists. Helena, Enrich Labs' AI marketing agent, does not invent fit. It keeps research, messaging tests, and channel experiments running while founders sit with users. Use it after you have a hypothesis, not instead of one. Compare execution tools later, including [Helena vs. Copy.ai](https://www.enrichlabs.ai/blog/helena-vs-copy-ai). * * * ## Why startups still die without it CB Insights' March 2026 report covers **431 VC-backed companies** that publicly shut down since 2023, excluding prior exits. Combined equity before death: **$17.5B**. Median raise: **$11M**. Average: **$48M**, pulled up by a long tail. Median time from last fundraise to death: **22 months** ([CB Insights](https://www.cbinsights.com/research/report/startup-failure-reasons-top/)). Capital running out tops the list at **70%**, but it is almost always the final cause of death. Among **385 companies** with identifiable reasons: - **Poor product-market fit: 43%** - **Bad timing: 29%** - **Unsustainable unit economics: 19%** Percentages exceed 100% because post-mortems cite more than one cause ([CB Insights](https://www.cbinsights.com/research/report/startup-failure-reasons-top/)). An earlier CB Insights post-mortem set of 101 startups is still widely cited at **42% "no market need"** ([DigitalOcean summary of that report](https://www.digitalocean.com/resources/articles/top-reasons-startups-fail-and-how-to-avoid-them); [original PDF](https://s3-us-west-2.amazonaws.com/cbi-content/research-reports/The-20-Reasons-Startups-Fail.pdf)). Two-thirds of PMF failures in the 2026 set were early-stage companies that never found a market. **Twenty Series B+ companies** also named poor PMF as a primary cause. They raised on traction that never widened. Zume raised **$446M** through Series C, pivoted from robot-made pizza to sustainable packaging, and still failed to find a viable market ([CB Insights](https://www.cbinsights.com/research/report/startup-failure-reasons-top/)). Sector mix in the same file: - **Healthcare and biotech:** 62 companies (14%), **$5.1B** destroyed - **Fintech:** 57 companies, **$4M** median raise - **Food and agriculture:** 54 companies, roughly a third in alt-protein or cultivated meat Warning signs: among companies with 12 months of Mosaic scores, **72% declined** in the year before death, dropping **15%** on average. Partnership activity among 206 companies with tracked relationships dropped **44%** from 12-24 months before death to the final 12 months. Two-thirds of companies with headcount data were shrinking in the last six months ([CB Insights](https://www.cbinsights.com/research/report/startup-failure-reasons-top/)). Fit is not a Series A checkbox. It can fail after you already book [annual recurring revenue](https://www.enrichlabs.ai/blog/annual-recurring-revenue-complete-guide-2026) if that revenue sits in a pocket that will not expand. Watch [churn rate](https://www.enrichlabs.ai/blog/churn-rate-complete-guide-2026) and [net revenue retention](https://www.enrichlabs.ai/blog/net-revenue-retention-complete-guide-2026) as lagging cousins, not substitutes. * * * ## Product-user fit comes first Peter Lauten and David Ulevitch at a16z draw a hard line. If 200 people love the product and you still cannot size the rest of the market, you have **product-user fit**, not product-market fit. Product-user fit is the extent to which you built the right product for the right user. Self-declared PMF in hiring decks and fundraises is often this earlier state ([a16z](https://a16z.com/product-user-fit-comes-before-product-market-fit/)). They watch three things that do not require scale: - **Durable retention.** Early adopters experiment. Do they stay every day? For churn, do you know why? - **Engagement depth that matches the job.** An hour a week can be perfect for a 1:1 tool and terrible for a daily workflow. Look at DAU/MAU, session length, and time per day. - **Testimony.** Would they be very disappointed without it? Superhuman's score is the qualitative version of this ([a16z](https://a16z.com/product-user-fit-comes-before-product-market-fit/); [First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). Enterprise path they describe: build something people want, then ask how many other people want it, then get the product in their hands fast. Bottom-up SaaS can jump from user fit to market fit when end users vote with their feet instead of IT procurement ([a16z](https://a16z.com/product-user-fit-comes-before-product-market-fit/)). That jump is the same discipline as a tight [ideal customer profile](https://www.enrichlabs.ai/blog/ideal-customer-profile-complete-guide-2026) plus [brand strategy](https://www.enrichlabs.ai/blog/brand-strategy-complete-guide-2026) that speaks in the fans' words. * * * ## The Sean Ellis 40% test Ask users who recently experienced the core product: **How would you feel if you could no longer use \[product\]?** Options: very disappointed, somewhat disappointed, not disappointed. Ellis ran early growth at Dropbox, LogMeIn, and Eventbrite. After benchmarking nearly 100 startups, companies that struggled to grow almost always sat under 40% very disappointed. Companies with strong traction almost always exceeded it ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/); [Ellis on Medium](https://medium.com/growthhackers/using-product-market-fit-to-drive-sustainable-growth-58e9124ee8db)). Hiten Shah posed the question to **731 Slack users** in a 2015 open research project. **51%** said they would be very disappointed without Slack, at roughly half a million paying users. First Round cites that as proof that 40% is a high bar ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). Who to poll: users who used the product **at least twice in the last two weeks**. Superhuman had 100-200 users. Directionally useful results start around **40 respondents** ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). Superhuman's four questions: 1. How would you feel if you could no longer use Superhuman? Very / somewhat / not disappointed. 2. What type of people would most benefit from Superhuman? 3. What is the main benefit you receive from Superhuman? 4. How can we improve Superhuman for you? Do not survey the same person twice if you want the 40% line comparable over time ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). A free runner lives at [pmfsurvey.com](https://pmfsurvey.com/) (Sean Ellis and GoPractice). Playbooks that walk the same template include [LearningLoop](https://learningloop.io/plays/product-market-fit-survey) and Vohra's talk notes at [Business of Software](https://businessofsoftware.org/talks/product-market-fit-engine/). * * * ## Superhuman's product-market fit engine By summer 2017 Superhuman had coded for two years, grown to 14 people, and still had not launched. Vohra did not want to throw the product out and see what stuck after that sunk cost. Classic PMF essays described post-launch lagging indicators. Superhuman was pre-launch, avoiding press, and not onboarding more users on purpose ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). ### Segment to high-expectation customers First survey: **22% very disappointed**. They grouped answers, assigned personas, and kept founders, managers, executives, and business development. That jump alone moved the score by **10 points** toward **33%** in the narrowed set ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). Question 2 from very disappointed users paints the high-expectation customer (HXC), using Julie Supan's idea: the most discerning person in the demographic, who enjoys the product for its greatest benefit and spreads it. Superhuman's HXC, "Nicole," reads 100-200 emails a day, sends 15-40, prides herself on being responsive, and hits Inbox Zero a few times a week ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). Paul Graham, quoted in the same piece: launch with users who urgently need what you make. Serve a small number of people who want it a large amount, not a large number who want it a little. Maxima in startup ideas are adjacent, not isolated spikes ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). Graham's broader rule: a startup is a company designed to grow fast, which only works after you made something people want ([Startup = Growth](https://www.paulgraham.com/growth.html)). ### Convert fence-sitters who already feel the main benefit Very disappointed users named speed, focus, and keyboard shortcuts. Vohra ignored people who would not be disappointed at all. They request features for use cases you will never win ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). Somewhat disappointed users split again. If speed was not their main benefit, ignore them. If speed was the main benefit, read question 4. First blocker: no mobile app. Then integrations, attachments, calendaring, unified inbox, search, read receipts ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). ### Split the roadmap 50/50 Half the work doubled down on love: sub-50 ms UI, more shortcuts, Snippets automation, design flourishes. Half the work removed blockers: mobile, integrations, attachments, calendar, unified inbox, search, receipts. Rank low/medium/high cost vs impact. Ship low-cost, high-impact first ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). If you only polish what fans love, the score does not rise. If you only chase missing features, competitors copy the benefit that made fans fanatic. ### Make the score the OKR They surveyed new users on a cadence, never twice, and made "very disappointed %" the product team's only key result. From 22% overall, to 33% after segmenting, to **58% within three quarters**. NPS rose with it. The team grew to 22. Investors asked in ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). The score can drop as you leave early adopters. SaaS companies have to keep improving the product as the pool expands. Superhuman rebuilds the roadmap every quarter with the same engine ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). Call it [growth hacking](https://www.enrichlabs.ai/blog/growth-hacking-complete-guide-2026) in Ellis's original sense: measure a leading indicator, then scale. It is not a [social media marketing strategy](https://www.enrichlabs.ai/blog/social-media-marketing-strategy-complete-guide) bolted on before anyone cares. * * * ## Product-market fit vs GTM, PLG, and demand gen Concept Question it answers When it matters Product-market fit Do a defined set of users need this enough to be very disappointed without it? Before you scale spend [Ideal customer profile](https://www.enrichlabs.ai/blog/ideal-customer-profile-complete-guide-2026) Which firms look like the ones who already love it? While you segment survey fans [Go-to-market strategy](https://www.enrichlabs.ai/blog/go-to-market-strategy-complete-guide-2026) How do we reach and win that market? After a repeatable yes [Product-led growth](https://www.enrichlabs.ai/blog/product-led-growth-complete-guide-2026) Can the product itself acquire and expand users? After core value is obvious in-product [Demand generation](https://www.enrichlabs.ai/blog/demand-generation-complete-guide-2026) How do we create pipeline at volume? After conversion and retention hold [B2B lead generation](https://www.enrichlabs.ai/blog/b2b-lead-generation-complete-guide-2026) How do we capture that pipeline as contacts? After the offer converts [Sales enablement](https://www.enrichlabs.ai/blog/sales-enablement-complete-guide-2026) How do we help reps tell the fan story? After the HXC language is real Andreessen's BPMF vs APMF split is the operating rule. Before fit, ignore almost everything else, including polished HR and a thorough marketing plan that still heads off a cliff ([pmarchive](https://pmarchive.com/guide_to_startups_part4.html)). After fit, push growth. Vohra's close: when you hit the score, grow as fast as you can. It will feel uncomfortable. You now have evidence ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). Premature paid [LinkedIn advertising](https://www.enrichlabs.ai/blog/linkedin-advertising-campaign-complete-guide-2026) or [B2B marketing automation](https://www.enrichlabs.ai/blog/b2b-marketing-automation-2026-guide) into a 22% score burns cash into the 43% PMF-failure bucket. Pair Ellis with [marketing attribution](https://www.enrichlabs.ai/blog/marketing-attribution-complete-guide-2026) only after conversion holds. [Content marketing for SaaS](https://www.enrichlabs.ai/blog/content-marketing-for-saas-complete-guide-2026) should repeat fan language from question 3, not a generic category pitch. * * * ## How to run the survey in B2B SaaS 1. Define "experienced the core." Two meaningful sessions in 14 days, or one completed job-to-be-done (first campaign live, first dashboard shared, first invoice sent). 2. Cap the sample at people who match a draft ICP. Random signups inflate "not disappointed." 3. Send the four Superhuman questions. Keep the first question identical so 40% stays comparable. 4. Stop at ~40 responses for a directional read; keep collecting as you add cohorts. 5. Tag each respondent with persona, company size, and whether they hit the activation event. 6. Build the HXC from question 2 language, not from a deck you wrote last year. 7. Word-cloud question 3 for very disappointed users. That phrase becomes homepage copy, [sales enablement](https://www.enrichlabs.ai/blog/sales-enablement-complete-guide-2026) talk tracks, and [social media brand voice](https://www.enrichlabs.ai/blog/social-media-brand-voice). 8. Act on question 4 only for somewhat disappointed users who already name the same main benefit. 9. Put the percentage on a weekly dashboard next to [ARR](https://www.enrichlabs.ai/blog/annual-recurring-revenue-complete-guide-2026) and [social media KPIs](https://www.enrichlabs.ai/blog/social-media-kpis-complete-guide), not buried in a research folder. 10. Repeat on new users each quarter. Early adopters forgive gaps. The next 100 will not. Helena can draft the survey, cluster open text, and ship messaging tests on the phrases fans actually use. It still cannot sit in the customer call. Founders do that. If you later automate outbound, read [sales automation AI](https://www.enrichlabs.ai/blog/sales-automation-ai-complete-guide-2026) after the score clears 40%, not before. * * * ## Signs you do not have it yet Use Andreessen's feel test as a weekly audit ([pmarchive](https://pmarchive.com/guide_to_startups_part4.html)): - **Sales cycles stretch** and a large share of deals never close. - **Usage is flat** even when you add seats or ads. - **Word of mouth is a hope**, not a channel. - **Press, if you get it, is polite** and forgettable. - **Support hears** "nice, but we still use the old tool." - **Ellis score under 40%** among activated users. - **You cannot name an HXC** in one paragraph. - **[Churn](https://www.enrichlabs.ai/blog/churn-rate-complete-guide-2026) in month two** is a feature, not a bug. - **[NRR](https://www.enrichlabs.ai/blog/net-revenue-retention-complete-guide-2026) is stuck** because expansion has nowhere to go. - **Brand claims outrun product.** [Brand protection](https://www.enrichlabs.ai/blog/brand-protection-complete-guide-2025) and [DTC branding](https://www.enrichlabs.ai/blog/dtc-branding-complete-guide-2026) cannot paper over a missing job-to-be-done. Do not treat a successful fundraise as fit. Superhuman could raise and still sit at 22%. CB Insights' later-stage PMF failures raised on a wedge that never became a market ([CB Insights](https://www.cbinsights.com/research/report/startup-failure-reasons-top/)). Andy Rachleff's framing, restated in a16z's "12 Things" essay: start with who is desperate for what you will build, then the business model that delivers it ([a16z](https://a16z.com/12-things-about-product-market-fit/)). * * * ## Tools that help you measure (not fake) fit Use research and survey tools. Do not buy a growth stack to hide a 22% score. 1. **Helena (Enrich Labs).** Drafts the four-question survey, clusters open text from fans vs fence-sitters, and ships landing and [SEO / GEO](https://www.enrichlabs.ai/blog/best-ai-for-seo-geo-2026) tests in the language of question 3. First choice for teams that already talk to users and need execution without extra headcount. 2. **[pmfsurvey.com](https://pmfsurvey.com/).** Free Sean Ellis / GoPractice runner for the core question. 3. **Typeform or equivalent.** Superhuman emailed a Typeform with the four questions ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). 4. **Your product analytics.** DAU/MAU and session depth for product-user fit ([a16z](https://a16z.com/product-user-fit-comes-before-product-market-fit/)). 5. **A human interview cadence.** The engine fails if nobody reads question 4. Helena stays #1 because it ties measurement to shipping, which is the Superhuman loop. Copy tools without a user sample invent slogans. * * * ## A 90-day plan to raise the score ### Days 1-30: Measure Export users who hit activation twice in 14 days. Send the four-question survey. Aim for 40+ responses. Write the HXC from question 2. Publish the score to the whole company. Pause net-new paid [demand generation](https://www.enrichlabs.ai/blog/demand-generation-complete-guide-2026) experiments that assume conversion will magically improve. ### Days 31-60: Segment and roadmap Split very / somewhat / not. Ignore not-disappointed feature requests. Split somewhat by whether they name the same main benefit. Cost-impact rank a 50/50 roadmap. Kill one vanity project that serves a persona absent from the fan set. Tighten the [ICP](https://www.enrichlabs.ai/blog/ideal-customer-profile-complete-guide-2026) to match fans' firms. ### Days 61-90: Ship and re-survey Ship the highest-impact blocker and one doubling-down improvement. Re-survey only new activated users. Compare to the 40% line and to your baseline. If you crossed 40% in the HXC segment, write the [GTM](https://www.enrichlabs.ai/blog/go-to-market-strategy-complete-guide-2026) that sells that segment first, then [PLG](https://www.enrichlabs.ai/blog/product-led-growth-complete-guide-2026) if the product can acquire. If you did not, do not scale ads. Change product or market. Helena can run the messaging and landing tests once the HXC language is real. Pair that with a human on every call until the score moves. * * * ## FAQ ### What is a good product-market fit score? Ellis's benchmark is **40% very disappointed** among users who actually used the product. Superhuman treated it as the product OKR and later posted **58%** ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). ### Can you have PMF in one segment and not another? Yes. Superhuman's overall 22% hid a stronger pocket once they kept founders, managers, executives, and BD. Narrow first. Adjacent maxima come later ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). ### Is retention the same as product-market fit? Retention is a lagging cousin. People who would be very disappointed tend to stay. People who stay out of contract inertia can still score "somewhat." Use both. Pair Ellis with [churn](https://www.enrichlabs.ai/blog/churn-rate-complete-guide-2026) and expansion inside [ARR](https://www.enrichlabs.ai/blog/annual-recurring-revenue-complete-guide-2026). ### Should we scale marketing before 40%? Andreessen says obsess on fit before scale. Vohra tells investors not to push growth ahead of fit. CB Insights shows later-stage companies still die from PMF that never widened ([pmarchive](https://pmarchive.com/guide_to_startups_part4.html); [First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/); [CB Insights](https://www.cbinsights.com/research/report/startup-failure-reasons-top/)). ### How is this different from founder intuition? Intuition picks the first bet. The engine tells you whether fans, fence-sitters, and lost causes agree. Superhuman's calendar work moved up because users asked, not because the team lived in calendar ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). ### What if only 40 people answer? Vohra says you start to get directionally correct results around 40 respondents ([First Round Review](https://review.firstround.com/how-superhuman-built-an-engine-to-find-product-market-fit/)). Treat it as a compass, then keep collecting. * * * ## Conclusion Product-market fit is a market that pulls and a product that satisfies it. Measure it with Ellis's 40% question. Improve it the Superhuman way: serve the high-expectation customer, ignore lost causes, split the roadmap, and put the score on the wall. CB Insights' **43%** PMF share of failures is the cost of skipping that work. When the number clears 40% in a real segment, then write GTM, PLG, and demand gen. Until then, talk to users, ship the blocker, and keep Helena on research and tests so the team stays on the product.