How do you know if you have product-market fit?

What is the Sean Ellis 40 percent test?

It is one survey question, how would you feel if you could no longer use this product, scored as the share of users who answer very disappointed. The benchmark is 40 percent or more.

Ellis was the first marketer at Dropbox, LogMeIn, Eventbrite, and Lookout, per Koji's documentation on the test. In Vohra's First Round Review article, the startups Ellis benchmarked that struggled to find growth almost always came in under 40 percent, and the ones with strong traction almost always came in over it.

The words almost always leave room for companies on the wrong side of the line, so we treat 40 percent as a threshold to check.

Vohra also cites Hiten Shah, who put the question to 731 Slack users in a 2015 open research project. Fifty-one percent answered very disappointed.

How did Superhuman use the test to improve its product?

It split the answers by who gave them, then built for the users who would miss the product most and the ones closest to joining them. The score went from 22 percent in the summer of 2017 to 58 percent within three quarters, per Vohra's November 2018 First Round Review article.

The first gain came from counting differently. Superhuman took the personas in its very disappointed group (founders, managers, executives, business development) and scored only users who matched them, which moved the number from 22 percent to 33 percent.

The rest came from the roadmap. Superhuman ignored the not disappointed users and split the somewhat disappointed ones by whether speed was their main benefit. What held the speed group back most was the lack of a mobile app. Half the roadmap went to what very disappointed users loved, and half to what held the others back.

The percentage was tracked weekly, monthly, and quarterly. The product team's OKR had exactly one key result, and it was that number.

Is a survey score enough on its own?

No. It tells you whether your active users would miss the product, and it says nothing about what the next customer costs to reach.

Vohra says results start to be directionally correct at around 40 respondents. At 40, one answer moves the score 2.5 percentage points.

We have no clean answer for a business whose entire customer list is shorter than 40 names.

The score also stops at the product. Before we plan spend, we want the very disappointed percentage and who gave that answer, because that segment is where a channel budget should point.

What happens if you scale go-to-market spend before you clear 40 percent?

You risk what Startup Genome calls premature scaling. In its study of over 3,200 high growth technology startups, 70 percent had scaled prematurely, and 93 percent of those never broke 100,000 dollars in monthly revenue.

Startup Genome measured something other than the Ellis score. It defines premature scaling as a company behaving one stage ahead of where it actually is, which on the customer side means spending too much on customer acquisition before product-market fit. Startups it classed as inconsistent were 2.3 times more likely to spend more than one standard deviation above the average on customer acquisition.

The data is from 2011. It was gathered from February that year, and JF Gauthier's post is dated September 2, 2011.

This is where we draw the budget line. Under 40 percent very disappointed, channel spend stays at test size and the rest goes into the segment and roadmap work Superhuman did. Over it, we plan how to scale the channels that tested well.

Startups that scaled properly grew about 20 times faster than the ones that scaled prematurely. No startup that scaled prematurely passed 100,000 users.