Startup market research examples for business teams

What to take away

These startup market research examples are designs, not measured outcomes. Each one names a real company and pairs a decision with public data, customer evidence, a behavior test, and a stop rule.

Figma's pricing shift started with support tickets

Figma is a browser-based design tool founded in 2012. It runs a free Starter plan and paid plans for teams, priced per seat. Shared team libraries and longer version history sit on the paid tiers.

Figma moved collaboration features into its paid tiers after watching free-team usage patterns and support volume, not after a survey. The research question was narrow: which features do teams actually pay for? Figma's 2019 pricing update made that shift public.

The method transfers. Classify your own support, pause, cancellation, and refund records before you add acquisition spend. Interview people about their last decision to stop, not their general satisfaction. Then test clearer plan terms against matched cohorts and judge on retained contribution, not signups.

Retained contribution is also the deciding number in the community-led acquisition versus paid ads comparison.

Notion read its template community before it sold to it

Notion is a workspace tool that combines notes, documents, and databases. It has a free personal plan, paid per-member plans such as Plus, and a template gallery where users publish and duplicate pages.

Notion's template gallery grew from watching which user-built pages got copied and shared. That is public behavior, not stated preference, and it told the company which use cases to build for. Notion's help center documents the gallery.

Run the same check on your own product. Which user-created artifacts do other users copy? Which get abandoned?

Superhuman tested one narrow segment at a time

Superhuman sells an email app for Gmail and Outlook, launched at about $30 per month, and grew through an invitation-only waitlist. The waitlist doubled as the research tool.

Superhuman ran a waitlist and onboarding calls with a defined segment before opening the product broadly. The research question was who gets value fastest, not how big the market is. Founder Rahul Vohra described onboarding calls with new users during the invite period, plus a product-market-fit survey, in a 2018 First Round Review article.

Define eligible users, the service or product radius, and the staffing the promise requires. Interview recent buyers and recent non-buyers separately. Test a waitlist with real dates. Reject any segment your operations cannot serve.

Zapier researched the integration gap, not the market size

Zapier connects web apps through automated workflows called Zaps. It lists thousands of app integrations, and its free plan carries a monthly task limit before paid plans begin.

Zapier's early growth came from covering integrations its users asked for, one at a time. The team watched which app-pair requests repeated, then built against that list. Zapier's public integration directory lists the result.

The transferable move: measure one dense use case instead of claiming a large total market. For a marketplace, that means enough qualified supply to meet a truthful buyer promise in one geography before a second. The pilot starts only when supply can carry the promise.

Segment's interviews found a different problem

Segment is a customer data platform that began in 2011 as ClassMetric, a classroom analytics tool built by a four-person founding team. Interviews with the instructors who signed up for the beta pointed at a different problem: connecting and managing analytics tools.

The founders rebuilt around that problem instead of pushing the classroom product. Segment's public company history describes that 2011 pivot. Copy the interview set for a B2B purchase. Take a recurring purchase inside a finance team and talk to users, budget owners, procurement staff, and vendors who were rejected. With permission, map the documents and delays.

A manual pilot then tests whether a buyer supplies real records and schedules stakeholders. Praise is not the signal. A scheduled stakeholder meeting is.

Public data and compliance: BLS, Census, NIST, W3C

The Bureau of Labor Statistics' OEWS data overview describes employment and wage estimates by geography and industry. OEWS publishes those estimates for states, metropolitan areas, and occupations.

Those estimates frame a workforce question. They do not prove product demand or an employer's willingness to pay.

The Census Bureau's American Community Survey data profiles summarize social, economic, housing, and demographic measures for one geography. Check the year, the estimate, and the uncertainty before you use one to pick a launch area.

The NIST Privacy Framework starting guide outlines a voluntary process for identifying and managing privacy risk. Use it to assign data and response owners. It is not legal clearance; privacy obligations vary by state, and the FTC and state attorneys general set the rules in their scope.

The W3C information and relationships explanation says visual structure should also be available to assistive technology. Apply that check to startup market research outputs, then test the page with the users and technology in scope.

Quick comparison

Write the stopping rule before the test

Before any test, write what would disconfirm the assumption and who can stop spending. A written list of startup market research questions keeps decision, boundary, and stop rule in one place. For example, a growth lead stops spend on a channel when cost per qualified signup exceeds a set ceiling for two consecutive weeks.

Common questions

Are these real startup case studies?

No, and none of them claim a measured result. Figma, Notion, Superhuman, Zapier, and Segment are real companies, and the moves described here are public. The research plans attached to them are designs you copy.

Can a startup copy an example's sample size?

No. Population, segments, incidence, and risk all differ. Method and product maturity also differ. Copy the structure and set your own numbers.

What should transfer from an example?

Transfer the decision, the boundary, the evidence map, and the participant roles. Transfer the behavior test and the stopping rule.

Where do the public data files come from?

The Census Bureau and the Bureau of Labor Statistics publish them. Check the vintage and the margin of error before you build a decision on one.