Steve Blank's Customer Development: The 4 Steps, Run in 2027

You keep shipping features nobody asked for. Usage is flat, a promising pilot went quiet, and a mentor or investor told you to read Steve Blank. This guide from the BuildBetter team explains Blank's customer development method in plain language. It then gives you a run-book for 2027 that needs only a spreadsheet, a calendar, and a phone.

What Is Customer Development? (Steve Blank's Four Steps in Plain Language)

Customer development is a method, introduced by Steve Blank in The Four Steps to the Epiphany (2005), that treats a startup's business model as a set of untested hypotheses and tests them with real customers before scaling. It runs alongside product development and does not replace it. Blank's argument was that the standard product process (concept, development, alpha, beta, launch) had no structured way to find customers or markets. Teams built first and hoped buyers would show up.

The customer development model has four steps:

• Customer discovery: confirm the problem exists and matters to a specific customer.
• Customer validation: prove customers will buy or commit, and that the sale can be repeated.
• Customer creation: create demand on purpose once you know what works.
• Company building: move from a learning organization to an execution organization.

Steps 1 and 2 are search. Steps 3 and 4 are execute. Blank and Bob Dorf made this split explicit in The Startup Owner's Manual (2012), the longer and more practical successor to Four Steps. It underpins Blank's definition of a startup: “a temporary organization designed to search for a repeatable and scalable business model.”

The core idea fits in four words: get out of the building. Blank's point is that there are no facts inside your office, only opinions. Facts about the problem, the price, and the channel live with customers. He also adapted von Moltke: “No business plan survives first contact with customers.”

Customer development tests whether you have customers and a repeatable way to reach them before you spend money scaling a product.

Why Customer Development Is Hard: What Actually Breaks

Customer development fails less from bad theory than from bad execution: founders run interviews that cannot produce a “no.” CB Insights' 2021 analysis of 111 startup post-mortems found that 35% cited “no market need” as a reason for failure. Discovery exists to catch that failure early, but only if it is run honestly. The same problems show up again and again:

• Interviewing to confirm, not to disprove. “Would you use a tool that…?” gets polite yeses that predict nothing.
• Hypotheses never written down. If nothing is on paper, nothing can be killed, and every interview “kind of supports” the idea.
• The wrong people. Friends, other founders, and B2B users who are not the economic buyer all give you comfortable answers.
• Interest treated as validation. Rob Fitzpatrick, author of The Mom Test, calls compliments “the fool's gold of customer learning.” Waitlist signups and “send it when it's ready” are not commitment.
• Notes that decay. When nobody records the call, the founder summarizes it from memory, and memory drifts toward what the founder already believed.
• Skipped steps. Teams spend on marketing before validation has shown a repeatable sale. Blank calls premature hiring of sales and marketing the classic way startups burn cash.
• A phase instead of a loop. Discovery stops at launch, and the team loses contact with the problem it was solving.

Each step in the run-book below is designed to block one or more of these.

How to Run Customer Development in 2027: A Step-by-Step Run-Book

You can run all four steps of customer development with a spreadsheet, a calendar, and a phone. No software is required to start. The four-step structure belongs to Steve Blank. The specific numbers below are working rules drawn from practice and interview research, not Blank's own prescriptions. Adjust them to your market.

Step 1: Customer Discovery (Is the Problem Real?)

Customer discovery succeeds when you can name one segment that describes the same painful problem unprompted and has already tried to solve it.

Build a hypothesis sheet before the first call. Use one row per hypothesis and these columns:

• Customer segment
• Problem
• Current workaround
• Buyer vs. user
• Willingness to pay
• Channel
• Kill criteria (what you would need to hear to drop it)

The kill-criteria column matters most. Without it, you will reinterpret every answer as support. Many teams sketch the same assumptions on Alexander Osterwalder's business model canvas, which Blank adopted in his Lean LaunchPad course. The canvas works well for overview, and the sheet works well for testing.

Target about 20 problem interviews per segment. Griffin and Hauser (1993) found that 20–30 one-on-one interviews surfaced roughly 90–95% of customer needs in a category. Guest, Bunce and Johnson (2006) found thematic saturation within about 12 interviews in a fairly uniform group. Stop earlier if answers converge, and keep going if they diverge. For scale, NSF I-Corps teams running Blank's curriculum do about 100 interviews in seven weeks.

Run each interview like this:

• Record it with consent. Write a one-paragraph snapshot within an hour that covers the problem in their words, what they do today, what it costs, and who else is involved.
• Ask about past behavior: “Tell me about the last time…” This matches the story-based interviewing in Teresa Torres's Continuous Discovery Habits.
• Do not pitch or demo. The goal is to learn, not to sell.
• Close by asking for a referral to someone with the same problem. Whether they agree to refer you is itself a signal.

Exit criterion: a clear majority of one segment describes the same problem unprompted and has already built a workaround, such as a spreadsheet, hired help, or a hacked-together script. Blank calls these people earlyvangelists. They have the problem, know it, are looking for a fix, have improvised one, and have budget or can get it.

Step 2: Customer Validation (Will They Commit Before You Scale?)

Customer validation means getting commitment, not opinions. A commitment costs the customer something. Examples include a paid pilot, a signed letter of intent, a deposit, a calendar slot for onboarding, or access to their real data. Fitzpatrick's test applies here: a good meeting ends with a concrete next step, and “keep in touch” is a polite rejection.

• Build the smallest artifact that allows a yes or no. This could be a clickable prototype, a concierge service you deliver by hand, or a small working version.
• Write a sales roadmap. Record who signs, who blocks, which objections recur, and how long the cycle takes from first call to signature.
• Test repeatability. Can someone other than the founder close the next deal with the same story? If only the founder can sell it, you have charisma, not a repeatable sale.
• Follow the pivot rule. If validation fails, return to discovery with an updated hypothesis sheet. Don't keep pushing the same pitch harder.

Step 3: Customer Creation (Generate Demand on Purpose)

Customer creation should start only after validation shows a repeatable sale to a defined segment. Spend before that point scales a funnel that doesn't convert.

Blank argues that market type sets the demand strategy:

• Existing market: compete on performance against known alternatives.
• New market: educate buyers and define the category. This takes longer and costs more.
• Resegmented market: reposition an existing market around low cost or a niche.

The Startup Owner's Manual adds a fourth type, the clone market, which copies a model proven in another country. Run channel experiments with explicit budgets and conversion targets tied to your sales roadmap. Keep at least one customer conversation a week so positioning stays grounded in what buyers say now, not in what they said months ago.

Step 4: Company Building (From Learning to Execution)

Company building turns a founder-led search team into departments with clear missions, without cutting the customer feedback loop.

• Write down what you learned: segment, problem, buyer, and objections. New hires should inherit evidence, not folklore.
• Keep discovery as a standing habit, such as weekly interviews by the product trio (PM, designer, engineer). It should not be a phase you finished during year one.

What AI Changes About Customer Development (and What It Doesn't)

AI makes customer development faster at synthesis and prototyping, but it does not change where the evidence comes from.

What changes:

• Synthesis across all calls. AI can compare 20 transcripts against your hypothesis sheet in minutes and surface contradictions a tired founder would miss.
• Prototypes in hours. Coded prototypes and small working versions are now cheap enough to use in validation instead of slide decks.
• Faster prep and snapshots. AI can draft interview guides and post-call snapshots from recordings.

What doesn't change:

• You still have to get out of the building. AI cannot create primary evidence about your customers. Personas generated from real customer data can help you pressure-test question wording or explore follow-ups, but they are not a substitute for interviews in early discovery.
• Judgment stays with you. Deciding which hypothesis to kill is a founder decision. Check every AI summary against the source quote.

The risk: faster synthesis makes it easier to confirm bias at scale. Prompt the model to argue against your hypothesis. Ask “Which interviews contradict this?” rather than “Summarize support for this.”

Worked Example: A Six-Week Customer Discovery for a B2B Startup

This is an illustrative composite example, not a real company or case study.

Setup: two founders want to automate vendor security questionnaires for mid-market SaaS companies. They write five hypotheses covering segment, problem, buyer, willingness to pay, and channel.

Weeks 1–2: they run 12 problem interviews with security and compliance leads. The hypothesis that security teams own the questionnaire dies. In most calls, sales engineers do the work and security only reviews it.

Weeks 3–4: they run 10 more interviews, now aimed at sales engineers and revenue leaders. The hypothesis that small companies will pay dies, because companies under about 50 employees get too few questionnaires to care. The time-cost hypothesis survives in a different form: the real cost is deals stalled while buyers wait for answers.

Week 5: they show a clickable prototype to 8 interviewees. Five ask about pricing, and three agree to a paid pilot at a set monthly fee.

Week 6: they reach a decision point.

Decision: target sales-led SaaS companies with roughly 100–500 employees, sell to the revenue leader, and position on deal velocity rather than security hours saved. The founders move into validation with three pilots.

Lesson: the surviving idea is a different product for a different buyer than the founders started with. That is the method working as intended.

Customer Development vs Lean Startup vs Ship First: When to Sell First and When to Ship First

Customer development answers “should this exist and who pays?” Shipping a small working answer answers “does this actually do the job?” Most teams need both, in that order.

The Ship First loop works like this: a customer asks for something, you build a small working answer, they try it in their own workflow, and what happens next tells you what to improve. Lean Startup contributed the minimum viable product, validated learning, and the pivot-or-persevere decision.

Sell first when the problem or buyer is unproven, the sale is high-touch B2B, building is expensive, or the product is regulated or integration-heavy and a working version would take months.

Ship first when a known customer is asking for something specific, a small working answer takes days to build, and real usage in their workflow will tell you more than what they say.

Is The Four Steps to the Epiphany Still Relevant?

Yes. The core logic of The Four Steps to the Epiphany still holds: the business model is a set of hypotheses, and evidence comes from customers. Two decades of Lean LaunchPad and I-Corps cohorts have shown that the search-then-execute sequence works across industries.

One thing has aged. The book predates cheap cloud software and AI prototyping, so it assumes building is far more expensive than interviewing. In 2027 a working prototype can cost less than a week of interviews. That shifts some validation work from slides toward small working versions. It does not remove the need for discovery.

A suggested reading order: start with Ship First for a modern working loop, then Customer-Led Development for the fuller method. Finish with The Startup Owner's Manual, the more practical successor to Four Steps.

Common Customer Development Mistakes

Most customer development mistakes come from mistaking agreement for evidence.

• Asking “would you use this?” Ask “tell me about the last time this happened” instead. Hypothetical questions get hypothetical answers.
• Counting interviews instead of killed hypotheses. Twenty interviews that change nothing were twenty confirmations. Blank treats a killed hypothesis as progress.
• Interviewing users when the buyer decides. In B2B, a delighted user without budget authority is not validation.
• Accepting enthusiasm as commitment. If nobody has given money, time, data, or reputation, nothing is validated.
• Not recording calls. Summaries written from memory drift toward the founder's prior beliefs.
• Running discovery once. Customer contact drops after launch, and the product drifts away from the problem.
• Delegating discovery. Blank insists founders do it personally, because founders make the pivot call.
• Outsourcing judgment to AI synthesis. Read the source quotes before you act on a summary.

When You Need Tooling for Customer Discovery (and When You Don't)

Under about 30 interviews and one segment, a spreadsheet hypothesis sheet and a folder of recordings are enough. Tooling starts to pay off when several people run interviews, or when evidence is spread across calls, support tickets, and Slack. It also helps once you hold more than about 10–15 customer conversations a week, or when insights must become PRDs and tickets that other teams act on.

1. BuildBetter. BuildBetter captures calls (with a bot or with no-bot local recording), Slack threads, and support tickets as one evidence base. It then turns that evidence into PRDs and Jira or Linear tickets. It earns its place once discovery becomes a continuous team habit rather than a founder's six-week sprint. One caveat: teams that mainly need a dedicated research repository with deep manual tagging and highlight-reel workflows may find a specialist repository tool more mature for that use.
2. A shared spreadsheet or doc template for the hypothesis sheet and interview snapshots. This remains the core artifact at any scale.
3. A scheduling link and a consent script for recording. Together they remove most of the logistics that make 20 interviews feel hard.

No tool replaces getting out of the building.

FAQ: Steve Blank's Customer Development

What are the four steps of customer development?

The four steps are customer discovery, customer validation, customer creation, and company building. Steve Blank introduced them in The Four Steps to the Epiphany (2005). Discovery tests whether the problem is real. Validation tests whether customers will commit and whether the sale is repeatable. Creation scales demand. Company building turns the startup into an execution organization.

How many customer discovery interviews should I do?

A practical working target is about 20 problem interviews per customer segment. Stop earlier if answers converge, and continue if they keep diverging. Griffin and Hauser (1993) found that 20–30 interviews surface roughly 90–95% of needs. Guest et al. (2006) found that saturation often occurs within about 12 interviews in a homogeneous group.

What is the difference between customer development and Lean Startup?

Customer development (Blank, 2005) is the process for testing a business model with customers. The Lean Startup (Ries, 2011) builds on it by combining it with agile development. It adds the minimum viable product, the build-measure-learn loop, validated learning, and the pivot-or-persevere decision.

What does “get out of the building” mean?

It is Blank's shorthand for gathering evidence directly from customers instead of relying on internal opinions, plans, or forecasts. Interviews, observation, and sales attempts all count. Debate in a conference room does not.

What counts as customer validation?

Customer validation is a commitment that costs the customer something. That can be money (a paid pilot, deposit, or pre-order), a signed letter of intent, time (a scheduled onboarding), reputation (an introduction to their boss), or access to real data. Compliments, waitlist signups, and “send it to me when it's ready” do not count.

Can AI replace customer interviews?

No. AI speeds up synthesis and prototyping, but it cannot produce primary evidence about what your customers do and what they pay for.

Further Reading: Books on Customer Development and Discovery

• Ship First: How to Build Self-Improving Products (Spencer Shulem). Free, 24 chapters in three parts: the case for a working answer, building with better evidence, and making the next release better.
• Customer-Led Development (Spencer Shulem). The fuller method, with Ship First as its short companion. Kindle edition.
• The Four Steps to the Epiphany (Steve Blank, 2005).
• The Startup Owner's Manual (Steve Blank and Bob Dorf, 2012).
• The Lean Startup (Eric Ries, 2011).
• Continuous Discovery Habits (Teresa Torres, 2021).
• Inspired (Marty Cagan, 2008; 2nd edition 2017), on the four discovery risks: value, usability, feasibility, and business viability.
• Shape Up (Ryan Singer, Basecamp, 2019), free online, for turning validated problems into six-week bets.

Sources and Further Reading

• Wikipedia: Customer development
• Wikipedia: Lean startup
• Steve Blank's blog
• Shape Up by Ryan Singer
• CB Insights, “The Top 12 Reasons Startups Fail” (2021 update)
• Griffin, A. & Hauser, J. R., “The Voice of the Customer,” Marketing Science (1993)
• Guest, G., Bunce, A. & Johnson, L., “How Many Interviews Are Enough?”, Field Methods (2006)
• U.S. National Science Foundation, I-Corps program materials

Get Ship First Free

Ship First by Spencer Shulem, cofounder and CEO of BuildBetter, shows how to run the loop that follows validation. A customer asks, you build a small working answer, they try it in their own workflow, and what happens next tells you what to improve. It is a free PDF and EPUB with 24 chapters and 6 illustrations. Get Ship First.

Make Churn Optional

Once discovery becomes a weekly team habit, the evidence piles up across calls, tickets, and Slack. BuildBetter keeps it in one place and turns it into PRDs, tickets, and customer follow-ups your team will act on.

Make churn optional. Book a demo.