Market demand is not a single number or enthusiastic reaction. It is a set of dated observations about a defined customer, an important job, current alternatives, willingness and ability to change, and a credible route to purchase or adoption. Start with secondary data to bound the question, then gather direct evidence from the relevant people. Record what they do and spend as well as what they say. Design a test with a decision rule, and revise the idea when evidence points elsewhere.
Separate signals from assumptions
A broad report can show that a category exists; it cannot by itself establish that a particular customer needs a particular product. A large addressable-market estimate does not tell a founder whether a first segment is reachable. An interview statement such as “I would use that” is useful context, but it is weaker than an observed workaround, repeated task, budget allocation, change process or agreed next step. A signed expression of interest can still leave price, procurement, use, renewal and implementation unresolved.
The U.S. Small Business Administration recommends examining demand, market size, economic indicators, location, saturation and pricing, while distinguishing existing sources from direct research. NSF I-Corps presents customer discovery as experiential learning to assess technology market potential. Neither source says one data point or interview proves demand.
The signal ladder
Use the ladder to plan learning, not to assign universal scores:
- Context signal: a credible source shows that the industry, user group or workflow exists. Use it to define the research boundary.
- Problem signal: a specific person describes a recent event, frequency, consequence and current workaround.
- Behavior signal: the person spends time, money or political capital on a solution, or has attempted a change.
- Buyer signal: the budget owner, buying process, decision criteria and blockers are identified.
- Commitment signal: an authorized buyer agrees to a concrete next action, such as a scoped evaluation or procurement discussion. Do not call interest a purchase.
- Repeatability signal: comparable buyers show a similar need and path. A single account does not establish a repeatable segment.
- Retention/value signal: continued use or renewal supports recurring value. This requires time and observed behavior.
A practical research sequence
1. Narrow the market. Specify geography, organization type, user, buyer, workflow and time period. State who is out of scope. “Healthcare” is a sector; “outpatient clinic operations teams handling a defined recurring handoff” is a research boundary.
2. Read existing evidence. Look for official statistics, procurement notices, regulator requirements, public company filings, standards, academic literature and relevant industry publications. Note publisher, population, date, geography, method and limitations. Do not combine data sets that count different things without explaining the difference.
3. Interview for recent behavior. Ask the person to walk through the last instance: “What happened next?” “Who else was involved?” “How did you handle it?” “What did it cost in time, money or delay?” “What have you already tried?” “Who approves a change?” Avoid leading with a pitch or asking respondents to forecast hypothetical use.
4. Observe the job if permitted. Document steps, handoffs, exceptions, tools and rework. Do not collect confidential documents or sensitive data without verified permission and controls.
5. Test the riskiest assumption. A prototype is useful when it answers a defined question. Other tests include a workflow walkthrough, a manual service experiment, a price discussion tied to a real scope or a procurement review. Select a test that a plausible result could disconfirm.
6. Update the record. Keep raw evidence separate from interpretation. Write what changed, what remained uncertain and what decision follows. Record a stop decision as a useful result, not a failure to publish.
Interview prompts that uncover action
- “Tell me about the most recent time this happened.”
- “What did you use, and what happened when it did not work?”
- “How often does this occur, and what changes when volume increases?”
- “Who is accountable for the outcome?”
- “What have you tried or purchased already?”
- “What must be true for you to change the current process?”
- “Who else needs to review a pilot or purchase?”
- “May I follow up with a concise summary for correction?”
Use a consistent interview guide without forcing every respondent into the same assumed problem. If the person does not encounter the job, record that and move on.
Evidence ledger and decision rule
For each claim, record source/date, customer segment, observation, alternative explanation, confidence, missing evidence, next test and owner. Set a decision rule before the next test. For example: proceed to a bounded experiment only if multiple relevant stakeholders independently describe the same recent workflow, a budget owner can explain the change route, and an allowed test can measure a customer-valued outcome. The exact threshold belongs to the venture and its risk; it is not an industry benchmark.
Worked hypothetical
A fictional founder believes small manufacturers need automated order-exception handling. Secondary data establishes a preliminary segment, but not demand. Interviews with operations staff reveal that exceptions occur for several different reasons. Process observation shows the costly work is reconciling supplier data rather than writing a customer response. The founder revises the hypothesis, checks which buyer owns reconciliation and compares a spreadsheet improvement, a rules-based workflow and an AI-assisted option. No market outcome is claimed; the example shows how evidence can change the problem definition.
Common errors
- Treating a total-market slide as proof of reachable buyers.
- Pitching the proposed product before learning the current process.
- Interviewing only friendly peers, internal staff or people without buying influence.
- Counting free pilots as paid demand without tracking sponsor, scope and next decision.
- Reporting percentages from a small, selected sample as if representative.
- Ignoring current alternatives, procurement, integration, switching cost and change ownership.
- Continuing after evidence contradicts the original hypothesis because the idea has already been built.
FAQ
What is the strongest early demand signal? There is no universal strongest signal. A repeated, costly problem, verified with the relevant user and buyer and connected to a concrete next step is more informative than a general compliment. The appropriate evidence depends on the buying context.
Do I need a survey? Not always. Surveys can help quantify a clearly defined question, but they may not reveal workflow context. Interviews, observation and existing records can be more useful for discovering what to measure.
Does a letter of intent prove demand? It documents stated intent under its terms. It does not necessarily show budget approval, procurement completion, product use, payment or renewal.
When should a founder stop? Stop or change direction when the defined user does not experience the problem, a viable alternative already solves it, no buyer owns a change, or the remaining risks cannot be tested responsibly.
Source notes — reviewed 2026-10-31
- U.S. SBA, Plan Your Business: Market Research and Competitive Analysis — the questions and distinction between secondary and direct research.
- NSF, About I-Corps — customer discovery as a learning process for assessing market potential.
This guide is Silicon Founders editorial material for general educational purposes. Examples are hypothetical. It is not investment, legal, accounting, technical assurance or regulatory advice, and it does not describe an active investment, partnership, client mandate or company result. Source links are provided for context; the editorial framework is Silicon Founders’ synthesis.