Choose the First Customer and Use Case

Task

Choose the narrow initial ICP and use case for the transition.

Summary

Select a narrow initial customer profile and repeated use case rather than trying to productize everything.

Choose the First Customer and Use Case You Can Prove

Task ID: S1-08

Choose the initial ideal customer profile from customers, problems, results, and economics the company can already support with evidence. The output is not a description of everyone who might buy. It is a working decision rule that tells sales whom to pursue, delivery what to repeat, product what to build, and leadership which assumptions still require testing.

The decision behind the ICP

A consulting-led company often reaches a point where several customers appear promising, several problems could become an offer, and almost every completed project contains something that could become software. The founder can usually sell all of them through reputation, referrals, and judgment. That flexibility creates revenue, but it also hides the decision the company must make before it can become more repeatable.

The decision is not simply, “What market are we in?” It is:

Which type of customer, facing which specific situation, should the company organize its next offer and product around?

For this task, the required result is a narrow initial ideal customer profile, or ICP, tied to one use case and expressed as a one-page document with inclusion and exclusion criteria. The identified dependencies should supply the underlying sales, customer, delivery, and financial evidence; leadership approval is the governance target, not an external industry benchmark.

An ICP is often reduced to firmographic labels such as industry, employee count, revenue, or geography. Those attributes can help a salesperson find accounts, but they rarely explain why an organization buys, whether it will obtain the intended result, or whether the supplier can serve it profitably. Research on business-to-business segmentation has repeatedly found that observable company characteristics alone are insufficient. Buying behaviour, organizational strategy, use circumstances, service expectations, and decision processes can be more informative than size or sector by themselves.

A useful working definition is therefore:

The initial ICP is the narrowest observable group of organizations for which the company has credible evidence of a recurring problem, a similar buying situation, a repeatable result, workable delivery, acceptable economics, and practical access through its current sales channel.

The use case is part of that definition. It identifies the circumstance that causes the customer to act: the triggering event, the work that must be done, the result sought, and the constraints surrounding the decision. Jobs to Be Done theory makes the same essential distinction: customers make choices in particular circumstances because they are trying to make progress, not merely because they share demographic characteristics.

This matters during a transition from consulting-led work because the company is not yet choosing its permanent market position. It is choosing the best-supported starting position from which it can make sales, delivery, onboarding, pricing, and product decisions more consistent. The first ICP should be narrow enough to guide action but provisional enough to change when better evidence appears.

What the research says about narrowing

Segmentation is a decision process, not a labeling exercise

A 2021 systematic review examined 88 papers on business-to-business market segmentation and concluded that the field still lacks a single comprehensive formula for creating robust segments. The review describes segmentation as an ongoing process involving preparation, segmentation, implementation, and evaluation rather than a one-time analysis. That finding is important here: leadership cannot outsource the choice to a standard template or a universal score. It must compare evidence, make a bounded decision, and revisit it as results accumulate.

Earlier research reached a related conclusion. A study of 200 Dutch firms found that organizational strategy predicted industrial buying behaviour better than traditional variables such as company size and location in the market examined. Research on mature industrial markets also found that size, industry, and product benefits alone were often inadequate without understanding the customer’s tradeoff between price and service. These studies do not establish a universal segmentation formula, but they show why an ICP must describe behaviour and circumstances, not just the appearance of an account in a database.

That means two companies of equal size in the same industry may belong in different ICPs. One may have an urgent executive mandate, usable data, budget authority, and a strong need for a standardized result. The other may regard the problem as optional, require extensive custom integration, and purchase only through a lengthy procurement process. Their firmographics match; their likelihood of becoming good customers does not.

Start with the customer situation and expected result

The use case should be stated as a situation, not a feature list. A practical form is:

When [trigger or circumstance] occurs, [specific organization and buyer] needs to [complete a task or solve a problem] so that it can achieve [measurable result], despite [important constraint or current workaround].

For example, “mid-sized manufacturers need analytics” is too broad. A more useful use case might identify manufacturers whose operations teams must reconcile production and quality data before a weekly planning meeting, currently spend several employee-days assembling it, and need a dependable report by a fixed deadline.

This wording forces the company to identify the actual buying circumstance. It also reveals what the product or repeatable offer must do, which data or integrations it needs, how quickly value must appear, and who must care enough to pay.

Business purchasing further complicates the profile because the user, economic buyer, technical evaluator, procurement function, and executive sponsor may be different people. Organizational buyer research has long emphasized that industrial purchases involve multiple participants, varying objectives, and contextual influences. An ICP that names only “the customer” without identifying these roles is unlikely to support a repeatable sale.

Use hypotheses, tests, and disconfirming evidence

The initial ICP is a hypothesis about where the company can repeatedly create and capture value. It should be treated accordingly.

Recent randomized research provides unusually direct support for disciplined entrepreneurial testing. A 2024 study combined four randomized controlled trials involving 759 firms. Entrepreneurs taught to frame theories, state hypotheses, and test them were more likely to terminate weak ideas and make a small number of focused pivots rather than either refusing to change or changing repeatedly. The researchers also found evidence of better performance, while cautioning that results can vary by context.

For ICP work, the practical lesson is not that leadership needs laboratory experiments. It is that it should state what would have to be true for a segment to be attractive and what evidence would prove the belief wrong.

A candidate ICP might rest on hypotheses such as:

• The problem recurs at least monthly and has a visible operational or financial cost.
• A named budget owner is willing to fund the result.
• Most of the work can be delivered through the same process.
• The customer can provide the necessary data without exceptional intervention.
• The result can be demonstrated within the expected buying and onboarding period.
• Support requirements will not erase the expected margin.
• Similar prospects can be reached without relying exclusively on the founder’s personal relationships.

These are testable. “The market is growing,” “customers like the idea,” and “the product could be used by many industries” are not sufficient substitutes.

Learn from advanced customers without mistaking them for the whole market

Eric von Hippel’s research on lead users shows that customers experiencing an important need ahead of the broader market can reveal valuable product requirements and even develop their own solutions. Such customers can be especially useful when the company is looking for repeated pain and technically informed feedback.

Lead users also create a risk. Their urgency, expertise, internal resources, or tolerance for unfinished products may be unusual. A technically sophisticated customer that has already built half the solution internally may be an excellent design partner but a poor representation of an ordinary buyer.

The initial ICP can include lead users, but it should explain why their need and buying behaviour are likely to recur. Leadership should separate:

• evidence that a problem is important;
• evidence that other customers experience it;
• evidence that those customers will purchase a solution;
• evidence that the company can deliver the solution without depending on exceptional customer capability.

Build the evidence before writing the profile

The quality of the ICP depends on the records examined before anyone begins drafting. The team should assemble evidence at the level of individual opportunities and customers, not rely on general impressions from sales meetings.

The minimum useful dataset usually combines five views of the same work:

The team should review successful customers, failed projects, stalled opportunities, lost deals, and customers it does not want to repeat. Exclusion evidence is often more useful than a polished description of successful accounts. A company that studies only its best relationships may accidentally define the ICP around customers who received unusual founder attention, favourable pricing, or custom work that cannot be repeated.

Give stronger weight to behaviour than stated interest

Customer interviews are useful for learning language, circumstances, decision processes, and constraints. They are weaker evidence of actual demand when they ask people to predict whether they would buy.

A 2020 open-access meta-analysis of 77 studies found that hypothetical estimates of willingness to pay exceeded real willingness to pay by about 21% on average. The research concerned consumer goods rather than consulting-led business software, so the percentage should not be transferred directly to a business-to-business price decision. The broader warning remains relevant: statements made without a payment obligation can overstate real purchasing behaviour.

Evidence should therefore be weighted approximately in this order:

1. Renewals, repeat purchases, expansions, or repeated paid work for the same use case.
2. Completed paid projects with verified results and known delivery costs.
3. Signed contracts, deposits, or paid pilots with new customers matching the proposed criteria.
4. Lost-deal and churn evidence that clarifies disqualifying conditions.
5. Detailed interviews about past behaviour and current workarounds.
6. Expressions of interest, survey responses, waiting lists, and general market-size estimates.

This is not a rigid statistical hierarchy. A signed deal can still be a poor fit, and one failed project may reflect weak execution rather than a bad segment. The point is to separate committed behaviour from encouraging conversation.

Correct for the warm-network effect

Founder-led referrals and warm introductions are appropriate early sales channels. They reduce the trust a new supplier must establish and can lead to high-quality conversations. They can also distort ICP confidence.

Social networks tend to connect similar people, and referral behaviour frequently follows those similarities. A 2025 open field study of job referrals, for example, found substantial same-gender referral patterns linked partly to the composition of close friendship networks. That study is about employment, not customer acquisition, but it illustrates the general mechanism: network-derived samples are not random samples of the market.

For a consulting-led company, this leads to two opposite risks:

False confidence: Warm contacts buy because they trust the founder, tolerate an unfinished process, or accept custom work. The company interprets the sale as proof that a broader segment will buy under ordinary conditions.

False narrowing: The network contains mainly one industry or professional group. The company assumes that group is the best market because it is the only one from which it has obtained enough introductions.

The remedy is not to abandon referrals. It is to record the role the relationship played and deliberately seek a small amount of evidence beyond the founder’s closest network. A paid pilot from a second-degree introduction, a sale led by another employee, or a customer acquired without prior personal trust provides a different and often valuable test.

Test delivery and economics at the same time as demand

A segment is not attractive merely because customers want the result. The work must also be deliverable at a cost, speed, and quality the company can sustain.

Research on professional-service productization describes standardization and modularization as ways to define an offer, make delivery more reproducible, improve communication, and balance efficiency with necessary customization. Research on business-process standardization similarly frames standardization as a means of improving consistency and optimizing costs and benefits. Neither body of work implies that all judgment should be removed; rather, it supports identifying which parts should repeat and which legitimately require professional discretion.

For each candidate ICP, the team should reconstruct several completed engagements and distinguish among:

• work that repeated with little variation;
• configurable work that followed known rules;
• expert judgment that remains necessary;
• customer-specific exceptions;
• rework caused by unclear scope;
• delays caused by missing customer inputs;
• support that continued after the expected delivery period.

The narrow initial use case should concentrate demand around the first two categories while making the third explicit. If most of the value depends on uncontrolled exceptions, the company may have identified a profitable consulting niche, but not yet a repeatable offer or useful software starting point.

flowchart LR
    A[Sales, customer, delivery, and financial records] --> B[Candidate customer and use case]
    B --> C[Observable inclusion and exclusion rules]
    C --> D[Test against paid cases and failed cases]
    D --> E{Evidence sufficient for a working decision?}
    E -->|Yes| F[Leadership approves a provisional ICP]
    E -->|No| G[Narrow the hypothesis or collect missing evidence]
    F --> H[Use in sales, delivery, and product decisions]
    H --> I[Review results and update]

The diagram shows a repeating decision cycle: evidence produces a candidate profile; the profile is tested against real cases; leadership either approves it provisionally or identifies the evidence still missing; actual results then update the profile.

Write an ICP one-pager that changes decisions

A useful ICP one-pager is not a marketing persona. It should be specific enough that two employees looking at the same prospect will usually reach the same qualification decision.

The document should contain the following elements.

Customer and buying situation

Identify the type of organization using observable characteristics, but include only characteristics that affect the problem, buying process, delivery, result, or economics.

Relevant variables may include industry, operating model, regulatory environment, location, business maturity, installed technology, data availability, team structure, transaction volume, and implementation capability. Company size belongs in the profile only if size changes the need, budget, buying process, or delivery burden.

Name the primary user, problem owner, economic buyer, technical evaluator, and executive sponsor when those roles differ.

Trigger and use case

Describe the event or condition that causes the customer to act. Examples include a regulatory deadline, a volume threshold, a system migration, a failed audit, a new executive mandate, rising service costs, a merger, or an inability to meet a recurring reporting requirement.

Then state the task and result in ordinary language:

• what the customer needs to accomplish;
• what it does today;
• what the current approach costs or prevents;
• what result it expects;
• how quickly that result must appear.

Inclusion criteria

Inclusion criteria should be observable before or during qualification. Each criterion should have a reason.

A criterion such as “has at least 100 employees” is weak unless the team can explain why that threshold predicts the problem or the economics. “Produces the report weekly, requires at least two employees to reconcile the source systems, and has an operations leader accountable for the deadline” is more useful because it connects directly to the use case and buyer.

Good criteria commonly cover:

• presence and frequency of the problem;
• urgency or trigger;
• accountable buyer;
• willingness and ability to pay;
• necessary data, systems, or customer participation;
• acceptable implementation complexity;
• expected result;
• compatibility with the current delivery model;
• reachability through the current channel.

Exclusion criteria

Exclusions protect focus and economics. They should identify customers that look attractive at first but consistently produce poor sales, delivery, support, or product outcomes.

Possible exclusions include:

• no named problem owner;
• problem occurs too rarely to justify action;
• expected result depends on data the customer cannot provide;
• requirement for unsupported regulatory, security, or geographic conditions;
• procurement period incompatible with the company’s resources;
• demand for ownership of custom intellectual property;
• integration work that exceeds the repeatable implementation;
• expected service level that makes the account unprofitable;
• request is related to the product but belongs to a different use case.

Exclusion does not necessarily mean “never sell.” It may mean “do not use this customer to define the initial offer or product.” A large strategic account can remain a separate consulting opportunity without being allowed to distort the repeatable core.

Evidence and unresolved assumptions

The one-pager should name the evidence supporting the profile: paid customers, repeated projects, observed results, delivery records, renewal behaviour, known margins, and lost-deal patterns.

It should also list important assumptions still being tested. A statement such as “supported by four paid customers in one referral network; willingness to buy outside that network remains unproven” is more useful than an unsupported claim of high confidence.

How to compare candidate ICPs

The following table distinguishes common candidates that can otherwise look equally appealing:

The strongest initial choice is normally not the largest theoretical market. It is the candidate for which demand, result, delivery, economics, and access align with the fewest unsupported assumptions.

Measure confidence without pretending certainty

“ICP confidence” should describe the quality and consistency of the evidence behind the decision. It should not become a decorative percentage produced by averaging opinions.

A practical confidence review examines seven dimensions:

Leadership can classify each dimension as unknown, hypothesis, repeated evidence, or decision-ready. These labels are preferable to false numerical precision.

“Decision-ready” does not mean certain. It means the evidence is strong enough to make a reversible operating commitment: use the ICP in qualification, package one offer around the use case, measure results, and review the decision after a defined period or sample of customers.

Leadership approval should therefore answer four questions:

1. Is the chosen customer and use case better supported than the alternatives?
2. Are the remaining assumptions explicit and testable?
3. Are the exclusions strong enough to prevent custom work from quietly redefining the offer?
4. Will sales, delivery, and product actually use the profile to make decisions?

Approval is weak when it means only that executives like the wording. It is credible when leaders agree to its consequences: which opportunities will receive priority, which requests will be declined or separated, which delivery process will be improved, and which product capability will be built first.

The appropriate confidence threshold varies. A company selling a low-cost product with fast onboarding can test an ICP with more customers and lower individual risk. A company selling to regulated enterprises through long contracts may need deeper technical, legal, security, and implementation evidence before committing. A services-heavy offer may initially tolerate more expert judgment than a standalone software product, but it still needs to know where that judgment occurs and what it costs.

Examples, failure modes, and tradeoffs

Basecamp: a repeated internal problem became a clear initial use case

Basecamp’s official history describes its origin inside a web-design firm that was struggling to coordinate growing client work through email, chat, spreadsheets, calls, and scattered documents. The company built a project-management system for its own work, used it with clients, and observed clients asking to use it themselves. According to the company’s retrospective account, the product generated more revenue than web design roughly a year after release, leading the firm to leave design services.

The useful lesson is not that internal tools automatically become products. Basecamp had several reinforcing forms of evidence:

• the problem occurred repeatedly in real delivery;
• the firm experienced the problem directly;
• the tool was used in live customer work;
• clients independently expressed demand;
• the initial result was clear: coordinate projects, communication, feedback, and milestones more reliably.

Its use case emerged from observed work rather than an abstract market category. The case remains a company-authored account, and it does not disclose every failed assumption or economic detail. It should be treated as an illustration, not a controlled test or universal service-to-software formula.

Veeva: a narrow industry can organize requirements

Veeva’s public filings describe a company founded in 2007 on the premise that industry-specific business problems should be addressed through tailored cloud solutions. Its products were built around the business, functional, and regulatory requirements of life-sciences organizations rather than a generic customer-relationship-management market.

That focus remained visible years later. Veeva reported 1,552 customers at January 31, 2026, while continuing to describe its core business as industry cloud software, data, and consulting for life sciences.

The case suggests how a narrow vertical can align language, workflows, compliance requirements, buyers, product capabilities, and sales expertise. It does not prove that narrow vertical positioning caused Veeva’s performance, nor that every software company should choose one industry. A vertical is useful only when industry membership predicts important common needs and a repeatable buying and delivery pattern.

Common ways the work looks complete when it is not

The profile is only a list of firmographics. Industry and company size are easy to search, so they dominate the document. The team has not explained the trigger, buyer, problem, result, or delivery conditions.

The company describes its favourite customers rather than repeatable customers. The selected accounts are enjoyable, prestigious, or personally close to the founder, but they do not share a reproducible use case.

One large customer defines the market. A high-revenue account creates the impression of strong demand while depending on custom features, special pricing, executive access, or unusual implementation work.

Interest is treated as willingness to pay. Interviewees confirm that the problem is important, but no one has bought under realistic scope, price, and timing.

Demand is considered without delivery. The profile identifies customers who buy, but each engagement requires a different process or extensive senior involvement.

Gross revenue hides poor economics. The company compares contract values without accounting for presales work, implementation labour, rework, support, concessions, and founder time.

Exclusions are polite rather than operational. The document says “not a fit” but gives sales no observable reason to disqualify an opportunity.

Leadership approval freezes the hypothesis. The ICP becomes a political commitment. New evidence is interpreted defensively instead of being used to refine the profile.

Important tradeoffs

Narrowing increases clarity but can exclude revenue. That tradeoff is real. The goal is not to stop all exceptional work immediately. It is to prevent exceptional work from silently becoming the design centre of the next offer or product.

A company may also have too little evidence to choose confidently. In that case, the answer is not to write a broad ICP. It is to choose the best-supported hypothesis, state the low-confidence areas, and design the next sales or delivery tests to distinguish among the remaining candidates.

In a small or emerging market, very few customers may exist. The ICP may need to be defined more by use case, operating conditions, or buyer behaviour than by a statistically large segment. In a regulated or high-value market, one or two customers can provide deep technical evidence, but the company should remain explicit about what has not yet been replicated.

The decision to make

The work is complete when leadership can make a specific, reversible commitment:

For the next stage of the transition, we will prioritize this observable type of organization, when this triggering situation exists, because it repeatedly pays to achieve this result. We can deliver that result through this bounded process at acceptable cost. We will exclude these conditions, and we will test these remaining assumptions before broadening the market or building further product scope.

The recommended sequence is:

1. Reconstruct the relevant sales, customer, delivery, and financial evidence from the preceding work.
2. Group customers by trigger, use case, buyer, result, delivery pattern, and economics—not firmographics alone.
3. Identify two or three candidate ICPs and compare them using the same evidence dimensions.
4. Select the narrowest candidate with repeated paid demand and workable delivery, not the candidate with the largest theoretical market.
5. Write observable inclusion and exclusion rules.
6. Record evidence, counter-evidence, and unresolved assumptions directly in the one-pager.
7. Have leadership approve the operating consequences, not merely the wording.
8. Use the profile in real qualification, scoping, onboarding, delivery, and product decisions.
9. Review it after a defined body of new evidence, especially customers acquired beyond the founder’s closest network.

Before the company depends on the result, the following should be true:

• Sales can explain whom to prioritize and whom to disqualify.
• Delivery can identify the standard process, required inputs, and legitimate exceptions.
• Finance can estimate the full economics of serving the profile.
• Product can connect the use case to one bounded set of capabilities.
• Leadership understands which claims are proven, which are inferred, and which remain hypotheses.
• At least some evidence comes from committed customer behaviour rather than stated interest.
• The company can name conditions that would cause it to revise or abandon the profile.

The result is not certainty about the final market. It is a disciplined starting position: one customer, one buying circumstance, and one result supported strongly enough that the company can stop rebuilding its strategy for every sale.

Sources

Primary and official sources

• Basecamp, “Where We Came From,” official company history.
• Christensen Institute, “Jobs to Be Done Theory.”
• Harvard Business School Working Knowledge, “Clay Christensen’s Milkshake Marketing.”
• MIT, Eric von Hippel’s collected research on lead users and user innovation.
• Veeva Systems, Form 10-K describing its industry-cloud premise and life-sciences focus.
• Veeva Systems, fiscal-year 2026 results.

Open research

• Camuffo et al., “A Scientific Approach to Entrepreneurial Decision-Making: Large-Scale Replication and Extension,” Strategic Management Journal, 2024.
• Mora Cortez, Højbjerg Clarke, and Freytag, “B2B Market Segmentation: A Systematic Review and Research Agenda,” Journal of Business Research, 2021.
• Verhallen, Frambach, and Prabhu, “Strategy-Based Segmentation of Industrial Markets,” Industrial Marketing Management, 1998.
• Sheth, “A Model of Industrial Buyer Behavior,” Journal of Marketing, 1973.
• Schmidt and Bijmolt, “Accurately Measuring Willingness to Pay for Consumer Goods: A Meta-Analysis of the Hypothetical Bias,” Journal of the Academy of Marketing Science, 2020.
• Hederos et al., “Gender Homophily in Job Referrals,” Labour Economics, 2025.
• Shamsuzzoha, Blomqvist, and Takala, “Service Productisation Through Standardisation and Modularisation,” International Journal of Sustainable Engineering, 2023.
• Goel, Bandara, and Gable, “Conceptualizing Business Process Standardization: A Review and Synthesis,” Schmalenbach Journal of Business Research, 2023.