Test the Field Before You Plant It All: Five Phases to a Customer Profile You Can Trust

“For which of you, desiring to build a tower, does not first sit down and count the cost, whether he has enough to complete it?” Jesus asked that in Luke 14:28 about the price of discipleship, and the image he chose assumes something about wise builders: they sit down first. Most founders do the opposite with their customers. They plant the whole field on an untested guess, then discover at harvest that the soil was wrong. A customer profile you can trust is not written in one inspired evening; it is built through five phases of faithful diligence: research, segmentation, definition, validation, and continuous learning. At AVODA we have trained 600+ founders and watched 150+ start-ups launch, and the difference between the ones who plant wisely and the ones who replant every season is rarely talent. It is whether they tested the field before they bet the harvest on it.

Key Takeaways

• A trustworthy ICP moves through five phases: research what the market actually says, segment it into distinct groups, define one narrow profile, validate it with real paying customers, and keep learning after launch (1, 2).
• Validation is the phase most founders skip, and the most biblical one: pilot with roughly five real customers before betting the harvest, the same counting of cost Jesus commends in the tower builder (1, 4).
• The bar for finished research is high and specific: you understand your customer when you can predict about three quarters of what they will say before they say it (4).
• Four pitfalls kill most profiles: written too broad to be testable, never validated with money, frozen in time while the market moves, and held by the founder alone while the team sells to somebody else (2, 3, 5).
• In our market the research happens where the customers already are: WhatsApp threads, market days, and mobile-money records tell you more than an imported survey template (6).
• Our alumni data shows the stakes: of 99 alumni we followed up with, 45% were running no business at all. Plans fail for lack of counsel, and a customer profile is counsel you gather before the money is spent (7, Proverbs 15:22).

Why Does the Tower Builder Sit Down First?

Because towers are expensive and half-built ones are public. The first two phases, research and segmentation, are the sitting down.

Phase one: research. Before you describe your ideal customer, listen to your actual market. Not a formal survey imported from a Western template, but disciplined attention where your customers already live: the WhatsApp groups where orders and complaints move, the market days where cash actually changes hands, the mobile-money records that show real payment rhythms (6). Talk to people who bought, people who refused, and people who bought from someone else. Write down their words, not your summary, because the customer’s own phrasing is the raw material every later phase depends on. Proverbs 18:13 is blunt about the alternative: to answer before listening is folly and shame. Most bad ICPs are answers given before listening.

Phase two: segmentation. The research will show you that “the market” is really several distinct groups wearing one label. Schools that pay by LPO at term start and shops that pay cash daily are not one segment because both buy your product. Split them honestly by the attributes that change behavior: size band, payment rail, buying trigger, who decides (1, 5). You are not choosing yet; you are drawing the map of the fields so the choice, when it comes, is between named alternatives instead of a vague gesture at everyone.

How Do You Write the Profile, and How Do You Prove It?

Phase three: definition. Choose one segment and compress it into one falsifiable sentence: type of business or household, size band, the rail they run on, and the trigger that makes them ready now (1, 5). The discipline that the best operators teach is to be almost uncomfortably narrow, three attributes chosen because they are distinctive, and to hold that narrowness until the evidence earns an expansion (1, 2). If you cannot list twenty real names from your sentence, you have written a wish. Write the negative list too: who you will decline even when their money is real, because the wrong customer consumes the hours your true flock needed.

Phase four: validation. This is where the tower builder counts the cost, and it is the phase most founders skip. A profile on paper is a hypothesis; only money and usage can promote it to knowledge. Recruit roughly five real customers who match the sentence and serve them fully: real payment, real delivery, real complaints (1, 4). This is the design-partner practice of the strongest product teams: a few committed early customers whose reality shapes the product before scale (4). Watch three things: did they pay without heroic persuasion, did they use what they bought, and did they refer anyone unprompted? Teams that publish their ICP work report exactly this pattern: profile-matched customers activate better, stay longer, and refer more, while a broad profile is worse than none because it cannot fail a test (2). Five customers is small enough to afford and large enough to expose a false guess. The pilot’s whole logic: risk one seedbed, not the harvest.

What Keeps the Profile Alive After the Pilot?

Phase five: continuous learning. A validated ICP is true about a season, not forever, and this is where the last three pitfalls live. The profile that was accurate in January drifts by August: the market moves, competitors move, your own product moves. The published warning signature is worth memorizing: pipeline growing while win rates fall usually means the market you are attracting is no longer the market you defined (3). The profile can also fracture inside the team, with the founder holding one picture while whoever answers the WhatsApp line sells to a different one; a profile that lives only in the founder’s head governs nothing (5).

The remedy is a quarterly review, and we teach founders to treat it the way the church has long treated the examen: a regular, honest examination of what actually happened, held against what you intended, in the presence of God. Sit down with the quarter’s wins and losses. Which customers matched the sentence, and did they behave as promised? Which did not match, and what did serving them cost? Where the evidence and the sentence disagree, change the sentence, never the evidence. Then re-align the team on the revised profile so everyone is farming the same field. This is not corporate ritual; it is Proverbs 27:23 on a quarterly cadence, knowing the condition of your flocks afresh, because the condition changes.

None of this requires a marketing budget. It requires humility to listen before defining, courage to test with real money, and faithfulness to keep looking after the launch. Of the 99 alumni we studied, 45% were running no business at follow-up (7). Plans still fail for lack of counsel. The five phases are how a founder gathers counsel before the harvest is on the line.

Frequently Asked Questions

What are the five phases of building an ICP?

Research (listen to the real market in its own words), segmentation (split the market by attributes that change behavior), definition (one falsifiable sentence plus a negative list), validation (a paid pilot with roughly five matching customers), and continuous learning (quarterly review against wins and losses) (1, 2, 3).

Why five customers? Why not one or fifty?

One customer can be an accident; fifty is a scale bet you have not earned. Around five paying customers who match the sentence is small enough to serve deeply and large enough to expose a false profile, the same logic behind design-partner programs at strong product companies (1, 4).

How do I do customer research with no budget?

Go where the evidence already is: your WhatsApp threads, the market day, your mobile-money statement (6). Interview buyers, refusers, and competitors’ customers, and record their exact words. The bar: you are done when you can predict about three quarters of what a customer will say before they say it (4).

What if the pilot fails?

Then the pilot worked. A failed pilot with five customers costs you a seedbed; a failed launch costs the harvest. Change one attribute of the sentence, using the words the pilot customers actually said, and test again. That is counting the cost, not wasting time (Luke 14:28).

How often should we revisit the profile?

Quarterly, as a standing examen for the business, and immediately whenever wins and losses start disagreeing with the sentence: rising pipeline with falling closes is the classic drift signature (3). Review it with the whole team so the profile governs practice, not just the founder’s slides (5).

Related Reading

• Know well the condition of your flocks: ICP, personas, and target market
• Selling to businesses is not selling to households
• Your market moves in waves: finding the first followers
• A diagnostic that earns trust: assessment design

Sources and Evidence

1. Rachitsky, L., 2023. How to identify your ideal customer profile (ICP). Lenny’s Newsletter. https://www.lennysnewsletter.com/p/how-to-identify-your-ideal-customer. The narrow-first discipline, the three-attribute sentence, and the ~five-customer validation bar synthesized from 20+ B2B companies.
2. Vandervell, A., 2024. Defining our ICP is the most important thing we ever did. PostHog. https://newsletter.posthog.com/p/defining-our-icp-is-the-most-important. First-party evidence that profile-matched customers activate and retain better, and that a broad ICP cannot be tested.
3. Cojoy RevGen, 2026. How to Fix ICP Drift in B2B SaaS. https://cojoyrevgen.com/blog/diagnose-icp-drift-b2b-saas. The drift signature (growing pipeline, falling win rate) and the case for scheduled ICP review.
4. First Round Review. Design partners. https://review.firstround.com/articles/design-partners/. The design-partner pilot model and the predict-75%-of-what-they-say bar for finished customer research.
5. Close. Ideal Customer Profile vs. Buyer Persona: What’s the Difference? https://close.com/blog/ideal-customer-profile-vs-buyer-persona. The attribute framework for definition and the team-alignment function of a written profile.
6. Techeconomy, 2026. WhatsApp: The Operating System of African SMEs. https://techeconomy.ng/whatsapp-operating-system-for-african-smes/. Why field research in African markets starts in WhatsApp threads and mobile-money records rather than imported survey templates.
7. AVODA Institute internal alumni study, 2026. Follow-up of 99 alumni: 15% registered businesses, 25% informal, 45% no business. The cost of unvalidated plans in our own data.