Reconstruct the conditions behind the purchase
I would build an ICP from the situations in which customers can actually buy, adopt and benefit. Industry and headcount help you find companies; they do not finish the reasoning.
Review actual wins, losses and stalled decisions. Ask what was true before the purchase, who owned the workflow, why change mattered and what made adoption viable. Separate repeated conditions from exceptional relationships or custom scope.
The initial sample can be small, so state its limits. Your strongest wins may reveal a useful hypothesis without proving market-wide demand. Do not turn one enthusiastic customer into a universal profile. Include rejected cases to understand the boundary rather than define fit only from success.
Write observable criteria and private questions separately
A researcher can verify public product tiers, integration documentation or a dated event. They cannot verify internal dissatisfaction, budget or intent from those facts. Put the observation, inference, alternative explanation and validation question into separate fields.
Use evidence confidence to describe source quality, not purchase probability. Preserve date and provenance. An outdated job title should not be presented as current authority, and a funding round should not become a sales-budget estimate.
| Decision | Evidence to use | What changes next |
|---|---|---|
| Observed | Public enterprise access features and source/date | Supports account research |
| Hypothetical | Manual changes may create operational work | Needs a neutral buyer question |
| Buyer-confirmed | Relevant person describes actual workflow and consequence | Supports qualification when other conditions are met |
Work through an ICP row
Illustrative account condition: a B2B platform offers differentiated enterprise access, has a plausible engineering/product owner and can use the standard integration. Fact: the tiers are visible on a public pricing page. Hypothesis: access exceptions may consume engineering time. Alternative: a mature internal service already handles them.
The next question is 'How do plan changes reach the product today?' A hard disqualifier is an unsupported deployment requirement. If the buyer confirms that the problem is solved, retain that answer and reject the hypothesis. The account does not become an opportunity just because the public conditions matched.
Keep the organization and the person distinct
The ICP selects organizations. A persona describes a role's work and likely decision relevance. The actual committee maps verified participants in this account. A finance director can be a plausible persona at an unsuitable company; a strong-fit platform can still require engineering, product, security and finance.
Avoid fictional persona biographies. Ask what the role can validate, approve or block. Update the map in discovery and leave unverified authority visible. Titles help research; they do not settle the decision process.
Prioritize without pretending to know intent
Apply hard disqualifiers first. Classify remaining accounts as work now, research or justified later review. Consider operational fit, evidence freshness, contact coverage, adoption and economics. High theoretical contract value is not enough if the team cannot serve the account.
A public event is a reason to investigate. Ask what decision-changing question it creates. If it creates none, it may be decorative personalization. Avoid an opaque score that converts every event into urgency. Plain reasons are easier for a seller to inspect and correct.
Maintain the boundary with rejected evidence
When a handoff is rejected, record whether the cause was weak account fit, missing role, solved problem or unsupported adoption. Return the lesson to research. Do not delete the reason and re-research the same account next month.
Review a sample with product and commercial owners. Distinguish confirmed mismatch from missing information. A temporary constraint needs a legitimate review condition; it does not justify endless follow-up. When the product boundary changes, revise the rule and test the new interpretation before reopening every old account.
Test whether another researcher can apply it
Give a colleague the profile and five unfamiliar accounts. Ask them to classify each with source, reason and next action. Compare their choices with the seller's reasoning. Disagreement exposes vague conditions that need revision.
The acceptance test is reproducible judgment, not a large list. Once the rules are usable, connect the worksheet to the client-owned CRM and the qualification process. Keep target-account value outside qualified pipeline until buyer evidence supports it.
Select accounts from evidence, not resemblance
A useful ICP should help two people make the same account decision. Replace broad labels with an observable operating condition and keep a disqualifier beside each positive signal. A company that looks attractive can still lack the workflow, ownership or ability to adopt your product.
For approval-workflow software, “moving upmarket” is a hypothesis. Ask whether larger purchases have actually introduced difficult handoffs. Record what would change your view before expanding the account list.
| Field | Evidence to collect | Decision it supports |
|---|---|---|
| Buyer condition | A recent purchase with multiple reviewers. | Include only accounts where the workflow exists. |
| Problem | A buyer-described delay, rework or risk. | Keep the pain hypothesis or revise it. |
| Owner | The person responsible for improving that workflow. | Choose whom to contact and involve. |
| Adoption | A team able to implement the proposed change. | Proceed or defer the account. |
| Disqualifier | The current workflow is already reliable, or the required scope is unsupported. | Stop this play rather than manufacture urgency. |
Connect this decision to the complete startup sales-motion plan.
Sources and research notes
- Schematicproduct documentation
Company product description
- Schematic$6.5m funding announcement
Company-issued press release
- GitLabsales development handbook
Company operating handbook
- GitLabCommand Plan
Company operating handbook
Primary sources reviewed October 6, 2026. The operating recommendations and worked scenarios are Daavid’s analysis. Illustrative numbers are assumptions, not measured client results. Company marks identify sources and prior experience; they do not imply a customer relationship or endorsement.

