LeadHaste
AdvancedNo variables to fill — paste & go

Design an ICP hypothesis testing plan for a new product

Turns 'who should we sell this to?' from a debate into an experiment: explicit ICP hypotheses with reasoned mechanisms, head-to-head outbound test designs sized to your capacity, and pre-committed confirm/kill thresholds that prevent goalpost-moving. The learning log ensures even failed tests compound into targeting knowledge.

The prompt
You are a go-to-market scientist who treats early ICP definition the way a researcher treats a hypothesis: something to be tested cheaply and revised without ego, not defended in strategy meetings. You've watched new products waste their first two quarters 'targeting everyone to see what sticks' — which produces noise, not learning — and you've watched others commit to a guessed ICP so hard they ignored every signal it was wrong. Your method: 2-3 explicit ICP hypotheses, tested head-to-head with outbound, judged on pre-committed criteria.

Design my ICP testing plan. Produce:

1. HYPOTHESIS CARDS: 2-3 candidate ICPs, each stated as 'We believe [segment] will buy because [reasoned mechanism]' — with the mechanism drawn from my actual evidence, not vibes. Rank them by prior confidence.
2. THE TEST DESIGN: for each hypothesis — list size (enough for signal, small enough to run in 3-4 weeks), the message angle that tests the mechanism specifically, and the channel. The messages must differ only where the hypotheses differ, so results are attributable.
3. SUCCESS METRICS: pre-committed thresholds per hypothesis — reply rate, positive reply rate, meetings booked — and what counts as confirm, kill, or revise-and-retest. Set these BEFORE launch so I can't move the goalposts.
4. THE LEARNING LOG: the 5 fields to record per test so learning accumulates (including reply verbatims, which carry more signal than rates at small volume).
5. DECISION TREE: what to do in each outcome combination, including the awkward one where nothing works.

Before you write anything, interview me. Ask me these questions ONE AT A TIME, waiting for my answer each time:
1. What's the new product, and what problem does it solve?
2. What evidence do you have so far — early users, discovery calls, waitlist signups, related product data?
3. Which segments are you considering, and what's your instinct about each?
4. What outbound capacity do you have for testing (sending infrastructure, list tools, hours per week)?
5. When do you need a directional answer by?

Once you have my answers, produce the plan. If my capacity can't test three hypotheses properly, cut to two rather than running three badly.

How to use it

  1. 1

    Copy the prompt into Claude, ChatGPT, or any LLM.

  2. 2

    Gather every scrap of early evidence for question 2 — discovery call notes and waitlist patterns beat instinct.

  3. 3

    Write the success thresholds into a doc before sending a single email; pre-commitment is the discipline that makes this work.

  4. 4

    Run the test campaigns in parallel, not sequence — market conditions shift enough monthly to muddy sequential comparisons.

  5. 5

    Log reply verbatims religiously; at test volume, what prospects say outweighs what the rates show.

Best practices

  • Force each hypothesis to state a mechanism ('because they just lost their in-house option'), not just a segment — mechanisms are what you're actually testing.

  • Keep test messages identical except where hypotheses differ; otherwise you're testing copywriting, not ICPs.

  • Treat 'nothing worked' as data about message-problem fit before concluding the product has no market — the decision tree covers this.

  • Cap each test cycle at 3-4 weeks; ICP testing that drags loses urgency and clean comparison.

Example: what this looks like in practice

A founder launching a contractor-payments product has a waitlist of 60 and two instincts: general contractors managing subs, or specialty trade firms. The plan produces three hypothesis cards — GCs (mechanism: sub-payment complexity scales with project count), specialty firms (mechanism: cash-flow gaps from slow GC payments), and a construction-accountant channel hypothesis surfaced from waitlist patterns she'd overlooked. Capacity allows two proper tests, so the accountant hypothesis is queued. Two 300-contact campaigns run for three weeks with pre-committed thresholds of 2% positive replies. GCs land at 0.7% with replies saying 'our bookkeeper handles this'; specialty firms hit 3.4% with verbatims about payroll-day stress. The decision tree confirms specialty trades as the beachhead ICP, and the learning log notes bookkeepers as an influencer persona for later.

Best fit

Roles
Founder / CEOMarketerSales Leader
Company size
Solo founderStartup (1–10)
Audience
B2B
Industries
Any industry
Works with
Any LLM
Difficulty
Advanced

This prompt is one gear in a bigger machine. We orchestrate 20+ tools into outbound systems our clients own — and guarantee the results.

Apply for a Pilot Spot →
Prompt FAQ

Frequently asked questions

Don't pick one — test two or three. State each candidate segment as a hypothesis with a reasoned mechanism for why they'd buy, run small parallel outbound campaigns against each, and judge results on thresholds you committed to before launch. Three weeks of head-to-head testing beats a quarter of targeting everyone and squinting at the noise.