Go to market for startups: design a channel test you can learn from

This guide is for a founder choosing how to reach an initial Australian customer segment—not an established business buying an AI transformation or a developer seeking contract work. You will leave with a test record and a decision rule, rather than a list of every marketing channel.

If you cannot yet describe a problem someone has experienced, start with the first-customer conversation guide. That page helps you prepare and record individual conversations; this one helps you design and interpret a channel test. If you already have a buyer and need to scope delivery, use the prototype-to-paid-pilot guide.

1. Specify a buyer and an offer before a channel

AWS’s go-to-market guide links market research, competing alternatives, audience access and messaging. Those are planning considerations, not proof that your particular offer has demand.

Fictional worked example: a Melbourne founder is considering a tool that drafts quote-follow-up messages for small commercial maintenance businesses. The assumed buyer is the owner who approves follow-ups; the current alternative is a spreadsheet and manually written emails. These are hypotheses to check, not research findings about that sector.

The proposed test offer is a review-only demonstration using synthetic quote records, followed by an optional, separately scoped paid pilot. It does not send customer emails or promise more sales. Record the pilot’s actual price and conditions consistently across routes; “interested” is not acceptance of a price never shown.

For the fictional calculation, assume a fixed AUD 600 total pilot price for one review-only workflow and ten synthetic records, with one review session and no live integration. This is an invented test price, not an MLAI quote, market benchmark or claim that the work is profitable. A real proposal needs its own delivery estimate and explicit commercial terms.

Write your message around the task: “See how a draft follow-up is checked against the quote before anyone sends it.” Avoid “autonomous revenue engine” or a savings percentage you have not measured. Make clear what exists today, what is a prototype and what still requires development.

2. Choose a route you can access appropriately

business.gov.au’s marketing-plan guidance connects a target market, goals, activities and a budget, then calls for analysing results and refining the plan. It does not establish a universally best channel.

These are candidate routes, not access MLAI guarantees. Follow organiser rules and obtain appropriate permission before following up. Do not scrape attendee lists or treat membership as consent. Check requirements for your actual communication channel and jurisdiction before outreach; this is not legal advice.

Keep the offer and qualification rule consistent. If the audience, message and price all differ, document that you compared two bundles of choices—not the isolated effect of the channel.

3. Complete the same twelve fields before starting

Copy these fields into your own document or save the editable text worksheet and completed example. The file includes the same completed fields, stage definitions, record log and decision shown here, plus space for actual follow-up evidence. It is an editorial planning aid, not a validated scoring system.

Reader segment, geography and buying role:
Observed problem and source of that observation:
Current alternative, including doing nothing:
Offer, deliverable, exclusions and price conditions:
Single question this test should answer:
Channel and permission/organiser conditions:
Message and next action offered:
Fit rule and denominator for each stage:
Time, cash and delivery-capacity limits:
Start, end, follow-up and decision dates:
Observed counts, costs, objections and missing data:
Repeat / change / stop decision and next uncertainty:

Completed fictional channel-test record

Synthetic results for teaching only. Every ID, status, price, cost and date below is author-created, not an observed customer result. No campaign or pilot has run. The schedule deliberately describes a future fictional decision, not a forecast. This expanded log assigns statuses to illustrate the earlier aggregate counts; it is not newly discovered customer evidence.

Reader segment, geography and buying role

A Melbourne early-stage founder testing an offer for commercial-maintenance owners who approve quote follow-ups. This is the hypothetical buyer, not the audience of a contractor-recruitment article.

Observed problem and source of that observation

Delayed quote follow-ups are the assumed problem. No interviews or customer records were supplied. Qualification statuses below are author-assigned for the exercise, not validated demand.

Current alternative, including doing nothing

Spreadsheet plus manually written emails, or leaving follow-up unchanged; both are assumptions to check.

Offer, deliverable, exclusions and price conditions

A synthetic demonstration, then optional AUD 600 total review-only pilot: one workflow, ten synthetic records, one buyer review session, no live integration or sends. Assume one full payment before the pilot review. Price and terms are invented, not an MLAI offer or proof of profitability; real scope, delivery costs, tax/payment/refund terms and permission need agreement.

Single question this test should answer

Will verified-fit owners take the next step on this specific demonstration and pilot offer through either accessible route? The example cannot determine a causal channel winner or product-market fit.

Channel and permission/organiser conditions

Recipient-agreed introductions; a separate organiser-approved opt-in session. The 28 fictional IDs do not overlap. In a real test, deduplicate recipients and record appropriate permission and follow-up limits; community attendance is not blanket contact consent.

Message and next action offered

See how a draft follow-up is checked against the quote before anyone sends it. Invite a demonstration first; show the same pilot scope and price at every attended demo. Acceptance of an introduction is not acceptance of this offer.

Fit rule and denominator for each stage

Verify owner role, a recent quote-follow-up problem and authority to consider a pilot. Individually invite only verified-fit people. Count yes/no replies per invitee, attended demos per invitee, accepted terms per terms shown, payments per accepted offer, and completed ten-record reviews per paid buyer. Unknown fit, pending replies, pending terms and pending payments stay separate.

Time, cash and delivery-capacity limits

Across both routes: at most 12 acquisition hours, AUD 250 acquisition cash and two accepted pilot commitments. Invented planning value: AUD 60 per founder hour. Pilot delivery and support effort are outside that acquisition budget and remain unestimated; two slots are a promise ceiling, not proof the founder can deliver them.

Start, end, follow-up and decision dates

Fictional future schedule: test 14–25 September 2026; appropriately agreed follow-up until 2 October; decision on 5 October. Statuses below are an author-created snapshot at that fictional decision, not events that have happened or a forecast.

Observed counts, costs, objections and missing data

Use the 28 synthetic status rows below. Introductions: 5 of 8 fit, 4 of 5 invitees replied, 3 demos, 1 accepted pilot, 1 payment, 0 completed reviews. Session: 4 of 20 fit, 3 of 4 invitees replied, 2 demos, 1 accepted pilot, payment pending, 0 completed reviews. Acquisition: AUD 40 / 4 hours and AUD 180 / 8 hours. Objection reasons and delivery economics are not supplied; see the stage definitions and cost limits, not an invented customer story.

Repeat / change / stop decision and next uncertainty

Pause new acquisition. The test has 0 hours remaining, AUD 30 cash remaining and no capacity for new pilot commitments under its two-commitment limit. Resolve the delivery plan for existing commitments before making new promises. Any further outreach needs newly agreed capacity and an appropriate contact basis; the fictional follow-up window has ended. Do not treat a pause in acquisition as cancellation of accepted terms. Delivery effort, actual operating costs, refunds, support, continued use, repeat demand and demand beyond personal contacts remain unknown. These counts cannot establish product-market fit, a profitable offer or a winning channel.

4. Interpret counts without inventing a winner

A reply, an attended demonstration, acceptance of terms, payment and actual use answer different questions. In this example, “activation” means completing the agreed ten-record pilot review. No fictional buyer has done that. The session has no paid buyer yet, so its activation proportion is not calculable—not an observed 0% success rate.

Wide tables scroll sideways. You can focus a table region and use the arrow keys, or read the same rows in the text download.

Five of eight introduced people and four of twenty attendees met the fit rule. Those denominators describe different selection processes, so their percentages are not a fair experiment proving introductions are better. One payment is too little evidence for a stable acquisition forecast. Zero payments so far does not make acquisition cost zero.

Cash, founder time and delivery promises are separate limits

Cost arithmetic: if the founder uses an illustrative AUD 60/hour value for planning, introductions consumed AUD 40 + (4 × AUD 60) = AUD 280 in cash plus valued time. The session consumed AUD 180 + (8 × AUD 60) = AUD 660. Only AUD 40 and AUD 180 are cash spending in this example; valued founder time is an opportunity-cost assumption, not a salary payment.

Combined acquisition uses AUD 220 cash and 12 hours, valued here at AUD 720: AUD 940 cash plus valued time. There is AUD 30 cash headroom but 0 hours remaining. Two accepted pilots use the two-commitment ceiling, even though only one buyer has paid. That ceiling does not establish delivery feasibility: delivery and support effort still need estimating.

These totals exclude product development, pilot delivery, ongoing support, refunds, taxes and other overhead. Do not present them as fully loaded customer acquisition cost or compare them with the fictional AUD 600 payment as though the difference were profit. Record delivery and continued use separately before judging the business model.

5. Make a bounded next decision

• Repeat narrowly: relevant buyers completed the next step and the test stayed within limits. Repeat with a new, clearly described group; check whether personal connections were doing the work.
• Change one assumption: people fit the segment but misunderstand the offer, cannot obtain approval or reject the terms. Record the objection and change the relevant assumption, not every variable at once.
• Stop or pause: access is inappropriate, the promised result cannot be delivered, capacity is exhausted or the agreed spending limit is reached. More outreach does not repair these problems.

Decision for this example: Pause new acquisition. The test has 0 hours remaining, AUD 30 cash remaining and no capacity for new pilot commitments under its two-commitment limit. Resolve the delivery plan for existing commitments before making new promises. Any further outreach needs newly agreed capacity and an appropriate contact basis; the fictional follow-up window has ended. Do not treat a pause in acquisition as cancellation of accepted terms.

Delivery effort, actual operating costs, refunds, support, continued use, repeat demand and demand beyond personal contacts remain unknown. These counts cannot establish product-market fit, a profitable offer or a winning channel. The pending payment is still pending at the fictional decision date; an expired follow-up window does not convert it into a sale or a loss. Do not infer price objections from a bare “declined” status.

Keep unsuccessful tests. A useful record explains the exact offer, who saw it, what happened and what remains unknown. Do not let an AI summary turn tentative interest into a customer win.

Record a new decision before extending a real test

The download includes blank follow-up slots. Record internal ID, route, appropriate contact basis and expiry, stage, date, actual outcome, supporting source and next allowed action in an approved private system. Separately record any new acquisition hours/cash, delivery capacity, approval owner and date. Leave unknowns explicit; no slot is a requirement to collect unnecessary personal data.

Take one unresolved decision to peers

Bring an anonymised version of your worksheet and one question—such as whether your fit rule distinguishes a user from a buyer—to a relevant founder or AI discussion. Give useful context without sharing prospect identities, confidential quotes or contact lists. MLAI events are for participation and learning, not guaranteed customer introductions.

Explore upcoming MLAI events and check the listing for its topic, format and participation requirements.

If your next task is organising company evidence for an update or content plan, explore Founder Tools; opening a workspace requires an account. Keep pending experiments distinct from confirmed results. No worksheet is automatically imported, and a polished update is not customer evidence.

Sources checked 15 September 2026. The worksheet, synthetic dataset and decision framework are editorial aids; neither linked organisation has endorsed or validated them. AI-assisted editorial example; independent founder/source review pending. This page does not claim original customer research or a completed pilot.

Frequently Asked Questions