How to run a first marketing experiment before you build

Research without a test is a notebook. Building without a test is a bet you cannot afford to lose twice.

You already know how to find app ideas people actually need, research your app audience, and tell when a complaint is a build signal. This post is the next step: one marketing experiment before you build — not a launch plan, not ads, not a hire.

Written for solo app builders, not agencies.

What a first marketing experiment is (and is not)

A first marketing experiment answers one question with evidence you can collect this week:

Will people who match my audience take a small next step when I show up with a specific promise?

It is not:

• A full funnel, content calendar, or brand kit
• Paid ads you cannot afford to learn from
• “Post every day” without a hypothesis
• Auto-replies, outreach blasts, or hiring a growth person

It is a cheap, dated test: a promise in their language, a place they already hang out, and a clear pass/fail signal.

Preconditions (skip these and you are guessing)

Before you design the experiment, you should already have:

1. A recurring problem with dated evidence — not one viral rant (build signal bar).
2. A named audience — who speaks, where, with what workaround (audience research).
3. A one-line promise in their words (“stops X so you can Y”), not “AI-powered platform.”

If any of those are fuzzy, keep researching. An experiment on invented demand only teaches you how to waste evenings.

Design one experiment in five lines

Write this on a single page. If it needs a second page, cut it.

1. Hypothesis — “Solo founders who already complain about X will click / reply / join when I offer Y.”
2. Audience slice — one speaker cluster, not “everyone with the keyword.”
3. Channel — one room you can reach without inventing distribution (a subreddit, a thread type, a review reply habit, a waitlist shared in places they already are).
4. Offer — the smallest next step: a waitlist that names the pain, a useful manual reply with a soft follow-up, a one-page landing promise, a short checklist PDF. Prefer usefulness before pitch.
5. Pass / fail — numbers you can count by Friday (e.g. 5 ICP-fit conversations, 10 waitlist joins from people who match, or clear “not worth it” replies). Silence counts as data.

Pick one channel and one offer. Parallel “tests” are how solos fake progress.

Three honest experiments that fit a solo calendar

1. Promise page + waitlist

Ship a single page that states the complaint-shaped promise and collects email. Share it only where your audience already talks — or reply manually in threads that clear a worth-reply bar and offer the page as an optional next step. Measure joins and whether joiners sound like your ICP.

2. Manual reply thread

Find a handful of fresh, specific asks from people who match your audience. Answer helpfully by hand. Soft-offer a next step only after you were useful. Measure follow-up questions and DMs that want more — not vanity upvotes.

3. Workaround interview sprint

Message (politely) people who already described the workaround. Ask what they tried, what “done” looks like, and whether a named promise would be worth time. Measure earnest answers, not polite “cool idea.”

None of these require a marketing hire. All of them beat three months of features nobody asked for.

How to read the result

• Pass — ICP-fit people take the next step or describe urgency. You earned a tiny build or a sharper promise.
• Fail — silence, “nice” with no urgency, or interest from people you will never serve. Kill or rewrite the promise; do not “add features” to save it.
• Messy — mixed signals. Narrow the audience slice and rerun one experiment. Do not open a roadmap.

Either way, you learned cheaper than a repo full of unused screens.

Checklist

Before you write code for the product:

1. Do I have dated evidence of the same pain across independent sources?
2. Can I name who has it and where they talk?
3. Is my promise in their language, with one next step?
4. Is there exactly one channel and one offer this week?
5. What number or outcome would make me stop — not just “learn something”?
6. Only then: build the thinnest thing the experiment actually needs (often: nothing clickable yet).

If step 5 is “I’ll know it when I see it,” you do not have an experiment. You have vibes.

Soft product bridge (honest wedge)

earlyaf. is market research and marketing guidance for solo app builders: dated evidence, realistic opportunities, and a path toward a first marketing experiment.

Public wedge today: a free reviewed daily idea, plus research that helps you decide what to test. Reply Radar (a separate surface) listens, scores, and queues Reddit/X posts that may be worth a manual reply — it does not auto-reply, send DMs, or run outreach.

Soft next step: Find ideas for my audience, or use the free daily idea when you want a concrete prompt instead of a blank experiment page.

FAQ

Do I need a landing page to call it a marketing experiment?

No. A landing page is one instrument. A careful reply thread or five earnest conversations can falsify a promise faster. Use a page when you need a durable URL for the offer — not because templates say so.

Isn’t “talk to customers” enough?

Talking helps when evidence is thin. A first marketing experiment adds a public or semi-public next step and a pass/fail bar so you are not only collecting compliments. Interviews and experiments stack; neither replaces dated complaints.

Should I run ads for the first test?

Usually no. Ads teach creative and bidding when you already know the promise lands with the right people. For a solo first test, organic rooms plus a clear offer teach more per hour.

How does this connect to Reply Radar?

Reply Radar ranks posts that might deserve your attention so you can run the manual-reply experiment without drowning in noise. Scoring is not engaging. Engaging is still you, writing a human reply — never auto-reply or spray CTAs.