How to choose your first marketing channel

Six questions that disqualify most of the list in an afternoon, and what 33 real distribution plans said when we scored the same ten channels every time

Search for the best marketing channel for a software product and you will get a ranked list: SEO, paid search, LinkedIn, events, ads inside AI assistants, partnerships, cold outreach. The lists are competently written. They are also written for a company with a marketing hire and a budget, which is why they can afford to talk about layering a paid channel on top of an organic one.

If you are one person with a working product and no budget, the ranked list is the wrong shape of answer. You do not have the capacity to run two channels badly. You have one channel's worth of attention for the next quarter, which means the useful question is not which channel is best in general. It is which eight you can stop thinking about today.

What follows is the elimination order we use. We have run the same ten channels through the same questions for 33 products now, and the pattern is consistent enough to write down: the answer came back avoid for now 163 times and focus 74 times. Most channels are wrong for most products at any given moment. That is the finding, and it is good news, because eliminating is fast.

1. Can you afford to lose the first thousand?

This disqualifies paid ads, and it disqualifies them more completely than anything else on the list. Across 33 plans, paid ads was recommended as a first move zero times. Thirty-two got told to avoid it for now.

Ads are a multiplier. You point them at a page that converts and they buy you more of what already works. If you do not yet know which sentence makes someone sign up, the ads are not acquiring customers, they are purchasing an experiment, and it is the most expensive way to run one. The recurring pattern in the plans was a founder with a few hundred dollars, which is enough to generate noise and not enough to generate a statistically meaningful answer.

The test: do you have a conversion rate you would be happy to multiply, and enough budget to buy a few hundred clicks without flinching? Two noes, and this one is closed until next quarter.

2. Is your product actually a mobile app?

This one is binary and takes ten seconds, which is why it is high on the list. App store optimization scored lowest of all ten channels on average, and was avoided 28 times out of 33.

The reason is not that app stores do not work. It is that most of the products we see are web software, and for web software the app store is not a weak channel, it is not a channel at all. What makes this worth stating is the other end: for the three products where the answer was yes, app store optimization jumped straight to focus. Almost nothing else on this list behaves that way. It is either irrelevant or it is your whole game.

The test: does a person install this from a phone? If not, delete the row.

3. Is one customer worth an hour of your time?

This is the outbound question, and it is the most misunderstood one on the list. Outbound split harder than any other channel in the data: 11 focus against 18 avoid, which means the average score for outbound tells you nothing whatsoever about your own product.

The split is not about outbound being good or bad. It is arithmetic. Researching one company properly, finding a real reason to write, and sending something a human would answer takes somewhere between twenty and forty minutes. If that company will pay you two thousand a year, this is the highest-return hour in your week. If your product is ten dollars once, or five a month to a consumer, the arithmetic never closes and no amount of tooling will close it. The plans that said avoid were almost all consumer products or very low-priced ones.

The test: multiply your annual revenue per customer by a realistic conversion rate. If the result is less than an hour of your time is worth, outbound is closed no matter how good the software gets. If it clears the bar comfortably, founder-led outbound is probably your first channel and you can stop reading here.

4. Does anyone already search for this problem?

SEO was recommended as a first move five times out of 33, and four of those five expected months before any signal. Sixteen more got test later, which is the planner's way of saying yes but not now.

Two separate things disqualify it, and people usually only check the first. The obvious one is demand: if nobody types anything resembling your problem into a search box, there is no traffic to rank for, and this is more common than founders expect for genuinely novel products. The less obvious one is the SERP itself. Plenty of products solve a well-searched problem where every result on page one belongs to a company with a decade of backlinks. Ranking is possible. It is a two-year project, and you are asking about this quarter.

You are reading this on a page built to rank, so the position is not that SEO does not work. It is that SEO is a compounding asset for a product that intends to exist in two years, and a poor answer to the question of where next month's users come from.

The test: search your own problem in the words a customer would use. If the results are irrelevant, there is no demand yet. If they are all household names, there is demand and it is spoken for.

5. Do creators exist here, and does your buyer believe them?

Creator and short-form video was avoided 21 times out of 33, and chosen twice. Two different failures produce that number.

The first is supply. In a genuinely narrow vertical there may be four people producing content and none of them have an audience worth sponsoring. The second is credibility, and it bites hardest in exactly the category most builders are in: technical buyers discount a sponsored recommendation almost to zero. Consumer products with visible before-and-after moments are the ones that clear both bars, which is why the channel looks so powerful in the case studies and so weak in the data.

The test: name three creators your actual customer follows. If you cannot, this is closed. If you can, ask whether that customer has ever bought something because one of them was paid to mention it.

6. Do you have anything a partner wants?

Partnerships is the strangest column in the table: one focus, sixteen test later, sixteen avoid. It is almost never rejected outright and almost never chosen, which is the signature of something that would probably work later.

A partnership is a trade, and the other side does arithmetic about your audience before agreeing to it. With no users, you are asking for distribution and offering nothing back. This is why partnerships read as a growth channel to founders and as a support request to the companies they approach. It becomes real the moment you have something to trade, which is usually a few hundred engaged users.

The test: could you write the sentence explaining what the partner gets, in numbers, today? If not, this is a next-year channel.

What is usually left

Six questions, most of the board cleared. What survives is generally some combination of three things, and the data has a clear favourite among them.

The highest-scoring channel across all 33 plans was not a channel in the normal sense. It was trust packaging: the pricing page that answers the obvious objection, the screenshot that proves the product runs, the about page with a human on it, the evidence somebody else already paid. It wins because it is the only item on the list that pays off without any new traffic at all. If a hundred people visited last week and two signed up, the cheapest growth available is not another hundred visitors, and fixing the order you work in matters more than picking a heroic channel.

Next to it sits community distribution, tied on focus recommendations and expecting a signal in days rather than months for most products. Going where the conversation already happens is slow, unscalable, and works, which is why doing it without getting thrown out is a skill worth learning properly. Launch platforms come third: a real spike, a short one, and worth timing rather than rushing.

Three or four survivors is a normal result, and it is one more than you can actually run. The follow-on question is capacity rather than viability: how many channels one person can carry is a separate piece, and the short answer is two, paired so that one pays this month while the other compounds.

A caution about the numbers

Every figure here comes from what our planner recommended, not from measured outcomes. It is 33 products over two months, scored by one model reading landing pages. Nobody in that set has reported back with revenue, and a model trained on the internet has absorbed the internet's conventional wisdom about marketing along with everything else. Treat it as a well-informed opinion that happens to be consistent, and argue with it where your product is genuinely unusual.

The part that will survive the outcomes data, I think, is the ratio rather than the ranking. Two thirds of the available channels were wrong for two thirds of these products: not wrong forever, wrong this quarter, at this size, for this buyer. Working out which two thirds is an afternoon of honest questions, and it is the cheapest afternoon in early distribution.