Marketing is a query

The problem, encoded#

To build a successful business you need three things:

1. Something easy to build and sell.
2. A segment of people who fit what you build.
3. A segment you can actually reach within your budget and ability.

Some products are easier to sell than others. If you build on an existing system people already know (WordPress, Shopify, Google, Amazon), it's easier than inventing something new. Some are AI-proof and some are AI-exposed. And some segments are easy to reach, some hard, some impossible.

Think of 8 billion people as a database. Marketing is a way to query it:

• By dimensions: nation, language, age, gender.
• By interests: tennis, books, programming, WordPress, math.
• By intent: people searching for "website", "catalog", "AI for X".
• By existing communities: a Facebook group, an influencer's followers, a newsletter's readers.

There are 100+ ways to query, and they're twisted together. A fan page with 2,000 active members who buy what the owner recommends is one query. And every query has a cost: some need money (ads, sponsorships), some need work (posting, creating, building).

The test that makes it real#

"Women 25–35 who like tennis" is a filter with no index behind it. "The 2,000 active members of this group" or "people searching WooCommerce B2B pricing" is a query that returns real people. That's the difference between segments that exist in theory and ones that exist in reality.

This is close to what Brian Balfour calls the four fits: product ↔ market, product ↔ channel, channel ↔ business model, model ↔ market. Arriving at a known framework on your own is a good sign.

Four corrections#

1. You can't score options from your desk. Before you try something, your guesses for "easy to sell" and "reachable" are mostly noise. You only learn an option's real value by putting an offer in front of real people. So this isn't optimization (compute the best point, go there). It's exploration where each test costs something, which needs a different strategy.
2. "Easy for me" isn't a side factor. It's what makes the space searchable. Every option has a distance from you: the cost to learn it, reach people and earn trust. Seen that way, the infinite space shrinks to a small area around where you stand. And whatever is easy for you but hard for others is your edge.
3. AI changes the order. AI made building cheap, not reaching people. So often: find a query you can run first, then build for it. Owning a query (an audience, a customer list, a community's trust) is becoming the moat.
4. Time is missing. Does it compound (SEO, audience, reputation, recurring revenue)? Would you still want to do it in year 3?

How people actually search an infinite space#

Nobody searches the whole space. People search the area right around them and let the world hand them data. Stuart Kauffman calls this "the adjacent possible".

Start from what you have (effectuation)#

Saras Sarasvathy studied expert entrepreneurs and found they mostly don't go "goal → analyze market → execute". They go the other way:

• Bird in hand: start from who I am, what I know, who I know.
• Affordable loss: don't ask "what's the expected return?" Ask "what am I willing to lose on this try?" (2 weeks, $500).
• Crazy quilt: partner with people who commit early. They help set direction.
• Lemonade: use surprises as input instead of treating them as noise.

Why "incidents" work#

You often hear that people succeed because of some incident in their life, not from rational analysis. It isn't magic. Look at what one incident gives someone, in the terms above:

• Product: they know exactly what's needed, because they felt the pain.
• Segment: "people like me", so they already know where to find them. The query is known.
• Easy for me: they're already inside that world.
• Evaluation: the world tested the option for them, for free.

Two caveats. Survivorship bias: thousands had incidents and failed, and the winners tell their story backwards. You have to notice it: "chance favors the prepared mind" (Pasteur). Your analysis is what lets you recognize which incident matters.

Make incidents more likely#

Luck surface area ≈ doing × telling. The more you ship, and the more people see it, the more incidents come to you. In Richard Wiseman's research, "lucky" people mostly noticed more and said yes more.

Rent a query engine#

Marketplaces like WordPress.org, the Shopify App Store and the Chrome Web Store are ready-made indexes where people already search with clear intent. That's why building on existing systems is easier: you rent their query engine. Rob Walling's "stair step" is built on it: make one small product in a marketplace that brings its own buyers, repeat until you control your time, then move to recurring revenue.

A plan you can run this week#

1. Map your area on one page: skills, assets (products, code, servers, content), and the queries you already own (customers, email list, followers, communities that know you).
2. Collect 20 past incidents: repeated support questions, odd custom requests, tools you built for yourself, "can you do this for me?" moments.
3. List 10–20 ideas only 1–2 steps from you. Score each 1–5 on your dimensions plus AI-proof and compounding. Spend at most a day. Scores rank; they don't decide.
4. Run 3 small bets at once. Each needs an affordable loss (say, 3 weeks), a query you can actually run, and a real signal: someone pays, pre-orders or leaves an email. "Nice idea" doesn't count.
5. Keep an incident log. Write down anything surprising and review it weekly.
6. At the end of each bet, stop or double down. Winners get more weeks. Stop losers without guilt.

Analysis decides where you stand. The world decides what happens to you there. So stand close to what you know, and make lots of cheap contact with people.