Find one acquisition channel that works before you try a second

By Meet Patel · 2026-10-03 · 5 min read

Gabriel Weinberg and Justin Mares catalog nineteen customer acquisition channels in Traction. The list, as reproduced in a summary of the Bullseye framework, runs from search engine optimization, content marketing and email to sales, trade shows, community building, public relations, engineering as marketing and existing platforms. A three-person startup that reads it will usually feel an urge to try six of them at once, because each one looks plausible and each one has a success story attached.

The arithmetic of that urge is unfavorable, and this post works through it: why an early company should find one acquisition channel that works before it spreads effort, how the Bullseye process picks the first one, and a set of rules for running the tests so the results can be read.

What Traction proposes

The authors' Bullseye framework has three concentric rings. In the outer ring you brainstorm marketing ideas across the channels. In the middle ring you place the ideas that look best, based on your experience. In the inner ring you place the best-performing ideas and focus on them. Between the middle and inner rings sits a testing step, which the framework runs at high tempo with cheap tests and quick iteration.

The same book sets what it calls the 50% rule. According to a reader's excerpt, traction and product development are of equal importance and each should get about half of your attention. Taken together, the two ideas set a budget: roughly half of a founder's time is available for getting customers, and the Bullseye process decides where that half goes.

Why one channel first

The first reason is signal. Take a hypothetical startup with 20 hours a week for acquisition, trying cold outreach as one of six channels. Its share of the time is about 3.3 hours, enough for roughly 30 personalized emails a week. If the true reply rate is 3%, the team expects one reply a week. With 30 emails, a single reply more or less moves the observed rate by 3.3 percentage points, so four weeks of data (120 emails, three or four replies) cannot separate a 3% channel from a 6% channel.

Give that channel the whole 20 hours and the picture changes. At about 200 emails a week, four weeks produce 800 emails and around 24 replies at the same 3% rate. A reading based on 24 events can support a decision, while one based on three events invites whatever story the team already wanted.

The second reason is learning that accumulates. Within a channel, early weeks teach you things the later weeks use: which list sources are accurate, which subject lines open, which objections come up on calls. That knowledge carries forward only if the same person runs the channel repeatedly. Six channels run at one-sixth speed each learn at a fraction of the rate.

The third reason is attention, which the 50% rule already frames. Half of one founder's week is a fixed quantity, and every additional channel adds setup cost (accounts, templates, tracking) before it adds a single customer.

Choosing the middle ring

Brainstorming across nineteen channels produces a long list, and the narrowing step needs a rule. I would score each candidate on three questions. Can I reach the buyers there, in a place where they already look for a solution? What would one customer plausibly cost through this channel compared with what one customer pays? How quickly will a test produce a readable result?

The third question deserves extra weight early, because the scarce resource of a young company is learning time. Direct outreach to 100 named buyers can produce a readable result in a week. Search engine optimization may have a lower long-run cost per customer, but a signal can take months to appear. Long-lag channels are not excluded. The sensible order is to run first the one whose result arrives within weeks, and to start the slow ones once that channel is running.

Running the Bullseye with kill rules

The framework tells you to test the middle ring. What it leaves to you is how to judge a test. My rule is to write four things for each candidate before spending any time on it:

1. Cost to test: hours and money, with a hard cap.
2. Time to a readable signal: the number of weeks, and the minimum volume needed for a result you would trust.
3. The pass threshold: the number that counts as working, decided in advance.
4. The kill date: the day the test ends whatever the result looks like.

Fixing the threshold first removes the most common failure, in which every result is read as encouraging because nothing was defined as discouraging.

A worked example, with invented figures

Take a hypothetical three-person company that sells document-collection software to accounting firms. After brainstorming across the nineteen channels, it puts three in the middle ring: direct outbound to firm owners, a monthly webinar for practice managers, and a listing in the app marketplace of a popular accounting platform. Each gets 30 hours across three weeks.

• Outbound. Threshold: at least 3 booked calls per 100 contacts. Result: 4 per 100. Pass.
• Webinar. Threshold: at least 8 qualified sign-ups per session. Result: 3 on average. Fail.
• Marketplace listing. Threshold: at least 10 installs in three weeks. Result: 2. Fail.

The company moves to outbound and puts the full acquisition half of its time behind it. It also writes the exit condition: if the cost per booked call rises by half, or the reply rate halves, the channel has stopped scaling and the next one comes off the middle ring. Until then, the webinar and the listing stay parked. Parked is different from dead: a failed test measures the channel and the execution together, and the team records what it would change in the next attempt.

Where this goes wrong

• Picking the channel you enjoy. A founder who likes writing will rate content highly. The test needs a threshold that a preference cannot argue with. A content channel that did not pay back is described in I published 37 blog posts. They got 48 views.
• Reading one weak execution as a verdict. Poor copy and a poor channel look alike in the first test. Allow one revision before you kill a channel that otherwise looks plausible.
• Counting attention instead of customers. Followers, impressions and webinar registrations are inputs. The threshold should be a count of paying customers or booked sales conversations.
• Adding a second channel before the first saturates. The exit condition above exists for this reason. Switch when the numbers say the channel has stopped scaling, and leave boredom out of it.
• Choosing a channel that does not fit the audience. If you sell to a narrow group, the channel has to reach that group densely. This is the thread running through day-one density and the audience trap.

The principle

A company with limited hours has to choose how it learns, and a single channel run at full tempo teaches the most per hour. Until one channel works, the weekly question is which test comes next, with a threshold written before it starts. Once one works, the question becomes how long it keeps scaling. In both cases the company knows which channel is carrying its growth and which numbers would tell it to change.

Written by Meet Patel — startup operator and growth strategist in Dubai.

Read on themeetpatel.com