By Ryan Richardson · Published 8 October 2026
Resisting the obvious additions is step 59 (Part IX), marked hard difficulty, almost nobody does this, manual (there's no tool that resists an idea for you).
An exit-intent offer to people leaving looked right: standard practice, catching traffic that costs nothing extra. It made the funnel worse, and got removed. A more sophisticated statistical testing setup looked right: satisfying to build, genuinely better maths. It barely moved anything, while plain detail on the page moved a lot more. An always-on retargeting layer looked right: sensible on paper, everybody runs one. The platform's own broad targeting was already doing it, better than a manually built layer would have.
All three were satisfying to build. That's the only thing they had in common, and it's the actual warning: building something is its own reward, independent of whether it was the right thing to build, and that reward arrives before the data does.
Before adding anything, write four lines: what it's expected to do, and the number that would show it, run as a test with a defined off switch. Anything that can't be turned off cleanly isn't a test, it's a commitment dressed up as one. Decide in advance what result would trigger removal, then remove it when that result arrives. Treat any model's suggestions as hypotheses, never a plan: it will recommend every standard addition because it's read a great deal about them, with no way of knowing whether any of them worked for this particular funnel.
Run on existing tools, not just proposed additions: if this disappeared tomorrow, what would break? Run against five tools in weekly use, two passed immediately, three didn't, including one kept out of habit long after it stopped doing anything the others weren't already covering.
Adding something because it's standard practice rather than because a four-line prediction says it should help. Building an addition with no off switch, which turns every subsequent decision about it into a bigger one than it should be.
| Claim | Value | Source |
|---|---|---|
| Three obvious additions and their outcomes | exit-intent offer made the funnel worse; statistical testing upgrade barely moved anything; retargeting layer duplicated existing broad-targeting delivery | The Sixty Steps manuscript |
| Disappearance test result on existing tools | run against 5 tools in weekly use; 2 passed immediately, 3 didn't | The Sixty Steps manuscript |
| Matrix status for this step | Hard difficulty, almost nobody does this, manual tooling | The Sixty Steps matrix |