Do Half, Then Go on a Killing Spree
Build half the features, then remove things until the one core action runs at 10x velocity.
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 3
- Confidence
- 92%
Belsky's hard-won rule is that whatever you think you need to build, you probably only need half of it. The complement is active subtraction: whenever you add something, ask what you can replace or remove, and be willing to kill features and controls so the single most important metric accelerates. Making the whole product about one thing gets everyone doing that thing, and the core crank spins roughly 10x faster.
Origin
Scott Belsky's learning from operating Behance (launched 2008), captured in his widely shared thread of product lessons. He notes the same dynamic Matt Mochary describes about letting people go: things move faster with fewer of them.
Core principles
- 01Whatever you plan to build, you likely only need half of it.
- 02Anchor the whole product on the one thing you want users to do well.
- 03Adding should trigger a question: what can I replace or remove?
- 04Removing friction and options often increases the core metric, not decreases it.
- 05You usually only see what to cut afterward; sunsetting is the real skill.
How to run it
- 1
Name the one core action
Identify the single behavior that drives the business (for Behance, publishing projects, which drove traffic back to the site).
Pro tip Pick the metric that everything else is supposed to serve; it becomes your cut/keep referee.
- 2
Kill features and watch the core metric
Remove secondary features one at a time (Behance killed the tip exchange, then groups) and measure whether the core action rises.
Pro tip Killing the tip exchange, then groups, each caused more projects to be published; treat the metric response as the verdict.
Watch out Expect short-term complaints; they usually fade fast if the removal is right.
- 3
Make every addition earn its place by replacement
Institute a standing rule: whenever adding something, consider what to remove or replace so the product stays focused.
Pro tip Removing portfolio color controls drew 24 hours of angry messages, then silence, cleaner portfolios, and more of the core action.
Watch out Don't let 'someone wants this feature so badly' override the data; loud requests are not the same as core value.
In the wild
After launching overloaded, Belsky killed the tip exchange and project publishing went up; he then killed groups and publishing went up again, so the team kept removing features.
→ Focusing the product on one core action (publishing projects) made that crank run at roughly 10x velocity, which was the metric that drove the business.
Behance removed users' ability to customize portfolio colors to see if people would focus more on projects; for 24 hours users complained angrily, then never mentioned it again.
→ Portfolios looked cleaner and more consistent and users did the core action more.
Common mistakes
Adding without ever subtracting
Continuously bolting on features without removing anything dilutes the core action and slows the crank that actually drives the business.
Letting a few loud requests dictate the roadmap
The angry 24-hour reaction to removed color controls proved that intense complaints often signal habit, not value; caving to them would have re-cluttered the product against the metric.
Is it for you?
Best for
Product leaders with a feature-bloated product who need to sharpen focus on one core metric.
Not ideal for
Genuine platform products where breadth is the value proposition, or before you have data on what the core action even is.
From the transcript
“everything you think you need to do, you probably only need to do half of it”
“if you make the whole product about one thing, everyone does that, that core crank operates at like 10x the velocity”
“whenever you're adding things, consider what you can replace. Consider what you can also remove.”
“For 24 hours, we had people reaching out to us being like, "Damn you." ... After the 24 hours, we basically never heard about it…”
From the episode
Lessons on building product sense, navigating AI, optimizing the first mile, and making it through the messy middle
Scott Belsky (Adobe, Behance)