The Control Principle Prioritization Filter
When several designs solve the problem, ship the one that leaves your users most in control.
- Difficulty
- Advanced
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 92%
Substack prioritizes not by impact-vs-effort alone but by a first-principle tiebreaker: of all the ways to solve a problem, which one gives the writer and the reader the most control over their own experience? The obvious, algorithmic, extractive solution is deliberately rejected in favour of the version that holds the principle constant, even when it looks harder to pull off. It turns a vague company value into an executable prioritization filter that survives reorgs and roadmap churn.
Origin
Articulated by Sachin Monga (Head of Product, Substack) describing how founders Chris Best, Hamish McKenzie and Jairaj Sethi start from an opinion about how the internet should work. Monga credits architect Christopher Alexander's The Timeless Way of Building (1979) as the deeper source of the insight that when the people building a space are not the people occupying it, incentives corrupt the design.
Core principles
- 01Start from an opinion about how the world should work, not from a metric to move.
- 02The most obvious solution is usually the one that quietly takes control away from the user.
- 03All things being equal, ship the variant that preserves user control — even if it has more steps and looks less likely to work.
- 04A principle only counts if it can kill a feature that would otherwise 'probably just work'.
- 05Second-order effects are the real cost: things that work today can kill what your product is.
How to run it
- 1
Name the principle in one sentence
Write down the non-negotiable belief the product exists to express. At Substack it is: writers are in control of their work, their terms, and their relationship with subscribers; readers are in control of what they see.
Pro tip Phrase it so it can be violated. 'Be user-centric' can never be violated and is therefore useless as a filter.
- 2
Generate the obvious solution first
Write out the standard industry playbook version of the feature — the algorithmic feed unit, the platform-chosen recommendation, the auto-posting. Substack's obvious version was a Facebook-style PYMK unit: 'here are some Substacks you might like', inserted at the bottom of a post.
Pro tip Naming the obvious version explicitly makes the principled version's cost visible and deliberate rather than accidental.
- 3
Run the obvious solution through the control test
Ask: does this take control away from either side of the market? For the PYMK version the answer was yes — the writer no longer controls what appears in the space they own, and the reader who subscribed to one writer is served things unrelated to them.
Watch out If the answer is 'it takes a little control but the numbers are good', you have already lost the principle. Treat control as binary at the design stage.
- 4
Go back to the drawing board and ask for the maximal-control version
Reframe the question as: what is the simplest possible way to put the user maximally in control? Substack's answer was to just ask writers who they recommend and put that in the subscribe flow.
Pro tip The maximal-control version is often simpler to build than the algorithmic one — the hard part is that it depends on user behaviour, not on your code.
- 5
Ship the principled version even when it needs more things to be true
Accept that the principled version has more failure points (users must opt in, must choose well, must generate enough surface area). Ship it anyway and let the feedback loop of it working reinforce the principle.
Pro tip Log the decision and the reasoning. When the principled version works, it becomes the strongest internal argument for the next principled call.
Watch out Do not use the principle to justify shipping something with no distribution mechanism at all. The control-preserving version must still have a plausible path to scale.
In the wild
Substack saw organic cross-discovery happening (guest posts, reader profiles in comments) and knew a discovery loop would be powerful. The default playbook — a platform-generated 'Substacks you might like' unit modelled on Facebook's People You May Know, which Monga saw drive non-trivial growth at Facebook — failed the control test on both sides of the market. The team went back to first principles and instead simply asked writers who they recommend, surfacing those picks in the subscribe flow.
→ The control-preserving version drove millions of new subscriptions across tens of thousands of writers, and more than one in three new Substack subscriptions now come from the Substack network.
The team can see obvious levers — auto-posting content on writers' behalf, aggressive algorithmic discovery — that would move numbers immediately. Monga explicitly rules these out on the grounds that they would eventually kill what Substack is and produce nasty second-order effects.
→ Substack preserved the writer-in-charge promise while still growing network-driven discovery, keeping the platform differentiated from ad-driven feeds.
Common mistakes
Treating the principle as marketing copy rather than a filter
A principle that has never rejected a high-performing feature is decoration. The test of the control principle is whether it actually killed the obvious version of a feature that would have worked.
Judging the principled version on how many steps it requires
The initial objection to writer-curated recommendations was that too many things had to be true for it to work. That objection is a reason to pilot, not to kill — the feature took off far faster than anyone expected.
Is it for you?
Best for
Founders and product leaders at marketplaces or creator platforms where an obvious ad-style or algorithmic growth lever conflicts with the promise the product was built on.
Not ideal for
Commodity or infrastructure products with no two-sided trust dynamic, or teams in survival mode where any distribution mechanism is better than none.
From the transcript
“all things equal, like do the one that kind of holds constant this principle of control.”
“The most obvious way to do this would be something like, here are some Substacks you might like.”
“Does that kind of break this control principle like putting writers in charge, putting readers in charge?”
“There's a bunch of things that would probably just work, but they would eventually kill what Substack is or have all these nasty second-order effects…”
From the episode
Building Substack
Sachin Monga (Substack, Facebook)