LLenny's Podcast
← All frameworks
StrategyKarri Saarinen (co-founder, designer, CEO)

Magic and Science Decision-Making

Replace A/B tests and metric goals with deep shared customer understanding, then decide by informed intuition.

Difficulty
Advanced
Time to result
~ongoing to results
Steps
3
Confidence
87%

Linear makes product decisions without A/B tests, per-feature metric goals, or a data-driven mandate. The 'science' is having the whole team immersed in customers, research, and the current state of the product; the 'magic' is that once everyone is genuinely informed of customer reality, intuition and judgment can drive decisions without needing data to justify them. Data is used to answer specific questions, not to make the choice for you.

Origin

Karri Saarinen's description of Linear's product decision culture, contrasted explicitly with metrics-and-velocity-driven companies like Ramp.

Core principles

  • 01The 'science' is customer immersion, not experimentation: user research, customer Slack, support calls for the whole team
  • 02The 'magic' is that deep shared understanding makes informed intuition reliable without data to back every call
  • 03Pull metrics to answer a specific question ('I wonder what's going on'), not to set a number to move
  • 04People lean on data out of fear of a wrong choice; you must accept mistakes and fix them rather than outsource the decision to data

How to run it

  1. 1

    Immerse the whole team in customers

    Start projects with user research; have the team join the customer Slack, answer questions, and take support calls so everyone builds real customer understanding.

    Pro tip Founders answering complaints directly in customer channels keeps leadership's understanding first-hand.

    Watch out You can't expect everyone to have full understanding; usually one or two people with deep understanding of a given area is enough.

  2. 2

    Frame data around a question, not a target

    When you pull stats, do it to answer 'I wonder what's going on' or spot patterns, not to declare a metric you must increase.

    Pro tip Ask things like 'is there a pattern here, are these companies using this thing more,' rather than 'we need this number up X%.'

    Watch out Setting a metric goal per feature optimizes a number rather than solving the customer's actual problem.

  3. 3

    Decide by informed intuition and own the outcome

    With the team informed of customer reality, use judgment to decide, define success as customers agreeing the problem is solved, and accept you'll sometimes be wrong and fix it.

    Pro tip Success is customers enjoying the solution or agreeing the problem is solved, not the metric going up.

    Watch out Letting data make the choice for you can override a correct gut call and is often driven by fear of being wrong.

In the wild

Project updates shipped on judgment

Linear built the project-updates feature because they thought it would be really nice, not to move a metric. After using it they judged from experience and feedback that following updates could be more robust (email digests, search, filtering), and iterated on that basis.

The feature shipped and evolved through informed judgment rather than experiment-driven targets.

Contrast with Ramp's velocity model

Karri notes Ramp builds product with velocity, constant shipping, and measuring everything, an almost opposite approach to Linear's.

Both work; the takeaway is to pick one approach and do it fully, matched to the founder and domain.

Common mistakes

Setting a metric goal for every feature

Optimizing a specific number per feature (like engagement) drives you to move the metric instead of genuinely solving the customer's problem.

Using data to avoid owning the decision

Leaning on data out of fear can override a correct intuition and outsources a judgment call that should be yours; you must accept occasional mistakes and fix them.

Is it for you?

Best for

High-caliber, product-minded teams in a retention/trust business who can invest in deep customer immersion

Not ideal for

Large-scale consumer apps where optimizing a single engagement metric genuinely is the goal, or teams lacking direct customer access

From the transcript

we like to like talk about this inite like this like a mixture of like magic and science

45:30

we do talk to users a lot and like the whole whole like the uh any project we start with we do some like level…

45:30

usually we have some kind of question we want to answer it's like I wonder what what what is going on and then we look…

47:00

sometimes people use data a lot or too much because they just are they're worrying or they're afraid that will I make a wrong choice

49:00

the data didn't make that choice for us

49:30

From the episode

Inside Linear: Building with taste, craft, and focus

Karri Saarinen (co-founder, designer, CEO)