Pick a goal.
Improve a metric.
Make the system better.
But real systems don’t work like that.
Because when you optimize one thing,
you almost always change something else.
Optimization always narrows the system
Every optimization starts with a choice.
What matters?
- engagement
- speed
- conversions
- retention
Once that choice is made, everything else becomes secondary.
Optimization doesn’t improve the whole system.
It improves one dimension.
And that changes behavior.
Metrics don’t describe reality — they reshape it
The moment a metric becomes a target, behavior shifts toward it.
People — and systems — start optimizing for the metric itself, not for the underlying goal.
Which leads to:
- gaming
- distortion
- unintended outcomes
Incentives and optimization are the same system
Optimization doesn’t exist in isolation.
It’s built on incentives.
As described in Why Product Incentives Shape User Behavior More Than Features:
users follow what gets rewarded.
Optimization systems define those rewards.
Which means:
they don’t just improve behavior
they reshape it
Algorithms optimize — not for meaning, but for signals
Modern systems don’t optimize abstract goals.
They optimize measurable signals.
- clicks
- watch time
- interactions
But signals are not the same as value.
So the system learns:
what increases the metric,
not what improves the outcome.
This is exactly how ranking systems evolve, as shown in Algorithms Don’t Just Recommend — They Decide Visibility.
Unintended consequences are not rare — they are expected
In complex systems, unintended outcomes are not edge cases.
They are the default.
Optimization simplifies reality.
And simplified models produce incomplete results.
Systems optimize locally — consequences appear globally
Optimization happens inside the system.
But consequences don’t stay there.
A small adjustment:
- changes incentives
- shifts behavior
- cascades through the system
What starts as a local improvement
becomes a global effect.
This is similar to how information flow is shaped, as described in Why Ranking Systems Quietly Control Information Flow.
Attention becomes the target
Once attention is measurable, it becomes something to optimize.
And once it’s optimized, everything shifts toward capturing it.
This is exactly what’s described in The Economics of Attention in Product Design.
Not because attention is the goal.
But because it’s the easiest thing to measure.
The most important decisions happen before users see them
Users think they are interacting with a system in real time.
But optimization has already shaped the environment.
- what is shown
- what is prioritized
- what is hidden
This follows the same pattern described in The Most Important Decision Is the One You Never Made.
By the time interaction begins,
the outcome is already constrained.
The system looks better — until it doesn’t
From the outside, optimization looks like progress.
Better numbers.
Higher engagement.
Improved metrics.
But underneath, something else might be happening:
- quality declines
- behavior becomes distorted
- long-term outcomes degrade
Because the system is optimizing the wrong thing.
Optimization systems don’t just improve products.
They reshape behavior, incentives, and outcomes.
And because they operate through metrics,
their unintended consequences often look like success —
right up until they don’t.