Contents
Systems & Complexity
Section 1
The Core Idea
Optimization is improving a system toward a defined objective — making a chosen variable as good as possible given constraints. The objective might be cost, speed, throughput, or quality; the constraints might be time, capital, or physical limits. The discipline is explicit: name the objective, identify the constraints, then improve until the next improvement costs more than it returns. Without a clear objective, optimization drifts into local tweaks that don't move the needle. Without constraints, the "optimal" solution is often infinite scale or zero cost — useless.
In practice, most systems have one binding constraint at a time. The Theory of Constraints frames this: find the bottleneck, optimise there, then the bottleneck moves. Optimising a non-bottleneck is waste. A factory that speeds up a station downstream of the slowest station only builds inventory. A product team that optimises a feature few users touch burns cycles. The strategic question is always: what is the current constraint, and does this change relax it?
Optimization has a dark side: premature optimisation. Donald Knuth's line — "premature optimisation is the root of all evil" — applies when you optimise before the problem is stable, the objective is clear, or the bottleneck is known. Early-stage startups that tune database queries before product-market fit are optimising the wrong thing. The right sequence: validate the system, then optimise the constraint. Optimization also risks overfitting to the current environment. A process optimised for last quarter's demand can break when demand shifts. The best optimisers leave slack at the constraint and re-check the objective when the world changes.
Section 2
How to See It
Optimization shows up wherever someone is deliberately improving a metric against limits. Look for: a stated objective, measurement of that objective, and changes designed to move the number. When the same system keeps getting tweaked for the same goal, optimization is at work. When tweaks stop yielding gains, you're at a local optimum or the constraint has shifted.
Business
You're seeing Optimization when a fulfilment centre reconfigures pick paths to cut average time per order. The objective is throughput; the constraint is labour and space. The team measures picks per hour, identifies the slowest segment, and redesigns it. Gains appear until the next bottleneck — packing or shipping — binds. Then optimization moves there.
Technology
You're seeing Optimization when engineers profile an app and discover 80% of latency is in one function. They rewrite it, cut latency in half, then re-profile. The next hotspot becomes the target. The cycle repeats. The discipline is measuring first, optimising the bottleneck, then measuring again. Guessing where to optimise without data is premature optimisation.
Investing
You're seeing Optimization when a public company cuts costs to hit margin targets. The objective is earnings; the constraint is revenue and competitive position. Aggressive cost optimisation can improve margins short-term while damaging growth or quality — the constraint shifts from cost to demand. The question: is the current optimisation relaxing the right constraint for the thesis?
Markets
You're seeing Optimization when a trader builds execution algorithms to minimise market impact. The objective is execution quality (price vs benchmark); the constraint is liquidity and time. Tweaking order-slicing and venue selection is optimization. So is the decision to stop optimising when marginal improvement is smaller than the cost of more complexity.
Section 3
How to Use It
Decision filter
"Before optimising, name the objective and find the constraint. Optimise at the constraint. If you're not sure where the constraint is, measure. If the objective is fuzzy, clarify it first. Avoid optimising anything else until the bottleneck moves."
As a founder
Optimise the constraint that limits growth or profit. Usually that's distribution, retention, or unit economics — not the nth iteration of a feature. Measure the funnel, find where the biggest drop or cost is, and improve that. The mistake: optimising engineering velocity when the bottleneck is sales. The second mistake: optimising for a metric that doesn't map to the real objective (vanity metrics). Align the team on one primary objective and one constraint at a time.
As an investor
Check whether management is optimising the right variable. Companies that optimise for earnings when the thesis is growth are misaligned. Companies that optimise for scale when the constraint is retention burn capital. The due-diligence question: what is the binding constraint today, and is this team optimising there?
As a decision-maker
Use optimization when the objective is clear and the constraint is known. When either is unclear, invest in discovery first. Defer optimisation until you're confident you're not optimising a sub-system that doesn't limit the whole. Revisit the objective when the environment changes; the old optimum may be obsolete.
Common misapplication: Optimising a non-constraint. Resources go into improving something that isn't limiting the outcome. Throughput doesn't rise; cost or complexity does. The fix: identify the bottleneck with data, not intuition.
Second misapplication: Optimising too early. The system is still changing; the objective might shift. Locking in an optimised process before product-market fit or before the bottleneck is stable wastes effort and can lock in the wrong design.
Section 4
The Mechanism
Section 5
Founders & Leaders in Action
[Jeff Bezos](/people/jeff-bezos)Founder & CEO, Amazon, 1994–2021
Bezos drove optimization at scale by tying it to customer outcomes and constraints. Amazon's fulfilment and logistics are optimised for delivery speed and cost — the objective is customer experience and margin; the constraint has shifted over time from warehouse layout to last-mile capacity to inventory placement. The company runs experiments to find the next bottleneck and optimises there. Bezos's "two-pizza teams" and single-threaded ownership ensure optimisation efforts are aligned to one constraint per team rather than diffused across many.
[Andy Grove](/people/andy-grove)CEO, Intel, 1987–1998
Grove applied constraint-based thinking to strategy. In Only the Paranoid Survive, he framed strategic inflection points as moments when the old constraint (e.g. manufacturing capability) is replaced by a new one (e.g. design or software). Optimization had to follow the constraint: Intel optimised for manufacturing excellence when that was the bottleneck; when the constraint shifted to ecosystem and software, Grove redirected the company. The lesson: optimise the current constraint; when the world shifts, re-identify it.
Section 6
Visual Explanation
Optimization: one objective, one binding constraint. Improve at the constraint until it moves; then the next constraint binds. Optimising elsewhere does not improve the whole.
Section 7
Connected Models
Optimization sits with constraint-based thinking, efficiency, and feedback. The models below either define the constraint (Theory of Constraints, Bottlenecks), warn against misuse (Premature Optimisation), or refine the objective (Efficiency vs Effectiveness, [Feedback Loops](/mental-models/feedback-loops)).
Reinforces
Theory of Constraints
The Theory of Constraints says the system has one binding constraint; optimisation elsewhere is waste. Optimization is the act of improving at that constraint. The two are inseparable: TOC identifies where to optimise; optimization is the execution.
Reinforces
[Bottlenecks](/mental-models/bottlenecks)
Bottlenecks are the constraints that limit throughput. Optimization targets the bottleneck first. When you relieve it, the bottleneck moves. The discipline is repeatedly finding and optimising the current bottleneck.
Tension
Premature Optimisation
Premature optimisation is optimising before the problem is stable or the constraint is known. The tension: optimization is powerful when applied at the right place and time; applied too early or in the wrong place, it consumes resources and can lock in bad design.
Leads-to
[Marginal Gains](/mental-models/marginal-gains)
Marginal gains are small optimisations that compound. The mindset is the same: define the objective, find the constraint, improve. When the big levers are exhausted, marginal gains at the constraint still add up.
Leads-to
Efficiency vs Effectiveness
Efficiency is doing things right; effectiveness is doing the right things. Optimization can maximise efficiency at the wrong objective. The check: are we optimising for effectiveness (the right outcome) or only efficiency (cost or speed of the current process)?
Reinforces
Feedback Loops
Feedback loops measure the objective and guide the next optimisation. Without measurement, optimization is guesswork. Close the loop: measure, optimise at the constraint, measure again.
Section 8
One Key Quote
— Eliyahu Goldratt, The Goal"The bottleneck is the constraint. Everything else is secondary. Improve the bottleneck and the whole system improves."
Goldratt's formulation forces a single focus: one constraint at a time. Optimization without that focus scatters effort. The practitioner's job: name the objective, find the bottleneck, optimise there, then repeat when the constraint moves.
Section 9
Analyst's Take
Faster Than Normal — Editorial View
Optimization only pays when the objective and constraint are right. Most teams optimise something — but often the wrong thing. The discipline is asking: what is the one metric that, if improved, would move the outcome we care about? And what is currently limiting that metric? If you can't answer both, you're not ready to optimise; you're ready to measure and clarify.
Premature optimisation is still the default failure mode. Early-stage companies optimise infrastructure, code paths, and org structure before product-market fit. The bottleneck is usually distribution or retention, not performance. The fix: write down the objective and the constraint. If the constraint is "we don't have enough users," optimising server cost is premature.
The constraint moves. Today's bottleneck is not tomorrow's. Optimise the current one, then re-profile. Teams that keep optimising the same subsystem after the constraint has shifted are wasting capacity. Build a habit of re-identifying the constraint when results plateau.
One objective at a time. Multi-objective optimization is possible in theory but in practice leads to confusion and trade-off debates. Pick the primary objective for this cycle. Optimise for that. Revisit when the cycle ends or the environment shifts.
Section 10
Summary
Optimization is improving a system toward a defined objective given constraints. Name the objective, find the binding constraint, and optimise there. Optimising non-constraints doesn't improve the whole. Avoid premature optimisation until the problem is stable and the bottleneck is known. When gains plateau, re-identify the constraint — it has moved.
Section 11
Further Reading
01
Book
The foundational Theory of Constraints text. The bottleneck is the constraint; optimise there. Narrative format; the principle is clear and actionable.
02
Book
Grove on strategic inflection points: when the constraint shifts, optimisation must follow. Essential for linking optimization to strategy.
03
Book
Applies constraint-based thinking to IT and DevOps. Bottlenecks, flow, and optimisation in tech delivery.
04
Book
Standard treatment of optimisation in finance: NPV, capital structure, and the objective function in corporate decisions.
05
Article
The source of "premature optimisation is the root of all evil." Context: optimise when you have a clear bottleneck and measurement; not before.
Why this matters next
mental modelsTheory of Constraints
Optimization applied the Theory of Constraints mental model
mental modelsNarrative
Optimization applied the Narrative mental model
mental modelsTrade-offs
Optimization applied the Trade-offs mental model
mental modelsScale
Optimization applied the Scale mental model
mental modelsIntuition
Optimization applied the Intuition mental model
mental modelsQuality
Optimization applied the Quality mental model
Frequently asked questions
What is Optimization?
Optimization is a mental model used for better thinking and decision-making.
How do you apply Optimization?
To apply Optimization, identify situations where this framework is relevant, then use it as a lens to evaluate your options and decisions. The model is most useful when combined with other complementary mental models.
What category does Optimization fall under?
Optimization falls under the Systems & Complexity category of mental models. Other models in this category can be found on the Systems & Complexity hub page.
Why is Optimization important?
Optimization is important because it provides a structured way to think about problems that would otherwise be approached with intuition alone. Understanding this model helps you avoid common reasoning errors and make better decisions.
Continue exploring
LEANEMFLGLL(
Mental model
Lollapalooza Effects
When multiple cognitive biases or forces act in the same direction simultaneousl
Mental model
Antifragility
Beyond resilience — some systems actually gain from disorder, volatility, and st
Mental model
Emergence
Complex system-level properties that arise from simple interactions between indi
Mental model
Feedback Loops
Circular causal chains where the output of a system feeds back as input — either
Mental model
Gall's Law
A complex system that works is invariably found to have evolved from a simple sy
Mental model
Leverage (Systems)
Places within a complex system where a small shift produces large changes — Mead
More like this, in your inbox
I send a newsletter every week — free, no spam, unsubscribe anytime.