The Forest City Law | Why More Building Can Make a System Worse

More is not always more.

A system can have more buildings, more worksheets, more software, more lanes, more meetings, more data, more lessons, more staff, more rules and more features — and still become harder to use.

That sounds paradoxical only if we assume that the thing being added is the thing currently limiting performance.

The Forest City Law is a simple diagnostic principle: before adding more capacity, find out whether the present system can absorb, connect and use the capacity it already has.

The name is a metaphor used on eduKateSG for a recurring systems problem. It is not a judgement about any specific real-world development called Forest City, nor a claim that building is bad. Sometimes the correct answer is absolutely to build more. The law exists to make one distinction first:

Is the system failing because it lacks capacity — or because existing capacity cannot reach the place where it becomes useful?

Quick Read

When performance is weak, “add more” is an intuitive response. More practice. More content. More infrastructure. More staff. More technology. Sometimes that is exactly right.

But if the actual bottleneck is somewhere else, more can create congestion, maintenance burden, confusion or unused inventory.

The better sequence is:

Define the real demand.
Identify the receiver — who or what must ultimately benefit.
Inventory what already exists.
Find the limiting point.
Repair connection, routing, feedback or utilisation if that is the constraint.
Test under real conditions.
Then add the smallest justified increment if a genuine capacity gap remains.

This is not austerity. It is sequencing.

A Building Is Not the Same Thing as a Function

Suppose a city has a large number of apartments but too few residents, weak transport connections, inadequate services and little employment nearby.

It unquestionably has housing units.

But whether those units function as a healthy city depends on much more than the number of structures.

People need reasons and means to live there. Water, power, transport, schools, shops, employment, healthcare, maintenance, finance, safety and social life have to meet the built space. A building can exist physically while remaining weakly connected to the demand that would make it useful.

The same mistake appears in many domains because visible inventory is easy to count.

A school counts worksheets.
A company counts software licences.
A government counts kilometres of infrastructure.
A learner counts chapters completed.
A household counts hours spent studying.
A service counts forms processed.

Those are real quantities. They are simply not the same as useful outcomes.

Inventory tells us what is present. Function tells us what the present things can actually do.

The Bottleneck Changes the Value of Everything Upstream

Imagine an airport with excellent check-in capacity but a security lane that can process only a fraction of the passengers arriving each hour.

Adding more check-in kiosks may make check-in faster.

It may also move passengers into the security queue faster.

The local metric improves. The whole journey does not.

Now imagine the security lane is expanded but baggage handling cannot cope. The bottleneck moves again.

This is why system improvement is not simply a competition to maximise every component. A component’s value depends partly on where the constraint sits.

Education behaves the same way.

A student may have weak performance because:

the concept was never understood;
the prerequisite is missing;
knowledge is understood but not retrievable;
the learner cannot recognise when the knowledge applies;
the representation is confusing;
working-memory load is too high;
feedback is not being converted into correction;
the learner can perform only when the question resembles the practice example;
time pressure collapses an otherwise sound method.

All nine can produce a low mark.

Only some of them are repaired by giving the student more of the same practice.

The Law Is Not “Do Less Practice”

This distinction is important because practice is essential to learning.

Research-based guidance from organisations such as the Australian Education Research Organisation, the U.S. What Works Clearinghouse, the American Psychological Association and the Education Endowment Foundation consistently supports well-designed practice, retrieval, worked examples, feedback, spaced review and gradual movement towards independence.

The Forest City Law does not argue against any of that.

It asks whether the practice being added addresses the limiting problem.

A novice may need more worked examples because trying to discover every method independently overloads attention. Later, the same learner needs guidance to fade so that the method can be selected without a teacher labelling the path. Still later, mixed and varied practice is needed so that the student learns to discriminate between methods rather than merely repeat one method many times.

“More practice” is therefore too vague.

More of what, at which stage, for which learner, against which bottleneck?

A Mathematics Example: When More Topical Questions Stop Helping

Consider a Secondary Mathematics student who performs well on a worksheet titled “Quadratic Equations”.

The title itself provides information.

The student knows before reading Question 1 that a quadratic method will probably be useful. Several similar questions follow. The representation is familiar. The technique becomes fluent.

This is useful practice.

Now place the same concept inside a full examination. The question describes a rectangle, a changing dimension and an area condition. Nothing says “quadratic equation”.

The learner must construct the equation before solving it.

If performance collapses here, another fifty labelled quadratic questions may increase fluency without repairing recognition and modelling.

The bottleneck has moved.

A more useful next practice set might mix linear equations, quadratics, simultaneous equations and non-equation problems. The learner must decide which structure is present. The practice now targets selection.

Later, a good teacher may ask the student to explain why one representation is useful, compare two methods, predict which mistake is likely and check the answer against the original situation.

The quantity of practice can remain high.

The function of the practice has changed.

English Vocabulary: A Large Word List Can Be an Empty City

Vocabulary makes the metaphor especially visible.

A student can memorise hundreds of definitions.

That is genuine knowledge.

But usable vocabulary also requires pronunciation, grammatical behaviour, collocation, register, connotation, semantic neighbours, context and retrieval under real writing or speaking conditions.

A word may be “installed” in memory yet unavailable during composition.

Another may be retrieved but used with the wrong preposition.

Another may be technically synonymous with a common word but emotionally or socially inappropriate in the sentence.

A bigger list is useful only if the learner is also building routes into usage.

That can mean reading, sentence construction, contrast, retrieval, rewriting, speaking, feedback and repeated encounter across different contexts.

Vocabulary size matters. Vocabulary connectivity matters too.

Science: Facts Are Capacity; Explanatory Models Make Them Work Together

A Science learner may know many correct facts and still struggle with unfamiliar application questions.

Why?

Because Science questions often require the learner to organise facts into a causal explanation.

Knowing that evaporation occurs, that temperature affects particle energy and that moving air removes water vapour from a surface is useful. Explaining why wet clothes dry faster under particular conditions requires those facts to connect into a mechanism.

If the limiting point is causal modelling, adding isolated fact-recall items alone may not repair the problem.

Practice must eventually ask the learner to predict, explain, compare, interpret evidence and identify where a model stops applying.

Five Scarce Resources That “More” Can Accidentally Consume

Additional capacity is never completely free. Even when money is abundant, it can consume other scarce resources.

Attention. More dashboards, alerts, textbooks, features or study resources can make it harder to identify what deserves focus.

Time. Every new process, lesson, feature or facility has to be learned, maintained or navigated.

Integration capacity. New components create interfaces with existing ones. The cost of integration can grow faster than the number of components.

Maintenance capacity. New assets do not disappear after launch. They need inspection, repair, replacement, staffing, security, documentation and eventual renewal.

Feedback capacity. More activity produces more signals. If nobody can interpret them, the system becomes noisier without becoming wiser.

This does not mean expansion should stop.

It means expansion has an operating cost beyond construction.

For the time dimension — why successful systems eventually need maintenance, replacement and redesign — see What Makes Singapore Work | The Renewal Habit.

More Connections Can Also Make a System Worse

If disconnected systems are a problem, it is tempting to connect everything.

That creates another trap.

Every connection can carry useful information.

It can also carry error, delay, noise, privacy risk, cyber risk and cascading failure.

A well-designed system therefore does not maximise the number of links. It designs the links that matter, preserves appropriate boundaries and makes responsibility clear.

The specialist deep dive is What Makes Singapore Work | The Handoff: identity, current state, timing, meaning, accessibility, responsibility and the return signal all have to survive the boundary.

The Receiver Decides Whether Capacity Is Real

A ramp is useful capacity only if it produces a usable route.

A digital service is useful capacity only if the intended people can complete the task.

A tutoring resource is useful capacity only if the learner can absorb and apply it.

A new hospital bed is useful capacity only when staffing, equipment, support services and patient flow allow it to function safely.

This sounds obvious. It is easy to forget because installed capacity is visible and receiver experience is distributed.

Installed does not automatically mean available. Available does not automatically mean accessible. Accessible does not automatically mean usable.

When Building More Is Exactly the Right Answer

A diagnostic rule is only useful if it can also tell us when not to apply it.

There are genuine capacity shortages.

A class may simply be too large for the available teachers and space.
A water system may genuinely need more treatment or storage capacity as demand rises.
A hospital may need more beds or specialist staff.
A growing neighbourhood may need another school, bus route or clinic.
A learner may need substantially more retrieval practice because knowledge is still not fluent.
A company may need more computing capacity because demand has genuinely saturated the existing system.
An accessible route may not exist at all and must be physically built.

In these cases, refusing to build because “utilisation should be improved first” becomes a false economy.

The law is not a command to squeeze every asset to 100% utilisation.

That would be fragile.

Unused Capacity Can Be Valuable

Some spare capacity is not waste.

It is a buffer.

A hospital needs surge capacity for unusual demand. A transport system needs recovery options. A water system needs resilience against weather variability and disruption. A student needs cognitive room to solve a novel problem rather than operating permanently at overload.

This distinction is essential:

Idle because nobody can use it is different from spare because the system deliberately preserves resilience.

The first may indicate poor fit.

The second may be excellent design.

High utilisation sounds efficient, but a system operating at its absolute limit has little room for maintenance, variation, mistakes or shocks.

The Difference Between a Bottleneck and a Buffer

A bottleneck restricts the important flow.

A buffer protects the important flow.

Both may look like “unused capacity” or “waiting” in a simple utilisation report.

That is why context matters.

If an emergency generator spends most of its life unused, that is not necessarily poor utilisation. Its job is not to maximise daily operating hours. Its job is to preserve power when the normal source fails.

If a student spends time checking an answer rather than producing more answers, that time may look less productive by volume and more productive by error prevention.

Good systems are designed around purpose, not one universal utilisation metric.

A Parent’s Version of the Forest City Diagnostic

When a child is struggling, parents understandably want to act.

More tuition hours. Another assessment book. More worksheets. More notes. Another online platform.

Any of these might help.

Before adding them, ask five questions.

What exactly is the child unable to do? “Weak in Maths” is too broad.
Can the child explain the underlying idea?
Can the child retrieve it without a cue?
Can the child recognise it when the question looks different?
Does feedback change the next attempt?

If the child cannot explain the idea, reteaching may be needed.

If understanding is good but retrieval is slow, spaced retrieval practice may be needed.

If topical practice is strong but mixed papers are weak, interleaving and discrimination may be needed.

If the method is sound but execution is error-prone, fluency and checking routines may be needed.

If the student understands everything at home and collapses under examination timing, practice under realistic constraints may be needed.

The answer can still be “more”.

But now it is more of the thing the learner actually needs.

A Teacher’s Version: Do Not Build Past the Weakest Transition

Teachers face the same problem at class scale.

A scheme of work may keep moving because the calendar says the next chapter is due. Yet a prerequisite fracture can make later teaching increasingly expensive.

If algebraic manipulation is unstable, Additional Mathematics can become a constant struggle even when the new concept itself is explained clearly. If reading comprehension is weak, Science questions may look like content failures. If sentence control is unstable, adding sophisticated vocabulary may produce awkward rather than stronger writing.

The answer is not to stop the curriculum indefinitely.

It is to locate the high-leverage repair: the prerequisite or handoff whose improvement unlocks many later tasks.

A Business Version: Feature Count Is Not Product Value

Software products often demonstrate the same law.

Users struggle, so another feature is added.

The feature needs another menu item.

The menu becomes harder to navigate.

Documentation grows.

Support burden grows.

More users become confused.

The team adds onboarding to explain the feature complexity created by earlier additions.

The original user problem may have been a confusing workflow rather than missing capability.

Sometimes the best product work is additive.

Sometimes it is connective.

Sometimes it is subtractive.

A Public-System Version: Local Improvement Can Move the Queue

Public infrastructure is especially vulnerable to local optimisation because each component is large enough to deserve its own performance indicators.

But people experience journeys.

A faster rail line can still produce a poor commute if the interchange becomes crowded or the final walking route is inaccessible. More road capacity can improve one segment while moving congestion elsewhere. More service counters can process people faster while the approval stage remains the constraint.

The broader Singapore systems synthesis is explored in What Makes Singapore Work? The Return Path. The common idea is simple: local capability matters only when it survives the whole route to the receiver and reality can report back what happened.

The Nine-Step Forest City Sequence

When a system feels weak, use this sequence before deciding that the answer is expansion.

1. Define the demand

What must become possible? A better mark? Shorter waiting? Safer access? More water? Higher reliability? Faster recovery? Better writing? Do not begin with the proposed solution.

2. Identify the receiver

Who or what ultimately needs to use the capability? The average commuter, a wheelchair user, a stressed parent, an examination candidate, a hospital patient and a logistics operator can expose different versions of the same system.

3. Inventory existing capacity

What resources, knowledge, routes, staff, time, tools and infrastructure already exist?

4. Measure utilisation carefully

What is unused because it is inaccessible or badly matched? What is deliberately spare because it provides resilience? Do not collapse those two states.

5. Find the limiting point

Where does the important flow slow, fail or become lossy?

6. Repair the smallest high-leverage constraint

That may be teaching, connection, representation, maintenance, staffing, accessibility, routing, timing or feedback.

7. Test under real conditions

Topical success, laboratory success and demonstration success can all disappear when real variation arrives. Test the system where it actually has to work.

8. Observe the return

Did the receiver improve? Did the queue move? Did the student transfer? Did reliability change? Did a new bottleneck appear?

9. Build the justified increment

If genuine capacity is still insufficient, expand it. Now the addition has a defined job.

What the Forest City Law Gets Wrong If Applied Badly

Every compact rule can become dangerous when turned into a slogan.

Bad use 1: “Never build more.” Wrong. Genuine capacity shortages require capacity.

Bad use 2: “Maximise utilisation first.” Wrong. Resilient systems need buffers, maintenance windows and spare routes.

Bad use 3: “The bottleneck is always obvious.” Wrong. The visible queue may be caused by an upstream or downstream constraint.

Bad use 4: “One bottleneck explains everything.” Wrong. Complex systems can have multiple interacting constraints, especially under different operating conditions.

Bad use 5: “Efficiency is the only goal.” Wrong. Safety, fairness, resilience, accessibility, learning and long-term adaptability may justify apparent inefficiency.

Bad use 6: “If the receiver struggles, the system is wrong.” Not automatically. Some tasks are genuinely difficult. Good systems reduce unnecessary friction without pretending every legitimate requirement can disappear.

How to Know Whether the Intervention Worked

Measurement should match the job.

Throughput: can more useful work reach completion?
Waiting: has time at the limiting boundary fallen?
Rework: are fewer tasks repeated because of errors or missing information?
Transfer: can learning survive a new context?
Reliability: does performance remain acceptable over time?
Recovery: can the system preserve its important function when the normal route fails?
Accessibility: do improvements reach receivers outside the assumed average?
Maintenance burden: did the solution create a larger future support problem than expected?

Most importantly, do not let a proxy replace the underlying reality.

Hours studied are not learning.

Questions completed are not transfer.

Features shipped are not user value.

Buildings completed are not a functioning neighbourhood.

The proxy can be useful. It simply has to stay attached to the thing it is supposed to represent.

Reader Experiment: Find One Place Where You Are Adding Past the Bottleneck

Choose one system you know well.

Your child’s study routine. Your office workflow. Your household. A school club. A digital product. A daily commute.

Write down everything being added because performance feels inadequate.

Then ask:

What is the desired outcome?
Where does the flow actually stop?
What existing capacity is unused?
Why is it unused?
What burden does each new addition create?
Would removing, reconnecting or redesigning something outperform another addition?
If more capacity is genuinely needed, what is the smallest addition with a clearly defined job?

The exercise is not about becoming minimalist.

It is about making additions earn their place.

FAQ: The Forest City Law

Is the Forest City Law against building?

No. It is against assuming that construction automatically solves the problem. If the real constraint is insufficient capacity, building is the correct response.

Is it a claim about the real Forest City development in Malaysia?

No. On this page, “Forest City” is a teaching metaphor for installed capacity that does not automatically become functioning flow. Readers should not treat the metaphor as an assessment of any particular real-world project.

Does the law say students should do fewer questions?

No. Practice is essential. The question is whether the practice targets understanding, retrieval, selection, fluency, transfer or another genuine need. A student may need more questions — but not necessarily more of the same question type.

What if I cannot identify the bottleneck?

Use small tests. Change one plausible constraint and observe the effect. Ask the receiver. Compare successful and failed cases. Look for where waiting, errors or repeated help appear. Uncertainty is a reason to diagnose, not a reason to guess with a large intervention.

Can a system have several bottlenecks?

Yes. Different constraints can dominate at different times or demand levels. Improving one can reveal another. That is normal.

How is this different from The Handoff?

The Forest City Law asks whether more capacity is actually needed and where the limiting point sits. The Handoff goes deeper into what must survive when capability crosses a boundary between systems.

How is this different from The Renewal Habit?

The Forest City Law is mainly about where to add. The Renewal Habit is mainly about when existing capacity must be maintained, replaced or redesigned.

The Final Distinction

Systems fail from scarcity.

They also fail from disconnection.

They fail from poor routing.

They fail from overload.

They fail from stale assumptions.

They fail because the receiver changed.

They fail because nobody noticed that the local success did not survive the whole journey.

“Build more” solves one family of failures.

It can worsen another.

Do not build past the bottleneck. Make the existing system work as a system — then build the capacity reality actually asks for.