Singapore School Mathematics Operating Manual · Chapter 41
Many mathematical systems contain several constraints, but only one or two actually control the outcome.
A factory can have ten machines, yet one slow stage sets the maximum throughput. A geometric construction can satisfy many inequalities, yet one boundary determines feasibility. A revision plan can contain many topics, yet one weak prerequisite blocks progress. A transport network can contain many links, yet one narrow section controls capacity.
This chapter develops the mathematics of bottlenecks and limiting constraints. The central question is: which restriction is actually controlling the result?
1. A bottleneck is the active limiting stage
If three processing stages can handle 100, 80 and 120 units per hour, the full line cannot exceed 80 units per hour without buffering or redesign.
The 80-unit stage is the bottleneck.
2. The minimum can control a whole system
For a serial pipeline with capacities c₁,c₂,…,cₙ, a simple throughput ceiling is min(c₁,c₂,…,cₙ).
The slowest stage limits the chain.
3. Not every constraint is active
Suppose x≤100, x≤80 and x≤120.
The effective upper bound is x≤80.
The other constraints are true but inactive for this decision.
4. Redundant constraints do not change the feasible set
x≤100 adds nothing once x≤80 is already enforced.
Recognising redundancy prevents wasted attention.
5. Active constraints often have zero slack
If the optimum occurs at x=80 under x≤80, slack is zero.
The constraint is binding.
The Decision Margins and Slack chapter develops this quantitatively.
6. Improving a non-bottleneck may not improve the system
If stage capacities are 100,80,120 and the 120 stage is increased to 200, system throughput remains 80.
Local improvement did not change the limiting constraint.
7. Improving the bottleneck can move the bottleneck elsewhere
If capacities become 100,110,120, the new limit is 100.
The bottleneck has migrated.
Optimisation is therefore iterative.
8. A chain is limited by its weakest required link
If every stage is necessary, one failing stage can prevent completion.
This is different from systems with parallel alternatives, where redundancy may bypass failure.
9. Series and parallel structures behave differently
In series, all required stages must succeed.
In parallel, multiple routes may provide alternatives.
System architecture determines whether a local weakness is globally limiting.
10. Geometry contains limiting inequalities
For a non-degenerate triangle, the longest side must be less than the sum of the other two.
Among several side constraints, the one nearest violation may be the limiting geometric condition.
11. Integer problems have discrete bottlenecks
If a box holds at most 12 items and 97 items must be packed, minimum box count is ceiling(97/12)=9.
The capacity per box limits the decision.
12. Time can be the bottleneck
If a paper has 120 minutes and a candidate spends too long on one section, time rather than mathematical ability can become the limiting resource.
The same Mathematics applies to scheduling: finite resource, competing demands, active constraint.
13. Budget can be the bottleneck
If a project satisfies space and time constraints but exceeds budget, budget controls feasibility.
The relevant bottleneck is whichever active constraint first blocks the candidate.
14. Data quality can bottleneck a model
More sophisticated calculation cannot recover information absent from the data.
If measurement resolution is coarse, it may be the limiting factor in answer precision.
15. Model quality can bottleneck prediction
If model assumptions are weak, improving numerical precision may not improve the real conclusion.
The bottleneck sits in structure, not arithmetic.
16. A proof can have a logical bottleneck
If one unproved implication is essential, the entire proof remains incomplete even if every other step is correct.
The weakest unsupported link controls validity.
17. Learning has prerequisite bottlenecks
A student may practise quadratic questions repeatedly but continue to fail because algebraic factorisation is weak.
The visible topic is not always the limiting skill.
18. Dependency graphs reveal bottlenecks
A node used by many later steps has high leverage.
If that node is slow, uncertain or error-prone, it can limit the whole workflow.
19. Bottlenecks can be structural or temporary
A structural bottleneck comes from system design.
A temporary bottleneck may arise from one unusually difficult instance, limited time or temporary capacity loss.
20. The narrowest margin deserves attention
If three constraints have slack 20, 7 and 1, the one with slack 1 is closest to failure.
It may become the next bottleneck under small perturbation.
21. Sensitivity can change bottleneck identity
A small parameter change may make a previously inactive constraint active.
This is why bottleneck analysis should be repeated after major changes.
22. Objective choice changes the bottleneck
Optimising cost may expose a different limiting factor from optimising time.
There is no universal bottleneck independent of the system objective.
23. A bottleneck can be hidden by averages
Average capacity may look high while one stage remains low.
System throughput depends on the structure, not average capacity alone.
24. Bottleneck analysis should come before blanket improvement
Improving everything equally can waste resources.
First identify the active limiter, then ask whether changing it changes the global outcome.
25. A practical bottleneck audit
Ask:
What output is being limited? Which constraints are active? Which has the least slack? Which stage has the lowest required capacity? If I improve this component, does the total system improve? If so, what becomes the next bottleneck?
26. Independent practice
1. A three-stage system has capacities 50,70,40 units/hour. Find throughput ceiling.
2. Improve the 70 stage to 100. Does throughput change?
3. Improve the 40 stage to 60. What is the new bottleneck?
4. Under x≤12, x≤9 and x≤15, which constraint controls?
5. Explain why an inactive constraint may still matter after parameters change.
6. Give one example of a data-quality bottleneck.
7. Why can average capacity be misleading?
8. What is the main question after improving a bottleneck?
27. Worked answers
1. min(50,70,40)=40 units/hour.
2. No. The 40 stage still limits throughput.
3. Capacities become 50,70,60, so the 50 stage is now the bottleneck.
4. x≤9.
5. The active constraint can change when other limits are relaxed or parameters move.
6. Measurements recorded only to the nearest 10 units can limit the precision of any later model.
7. A single low required stage can control throughput even when the average is high.
8. “What becomes the next limiting constraint?”
28. Continue through Batch 11
Use Reversibility and Optionality when changing a bottleneck may close future routes, Coupling and Interaction Effects when bottlenecks interact, and Local and Global Optimisation when improving one component may harm the whole system.
Return to the BTT Mathematics Hub for Batch 11.