How to be good at problem solving? The first move is wonderfully simple: do not rush to solve.
First, make sure you know what the problem is.
Did you know that many difficult problems become difficult for two completely different reasons? Some are genuinely hard to solve. Others are badly represented. The facts are scattered, assumptions are hidden, constraints are mixed with preferences, and the solver starts calculating before the shape of the problem is visible.
A gold-standard problem solver does not merely produce answers. A good problem solver builds a reliable path from uncertainty to a result that can be explained, checked and improved.
That is why this skill travels everywhere: Mathematics, Science, English, engineering, coding, business, logistics, education, public policy, household decisions and the daily work of keeping Singapore’s systems moving.
Did You Know? The Representation Can Be Half the Solution
Consider a messy word problem.
On the page it looks like a paragraph. In the solver’s head it may feel like ten unrelated facts.
Then somebody draws a diagram, labels the quantities and writes one relationship.
Suddenly the problem becomes smaller.
Nothing outside the problem changed.
The representation changed.
This is one of the deepest transferable ideas in problem solving: how you represent a problem changes what you can see and therefore what you can do.
The same principle works outside Mathematics. A timeline can clarify history. A process map can expose a bottleneck. A table can reveal a comparison. A decision tree can separate options. A sketch can expose a design constraint. A list of claims and evidence can repair an argument.
The Gold-Standard Problem-Solving Loop
A practical loop is:
- Notice — identify the gap between current state and desired state.
- Define — state the problem precisely.
- Represent — make the structure visible.
- Decompose — split the problem into manageable parts.
- Generate — produce plausible approaches.
- Choose — select an approach using constraints and evidence.
- Execute — carry out the next useful move.
- Verify — check whether the result is actually valid.
- Learn — identify what worked and what failed.
- Transfer — reuse the insight in a new context.
The loop matters more than any single clever trick.
A brilliant idea without verification can still be wrong.
A careful analysis without execution can still be useless.
A successful answer without learning can still leave the next problem untouched.
Step 1: Define the Gap
A problem exists because there is a difference between where you are and where you need to be.
Write both.
- Current state: what is happening now?
- Desired state: what must be true afterwards?
For example:
- Current: the student can solve routine linear equations but fails when fractions appear. Desired: the student can solve and verify linear equations with fractional coefficients.
- Current: a team misses deadlines because work arrives late between stages. Desired: each dependency is visible and critical handoffs occur before agreed cut-offs.
- Current: a presentation contains accurate information but the audience cannot identify the main point. Desired: the audience can state the central message and supporting reasons afterwards.
A precise gap is easier to solve than a vague complaint.
Step 2: Separate Facts, Assumptions, Constraints and Unknowns
This is one of the fastest ways to clean a messy problem.
Facts
Information supported by observation, data, rules or reliable evidence.
Assumptions
Things currently being treated as true but not yet established.
Constraints
Conditions the solution must respect: time, cost, rules, capacity, safety, format or available resources.
Unknowns
Information that would materially change the decision if you had it.
Now ask a powerful question: Which assumption is pretending to be a fact?
A surprising number of problems move immediately when that distinction is made.
Step 3: Restate the Problem in More Than One Way
A first description can trap the solver.
Try three versions:
- What is the problem as currently stated?
- What is the underlying mechanism?
- What outcome are we actually trying to improve?
For example, “students are careless” may become “students are losing marks through repeated copying, sign and unit errors under time pressure”.
That restatement is much more useful.
It suggests different interventions: checking routines, error classification, timed accuracy practice and better working layout.
Words determine what actions become imaginable.
Step 4: Represent the Problem
Choose the representation that exposes the structure.
- Diagram — spatial relationships, geometry, systems.
- Table — comparisons, categories, changing values.
- Equation — quantitative relationships.
- Timeline — sequence, history, causation.
- Flowchart — processes and decision paths.
- Dependency map — tasks that unlock or block other tasks.
- Argument map — claims, evidence, assumptions and counterarguments.
- Graph — trends and relationships between variables.
- Prototype — when understanding requires building or testing.
Do not worship one representation.
Advanced problem solving often means switching representations until the hidden structure becomes visible.
Step 5: Decompose Without Losing the Whole
Large problems overwhelm working memory.
Break them into smaller questions.
But decomposition has a danger: the parts may be solved beautifully while the whole problem remains unsolved.
So use two levels at once:
- Whole-system question: what must the total solution accomplish?
- Subproblem question: what smaller piece can be solved independently or sequentially?
A useful decomposition asks:
- Which part is known?
- Which part is uncertain?
- Which part blocks everything else?
- Which part can be tested cheaply?
- Which part depends on another part?
- Which part carries the most risk?
eduKate explores this directly in How to Break Complex Problems Into Smaller Problems. The important idea is not merely “make it smaller”. It is “make the dependency structure visible”.
Step 6: Use Domain Knowledge Before Generic Cleverness
Problem solving is not one magical mental muscle that operates identically everywhere.
A person solving a chemistry problem needs chemistry knowledge. A person debugging code needs programming knowledge. A person analysing a historical source needs historical concepts and source-reading skills.
General strategies help, but domain knowledge supplies the pieces.
That means becoming good at problem solving requires two kinds of learning:
- Domain knowledge — facts, concepts, procedures, vocabulary and patterns.
- Problem-solving process — defining, representing, decomposing, testing and verifying.
The two strengthen each other.
Knowing more lets you recognise more patterns. Better problem-solving habits help you organise and test what you know.
Step 7: Learn Heuristics Without Turning Them Into Formulas
A heuristic is a useful problem-solving move that often helps but does not guarantee success.
Examples include:
- draw a diagram;
- work backwards;
- look for a simpler case;
- identify an invariant;
- make a table;
- consider extreme cases;
- search for a pattern;
- introduce a useful variable;
- compare with a known problem;
- split the task into cases.
The danger is memorising heuristics as if every problem announces which one to use.
The real skill is selection.
eduKate’s How Mathematical Heuristics Work explores why a good heuristic is a move in reasoning, not a guaranteed formula.
Step 8: Generate More Than One Plausible Approach
The first solution that appears in your mind has a psychological advantage: it arrived first.
That does not make it best.
For important problems, deliberately generate alternatives.
Ask:
- Can I solve it directly?
- Can I transform it into a known problem?
- Can I test a smaller case?
- Can I reverse the direction?
- Can I approximate first?
- Can I remove a constraint temporarily to understand the core?
- Can I solve a simpler version and then generalise?
Multiple approaches create comparison.
Comparison improves judgment.
Step 9: Choose With Explicit Criteria
Once you have options, define the criteria.
A solution may be:
- accurate but slow;
- fast but fragile;
- cheap but difficult to scale;
- elegant but hard to explain;
- safe but conservative;
- powerful but dependent on uncertain information.
The “best” option depends on the actual problem.
Useful criteria include:
- accuracy;
- time;
- cost;
- risk;
- reversibility;
- robustness;
- explainability;
- resource requirements;
- ethical or regulatory constraints;
- ease of verification.
This prevents a common problem-solving error: optimising one attractive metric while damaging everything else.
Step 10: Prefer Small Informative Tests When Possible
When the cost of being wrong is manageable, a small experiment can be more valuable than a long argument.
A prototype, sample, pilot, trial calculation or partial implementation can reveal hidden constraints.
Ask:
- What is the smallest action that would teach us something important?
- What result would support the hypothesis?
- What result would weaken it?
- Can we reverse the test if it fails?
This turns uncertainty into information.
Of course, not every domain allows experimentation. High-stakes systems, safety-critical work and irreversible decisions require stronger controls. Problem solving includes knowing when experimentation is appropriate.
Step 11: Execute the Next Useful Move
Analysis can become a hiding place.
After enough structure is visible, act.
A useful action is one that either:
- solves part of the problem;
- reduces uncertainty;
- tests a key assumption;
- unlocks another task;
- produces evidence for the next decision.
Do not confuse motion with progress.
Ten low-value actions are not automatically better than one decisive experiment.
Step 12: Verify Independently
Verification is one of the clearest differences between casual problem solving and reliable problem solving.
Do not ask only, “Did I get an answer?”
Ask, “What evidence says this answer survives checking?”
Verification methods include:
- substitute the answer back into the original condition;
- solve using a second method;
- check units and dimensions;
- test boundary cases;
- estimate the expected magnitude;
- compare with known data;
- look for counterexamples;
- ask another person to reproduce the result;
- trace the process from output back to inputs.
In Mathematics, a solution may be checked by substitution.
In writing, check whether evidence actually supports the claim.
In software, test expected and unexpected inputs.
In a plan, examine what happens if a critical assumption fails.
The goal is independent pressure on the answer.
Step 13: Use Sanity Checks
A sanity check asks whether the result makes basic sense.
If a calculation says a school bus holds 3,200 students, something is wrong even before you locate the exact arithmetic error.
Useful sanity checks include:
- order of magnitude;
- direction of change;
- units;
- physical possibility;
- known limits;
- comparison with a simple baseline.
Sanity checks are fast because they use broad constraints instead of repeating every step.
Step 14: Search for Counterexamples
If you think a claim is always true, ask what would make it false.
If you think a solution works, ask where it breaks.
If you think a policy helps, ask who might experience the opposite effect.
Counterexamples are powerful because they attack overconfidence.
A robust solution is one that has survived attempts to break it.
Step 15: Keep an Error Log
An error log converts failure into a reusable dataset.
For each meaningful error, record:
- What was the task?
- What did I do?
- Where was the first wrong move?
- What type of error was it?
- What would have caught it earlier?
- What rule or cue should I remember next time?
Over time, patterns emerge.
Perhaps the student repeatedly misreads “at least”. Perhaps a team repeatedly starts production before the brief is stable. Perhaps a programmer forgets edge cases around empty inputs.
Once the pattern is visible, improvement can target the mechanism.
Step 16: Conduct a Post-Mortem on Good Results Too
People often analyse failures and celebrate successes.
But successful results also deserve analysis.
Ask:
- Why did this work?
- Which part was skill and which part was luck?
- What assumptions happened to hold?
- Which move should become a standard practice?
- Would the method still work on a harder variant?
Otherwise, a lucky success can teach the wrong lesson.
Step 17: Learn Transfer
Transfer means recognising that a structure learned in one place may help in another.
A student learns to decompose a Mathematics problem and later uses the same habit to break an essay task into claim, evidence and explanation.
A programmer learns to isolate a bug and later uses the same method to isolate a broken step in a business process.
A historian learns to separate source evidence from interpretation and later becomes better at evaluating online claims.
Transfer is not automatic.
Make it explicit.
After solving something, ask: Where else does this structure appear?
Problem Solving in Mathematics
Singapore’s Mathematics curriculum places mathematical problem solving at the centre of its framework, supported by concepts, skills, processes, metacognition and attitudes.
That is a useful model because it shows why solving is more than calculation.
A student needs:
- concepts to understand the mathematical objects;
- skills to perform procedures;
- processes to reason and select approaches;
- metacognition to monitor thinking;
- attitudes that support persistence and sense-making.
A mathematically strong solver therefore asks both “What can I do?” and “Why does this move make sense here?”
Explore the wider structure through How Mathematics Works.
Problem Solving in Science
Science adds a special requirement: claims must connect to evidence and models.
A useful science scaffold is:
- What phenomenon are we explaining?
- What evidence do we have?
- What model or concept connects the evidence?
- What alternative explanation is possible?
- What observation would distinguish between them?
This is why scientific problem solving is not simply remembering the correct chapter.
It is learning to use concepts as explanatory tools.
Explore How Science Works.
Problem Solving in English and Writing
Writing is full of problems hiding in plain sight.
What is the task really asking? Which evidence is relevant? What does the audience know? What order makes the argument easiest to follow? Which word expresses the distinction accurately?
A writing problem can be decomposed into:
- task interpretation;
- idea generation;
- selection;
- organisation;
- evidence;
- language;
- revision.
This makes improvement more precise than “write better”.
Problem Solving in Teams
Collaborative problems introduce another layer: no single person may hold all the information.
Now the group must solve the problem and the coordination system at the same time.
That connects directly to How to be Good at Teamwork. A team needs shared understanding, useful roles, communication and a way to integrate different pieces of evidence.
The OECD’s collaborative problem-solving framework similarly emphasised shared understanding, appropriate action and team organisation.
Problem Solving With AI
AI changes the speed at which options can be generated.
It does not remove the need for problem definition and verification.
A strong AI-assisted workflow is:
- define the problem yourself;
- state constraints explicitly;
- ask for multiple approaches;
- request assumptions;
- check important facts against reliable sources;
- test the proposed solution;
- use independent verification for high-stakes claims;
- record what remains uncertain.
AI is especially useful for brainstorming representations, generating test cases, finding alternative explanations and critiquing a draft solution.
It becomes dangerous when fluent output is mistaken for verified truth.
Explore related ideas in How to Solve Problems With Super Intelligence.
How to Measure Improvement in Problem Solving
Do not measure only the number of answers completed.
Watch the quality of the process.
- How quickly can you state the problem precisely?
- Do you identify important assumptions?
- Can you choose a useful representation?
- Do you decompose at meaningful boundaries?
- Can you generate more than one approach?
- Do you select using explicit criteria?
- Do you verify independently?
- Are repeated error types decreasing?
- Can you explain why the solution works?
- Can you transfer the method to a new-looking problem?
The mature solver becomes easier to audit.
The reasoning leaves a trail.
Common Problem-Solving Traps
Solving the Wrong Problem
The answer can be perfectly correct and completely irrelevant.
Premature Calculation
Numbers begin moving before the relationships are understood.
Hidden Assumptions
The solution depends on something nobody tested.
First-Solution Fixation
The first plausible approach becomes the only approach.
One-Metric Optimisation
A solution improves speed while damaging accuracy, safety or quality.
Complexity Theatre
The method becomes more elaborate than the problem requires.
No Verification
The solver stops when an answer appears.
No Learning Loop
The same error returns because nobody examined its cause.
A 30-Day Scaffold for Becoming Better at Problem Solving
Week 1: Define and represent
- Take one problem per day and write current state, desired state, constraints and unknowns.
- Represent it in at least two forms.
- Identify one hidden assumption.
- Do not solve until the structure is visible.
Week 2: Decompose and generate
- Break problems into subproblems.
- Mark dependencies.
- Generate at least two approaches.
- Compare them using explicit criteria.
Week 3: Verify and falsify
- Check answers with a second method.
- Use sanity checks.
- Search for counterexamples.
- Add meaningful errors to an error log.
Week 4: Transfer
- Take a technique from Mathematics and apply it to writing, coding or planning.
- Explain one solved problem aloud.
- Solve one unfamiliar variant.
- Review the month for repeated patterns.
The aim after 30 days is not omniscience.
It is a stronger operating system for uncertainty.
How Parents Can Help Children Become Better Problem Solvers
When a child is stuck, immediate rescue can remove the most valuable part of the task.
Instead, ask questions that reopen the problem:
- What do you know for sure?
- What is the question asking?
- Can you draw it?
- What have you tried?
- Where did the first confusing step appear?
- Is there a simpler case?
- How could you check your answer?
The parent is not withholding support.
The parent is helping the child keep ownership of the reasoning.
How Teachers and Tutors Can Teach Problem Solving
Students need access to expert reasoning, not only expert answers.
Useful teaching moves include:
- model how a problem is represented;
- compare two methods;
- show why a tempting method fails;
- ask students to predict before calculating;
- use worked examples, then fade support;
- ask for verification;
- classify errors;
- include unfamiliar transfer tasks.
The deepest improvement comes when the student begins asking these questions independently.
How This Connects to Singapore’s Learning and Civilisation Graph
Problem solving sits underneath many of Singapore’s visible strengths because complex systems constantly encounter constraints: land, transport, housing, water, education, trade, health, technology and changing demographics.
That does not mean every solution is simple or uncontested. It means large systems survive by continuously identifying gaps, modelling constraints, choosing interventions, measuring outcomes and adapting.
The same mechanics appear at a smaller scale in a student solving an equation.
Different stakes. Different domain knowledge. Similar discipline: understand before acting, make relationships visible, test the result.
Explore the larger national graph through Singapore | How the Country Works and Holds Together and History of Education in Singapore | An Atlas of Schools, Towns, Subjects and Learning.
Frequently Asked Questions
Are some people naturally better problem solvers?
People differ in prior knowledge, experience, working-memory demands, habits and domain expertise. But many useful problem-solving behaviours — representation, decomposition, generating alternatives, verification and error review — can be deliberately practised.
Why do I freeze when I see an unfamiliar problem?
Unfamiliarity removes pattern cues. Begin by lowering the problem’s complexity: state the goal, list knowns and unknowns, represent the relationships and solve a simpler case.
Should I always try multiple methods?
No. Routine low-stakes problems may not justify it. Multiple methods are especially valuable when verification matters, the first approach is uncertain or the cost of an error is high.
Is trial and error bad?
Unstructured guessing is weak. Controlled experimentation can be excellent when each trial is informed, the result teaches you something and the cost of failure is acceptable.
How do I get better at mathematical problem solving?
Build mathematical knowledge, study worked solutions, practise choosing heuristics, solve unfamiliar problems, explain methods and verify answers. Do not train only repetitive execution.
Does problem solving require creativity?
Often, yes — especially when standard methods fail. But creativity works best when supported by domain knowledge, constraints and verification.
Can AI make me a better problem solver?
It can help generate alternatives, critique assumptions and create test cases. You improve only if you still define, evaluate, execute and verify rather than outsourcing every difficult step.
What is the difference between problem solving and decision making?
Problem solving focuses on moving from an undesirable or uncertain state toward a workable resolution. Decision making focuses on choosing among alternatives. In real life they often overlap.
Helpful Reading Inside eduKate
- How Mathematical Heuristics Work
- How to Break Complex Problems Into Smaller Problems
- How to Solve Problems With Super Intelligence
- How Intelligence Works
- How Mathematics Works
- How Science Works
- How to be Good at Teamwork
Research and Public References
- Ministry of Education, Singapore — 21st Century Competencies
- Ministry of Education, Singapore — Secondary Mathematics Syllabuses
- OECD — Collaborative Problem Solving
How to Be Good at Problem Solving
Good problem solving is less about having a magical answer and more about building a disciplined relationship with uncertainty.
Define the gap.
Separate facts from assumptions.
Represent the structure.
Break the problem carefully.
Use what you know.
Generate alternatives.
Choose explicitly.
Test what can be tested.
Verify what matters.
Learn from the result.
Then carry the useful structure into the next problem.
That is the gold standard: not a person who instantly knows every answer, but a person who can enter confusion and steadily make the problem more understandable, more testable and more solvable.
Continue the series with How to be Good at Studying, How to be Good at Teamwork and How to be Good at Public Speaking.
Properly taught kids shine a bright light into the future.
