Additional Mathematics Synthesis Guide 42: Backward Reasoning and Target Decomposition — Let the Required Answer Choose the Route

Read the target → ask what would immediately produce it → ask what would produce that → stop when you reach available information → execute forward.

1. The target is a mathematical object

A question may ask for a value, proof, equation, coordinate, range, maximum, parameter, tangent, area or time. Each output type suggests a different family of useful predecessors.

For example, if the target is a tangent equation, the immediate needs are usually a point of contact and a gradient.

2. Backward planning is not reverse algebra

The final written solution should still proceed from known information to the conclusion. Backward reasoning is a planning process used before or between lines.

The learner asks what must be true just before the final step, not whether the algebra can literally be performed backward without care.

3. Tangent equation target

Suppose the target is “find the equation of the tangent to y=f(x) at x=a”.

• Final form needs y−y₁=m(x−x₁).
• Therefore we need the point (a,f(a)).
• We also need m=f′(a).
• Therefore we need to differentiate f.

The required answer has already chosen the architecture of the solution.

4. Worked tangent example

For y=x²+1 at x=2:

point=(2,5)

and f′(x)=2x, so m=4.

Hence

y−5=4(x−2),

or y=4x−3.

5. Maximum-value target

If the target asks for a maximum of a differentiable function on an interval, backward planning asks:

• Which candidate x-values can produce extrema?
• Stationary points require f′(x)=0.
• Closed intervals may also require endpoint checks.
• Classification may need derivative signs or f″.

The word “maximum” therefore activates a candidate-generation and comparison plan rather than mere differentiation.

6. Area target

If the target is an area between a curve and the x-axis, the immediate need is an integral over the correct interval, but first the learner must know where the region begins and ends and whether the curve crosses the axis.

So the backward chain may be:

area ← definite integral(s) ← interval boundaries ← roots/intersections.

7. Parameter target

If the question asks for a parameter such that a line is tangent to a curve, the target “parameter value” is reached through a tangency condition.

Possible backward routes include:

• repeated intersection root → discriminant zero;
• same point and same gradient → derivative condition.

The target does not dictate one unique method, but it narrows the method family.

8. Proof target

When asked to prove an identity, inspect the desired final form. If the target contains tanx, a useful predecessor may be sinx/cosx. If the target contains a single angle, addition or double-angle identities may compress the left side.

The target gives clues about which representation is likely to reduce distance to the conclusion.

9. Solve-for-x target

“Solve” is still too broad. Backward reasoning asks what form would make the equation directly solvable.

• product=0 → zero-product branches;
• quadratic=0 → factorisation/formula/completing square;
• one trig function=value → interval solution structure;
• aˣ=k → matching bases or logarithms;
• f(x)=g(x) → intersection equation.

10. Target decomposition reduces search cost

A student who knows many methods can become slower because every method competes for attention. Target decomposition removes irrelevant branches.

If the target is a minimum value of a quadratic, expanding a factorised form may be less useful than completing the square. If the target is roots, factor form may be preferable.

11. The shortest route is not always the best route

A route should be judged by reliability, not only line count. A slightly longer method may preserve structure, reduce sign risk and make checking easier.

Backward planning should therefore optimise for a trustworthy path to the target, not merely the fewest written steps.

12. Given information has roles

In a long question, every datum need not be used immediately. Some information defines the object, some determines a parameter, some locates an interval, and some verifies a later result.

Backward reasoning helps assign those roles instead of forcing every given number into the first equation.

13. Worked line-circle planning example

Suppose the target is the equation of a tangent to a circle at a known point P.

• Tangent equation needs tangent gradient.
• Tangent is perpendicular to radius OP.
• Radius gradient comes from centre and P.
• Then use the perpendicular-gradient relationship and point P.

The geometry chooses the algebraic dependencies before any calculation begins.

14. Worked logarithmic planning example

Suppose the target is to solve

ln(x−1)+ln(x+1)=ln8.

To reach x-values, a useful predecessor is a polynomial equation. To get that polynomial, combine the logarithms. Before doing so, record x>1.

The plan becomes:

x-solutions ← polynomial ← combined logarithm ← original log domain.

15. Trigonometric target decomposition

If an equation contains sin2x and sinx, the target “solve for x” suggests first reducing the equation to one trig function or factorised branches.

For sin2x=sinx:

2sinx cosx−sinx=0

sinx(2cosx−1)=0.

Now each branch has the familiar target form “one trig function equals a constant”.

16. Kinematics target decomposition

If asked for total distance, the immediate need is the path length over each direction interval. Therefore we need turning times, which come from v=0 plus sign changes. If only s(t) is given, differentiate first.

The dependency chain is:

total distance ← interval distances ← turning times ← velocity ← displacement.

17. A target may require several predecessor objects

To optimise a geometric quantity, one may need:

• a diagram;
• a constraint equation;
• a one-variable objective function;
• its derivative;
• candidate points;
• a final contextual interpretation.

Target decomposition turns one large problem into a dependency graph.

18. Work backward until you reach available information

Planning should stop when the needed predecessor is already given or can be obtained by a standard operation from given data. Continuing backward indefinitely is unnecessary.

The purpose is to find a bridge from the target to the known information, not to analyse the entire subject.

19. Then execute forward

Once the dependency chain is clear, write the solution in forward logical order. Establish the first needed intermediate result, pass it into the next stage, and continue until the target is produced.

This keeps the reasoning defensible to another reader.

20. The target can change the preferred representation

For y=x²−6x+5:

• target roots → factorise;
• target minimum → complete the square;
• target y-intercept → expanded form is already convenient;
• target tangent at x=a → differentiate the expanded form or any equivalent differentiable form.

The same object supports different best routes because the target changes.

21. Answer-form control

Backward reasoning should begin with the actual required answer form. “Find x”, “show that”, “find the coordinates”, “state the range”, “give the exact value”, and “find the maximum” are different contracts.

A correct intermediate quantity is not enough if it is not converted into the requested final object.

22. A reliable target-decomposition routine

1. Underline the required output and its conditions: exactness, interval, coordinates, parameter range, proof or interpretation.
2. Ask what object would immediately produce that output.
3. Repeat until you reach information already given or directly computable.
4. Choose among alternative predecessor routes using reliability and economy.
5. Record restrictions that must survive the chain.
6. Execute forward in a clear dependency order.
7. At each handoff, ask whether the intermediate result is in the form the next stage needs.
8. Finish by checking that the final line answers the original target exactly.

23. Common failure patterns

• Starting calculations before identifying the requested object.
• Using every given number immediately because it appears in the question.
• Choosing the first remembered method rather than the route aligned with the target.
• Planning a maximum question as “differentiate” but forgetting endpoints or classification.
• Planning an area question without first locating boundaries and sign changes.
• Finding an intermediate gradient but never forming the tangent equation.
• Finding candidate parameter thresholds but not stating the required interval of values.
• Working backward algebraically through a non-reversible step as though logic were symmetric.
• Choosing a short route that is fragile and hard to verify.
• Finishing with a correct result that does not match the requested answer form.

24. Practice set

1. If the target is a tangent equation at x=a, name the two immediate quantities needed.
2. If the target is a maximum on a closed interval, what candidates should be considered?
3. If the target is area under a curve, what must be known before integrating?
4. If a line is tangent to a quadratic, give one algebraic threshold condition.
5. Give one calculus tangency route.
6. For y=x²−6x+5, which form best exposes the roots?
7. Which form best exposes the minimum?
8. What is the backward dependency for total distance from a displacement function?
9. Why should a long question’s data be assigned roles rather than used immediately?
10. What does “execute forward” mean after backward planning?
11. For ln(x−1)+ln(x+1)=ln8, what domain must be recorded first?
12. What useful predecessor form can a log equation often be transformed into?
13. For sin2x=sinx, what factorised equation is useful?
14. If the target is a circle tangent equation at P, why is the radius relevant?
15. What does the word “exact” change about the final answer contract?
16. Why can the shortest algebraic route be a poor choice?
17. When should backward decomposition stop?
18. Why is a dependency graph useful in optimisation?
19. What is wrong with ending a tangent question at the gradient?
20. What final check should always be made?

Answers

1. The point of contact and tangent gradient.
2. Admissible stationary points and relevant endpoints.
3. The interval boundaries/intersections and whether sign changes require splitting.
4. Discriminant zero in the line-curve intersection quadratic.
5. Same point and same gradient: f(a)=line value and f′(a)=line gradient.
6. Factor form (x−1)(x−5).
7. Completed-square form (x−3)²−4.
8. distance ← interval displacements ← turning times ← velocity ← displacement.
9. Different data may define, constrain, determine or verify different stages; premature use can create unnecessary equations.
10. Write the justified solution from given information through the planned intermediates to the target.
11. x>1.
12. A polynomial or a one-function equation, depending on the structure.
13. sinx(2cosx−1)=0.
14. The tangent is perpendicular to the radius at the point of contact, so the radius gradient determines tangent gradient.
15. The result should remain in an exact symbolic form where possible rather than being replaced by a rounded decimal.
16. It may increase sign risk, hide structure or make independent checking harder.
17. When the chain reaches given information or a directly available standard operation.
18. It separates the objective, constraints, substitution and calculus stages so dependencies are explicit.
19. A gradient is only an intermediate object; the requested answer is the line equation.
20. Check that the final line answers the exact requested object with all domain, interval, unit and accuracy conditions satisfied.

25. What mastery looks like

Mastery means the learner can use the target to prune method search, identify the minimum set of intermediate objects required, choose a reliable representation for each stage and execute the plan forward without losing restrictions.

The transfer test is to give a multi-topic question whose chapter is not named. If the learner can begin by describing the dependency chain from the final target back to available information, rather than trying random procedures, backward reasoning has become a practical problem-solving architecture.

---

Continue through Batch 11

• Guide 41: Representation Switching — Words, Equations, Graphs, Geometry and Rates
• Guide 43: Linked Multi-Stage Problems — Hence, Result Handoffs, Parameters and Preserving Conditions
• Guide 44: Mathematical Modelling — Formulation, Variables, Assumptions, Parameter Meaning, Validation and Return to Context
• Additional Mathematics Directory
• BTT Mathematics Hub