Differentiation can estimate how much a function changes after a small input adjustment. Learn when a tangent approximation is useful and how to check its error.
Why this mathematical connection matters
Differentiation can estimate how much a function changes after a small input adjustment. Learn when a tangent approximation is useful and how to check its error.
A worked mathematical example
For f(x)=x² at x=10, f(10.1)≈100+20(0.1)=102, while exact f(10.1)=102.01. The error is 0.01.
Which concepts students should connect
Derivatives, tangent lines, linear approximation, exact finite differences, error bounds and units.
What the mathematical model assumes
Start by defining every variable and its units. Identify the domain, the fixed quantities and the information that the question does not supply. A correct equation is only as useful as the assumptions supporting it. A classroom geometric or calculus model does not independently establish real-world engineering performance.
How to check the calculation
Substitute the proposed result into the original conditions, use exact arithmetic before rounding, and compare the sign and size of the answer with the diagram or graph. If a derivative is involved, interpret its units as output change per input change. If a geometric scaling is involved, check whether the required quantity is one-, two- or three-dimensional.
A practical revision routine
First draw and label the mathematical situation. Next write the governing equation and its permitted domain. Solve one example with exact values. Change one input and explain how the output changes. Finally, explain a limitation of the model and solve a related question independently without copying the worked answer.
What parents can ask
Ask which mathematical idea makes the example work, which assumptions are necessary, what the answer’s units mean and how the learner would check a proposed solution. Ask whether the result is a mathematical prediction or a verified real-world measurement. These questions build reasoning rather than rote memorisation.
Singapore SEC G3 Additional Mathematics connection
The application draws on algebra, functions, graphs, trigonometry or calculus depending on the specific example. Advanced engineering or computing interpretation is enrichment, not a claim of a separately examined topic. Check the official 2027 G3 syllabus at SEAB 2027 G3 syllabus directory.
Small-group learning and further reading
A true three-student tutorial can distinguish an algebra error from a misconception about the graph or the model. The immutable tutorial reference is Secondary 1 Mathematics small-group tutorial in Clementi. Related Additional Mathematics reading is available at Additional Mathematics library.
Build the mathematical relationship
A tangent line provides a first-order approximation to a function near a known input. Begin with f(a+h)≈f(a)+f'(a)h. State the variables, coordinate convention and units. Explain how each coefficient changes the output. A model is not an observation; it is an assumed relationship whose consequences can be checked.
Calculate and verify an example
For f(x)=x² at a=10 and h=0.1, tangent estimate is 100+20(0.1)=102; exact result is 102.01. Substitution into the original equation verifies the numerical result. Where the problem involves a graph, check symmetry, intercepts, turning points and signs. A second method is valuable because it may reveal an arithmetic or interpretation error.
Go further with the mathematics
The exact error for f(x)=x² is h², since (a+h)²=a²+2ah+h². At h=0.1 it is 0.01; at h=0.01 it is 0.0001. The first-order approximation becomes better for smaller changes. For other functions, the size of the error depends on curvature and the interval. The approximation is not an exact identity unless the function is linear.
Explain what would change if the input doubled
Start from f(a+h)≈f(a)+f'(a)h. Replace the independent input with twice its original value, keeping all other parameters fixed. Simplify before making a numerical prediction. Do not assume that doubling an input always doubles an output: the exponent, function family and any fixed offsets determine the result.
Check units and the permitted domain
A length must be measured in compatible units. Area uses squared units and volume cubed units. A derivative has units of output per input. Identify whether the formula permits negative coordinates, only positive physical lengths or a restricted time interval. Exclude answers that violate the stated physical interpretation.
Separate an exact result from an approximation
The central worked example is For f(x)=x² at a=10 and h=0.1, tangent estimate is 100+20(0.1)=102; exact result is 102.01. Exact substitution proves an identity or checks a constructed function within its stated mathematical assumptions. It does not establish that a real manufactured object, physical measurement or engineered system matches the ideal formula. When a derivative approximation is used, state its approximate nature explicitly.
What students should be able to explain
A strong learner should identify the independent and dependent variables, construct the governing expression, substitute exact values, explain the graph and describe one limitation. They should be able to solve a modified problem with changed inputs, not merely repeat the same numerical steps.
A four-week study pathway
Week one: revise algebra, indices and coordinates relevant to the model. Week two: derive the governing relationship and complete three numerical checks. Week three: study graph shape, derivatives or scaling as applicable. Week four: solve a new example and write a short explanation of the model’s limitations.
Frequently asked questions
Does the mathematical formula describe every real object? No; its assumptions must hold. Can the formula be checked? Yes, substitute the original constraints and verify the dimensions. Is an attractive graph sufficient proof? No, algebraic verification is required. Is the advanced application a separate SEC examination topic? Not necessarily; the underlying functions, geometry and calculus are the assessed foundations.
Why study this in Singapore
The 2027 SEC G3 Additional Mathematics foundation includes relevant algebra, functions, graph interpretation, trigonometry and calculus. Consult the official SEAB syllabus directory for the exact assessed scope: SEAB 2027 G3 syllabus directory. The engineering, optics and approximation connections here are enrichment rather than examination predictions.
What a tangent approximation means
A tangent is a straight line that matches a differentiable curve’s value and instantaneous gradient at a chosen point. If f is differentiable at a, its tangent approximation near a is f(a+h)≈f(a)+f′(a)h, where h is a small input change. The formula predicts the new output by starting at the known output and adding gradient times input change. It is a local estimate, not a claim that a curved graph becomes exactly straight over a finite interval.
Why the derivative appears in the estimate
The derivative f′(a) is the limit of [f(a+h)−f(a)]/h as h approaches zero. For small nonzero h, this ratio is close to f′(a), so f(a+h)−f(a) is close to f′(a)h. Rearranging gives the tangent estimate. This derivation explains why multiplication by h is essential. The derivative alone is a rate per unit of input, not the total change. The approximation becomes more trustworthy as the input step becomes small and the function behaves smoothly near a.
A complete square-function example
Let f(x)=x² and estimate f(10.1) from a=10. We know f(10)=100 and f′(x)=2x, so f′(10)=20. The input change is h=0.1. The tangent prediction is 100+20(0.1)=102. The exact value is 10.1²=102.01. Thus the tangent underestimates by 0.01. This is a small error compared with the output, but it is not zero. The difference comes from the curvature of the quadratic.
Derive the exact error for a square
For f(x)=x², expand the exact new value: f(a+h)=(a+h)²=a²+2ah+h². The tangent approximation is a²+2ah. Subtracting gives exact minus estimate=h². The error is nonnegative for every real h and grows with the square of the step size. At h=0.1 it is 0.01; at h=0.01 it is 0.0001. This exact formula explains both the direction and magnitude of the error without relying on a calculator.
Why reducing the step improves the estimate
If the step size is divided by ten in the square-function example, the exact error h² is divided by one hundred. This quadratic reduction is a feature of a smooth function with a nonzero second-order term. It is not a universal guarantee for every function or every interval. Near a discontinuity, corner or rapidly changing curvature, a local linear approximation may fail or have a different error pattern. Always inspect the function and the size of the step before trusting a numerical estimate.
The tangent line in point-gradient form
At x=a, the tangent to y=f(x) passes through (a,f(a)) and has gradient f′(a). Its equation is y−f(a)=f′(a)(x−a). Rearranging gives y=f(a)+f′(a)(x−a). Substituting x=a+h yields the tangent approximation immediately. For f(x)=x² at a=10, the tangent is y−100=20(x−10), or y=20x−100. At x=10.1 this line gives 102. The tangent equation and the differential estimate are two ways to express the same local linear model.
A square-root estimate without a calculator
Estimate √101 using f(x)=√x near a=100. The known value is f(100)=10, and f′(x)=1/(2√x), so f′(100)=1/20. With h=1, the tangent estimate is 10+1/20=10.05. The exact √101 is approximately 10.0498756, so the estimate is slightly high. This example is useful because a nearby perfect square gives a convenient starting value and derivative. The method provides a sensible approximation while retaining the meaning of each step.
Why the square-root tangent overestimates
For x>0, the second derivative of √x is −1/(4x^(3/2)), which is negative. The graph is concave downward. Its tangent at a positive point lies above the nearby curve, so the tangent estimate tends to overestimate the true square-root value. This explains why 10.05 is greater than √101. The sign of curvature provides a qualitative error check before numerical calculation. A student who gets a lower estimate should revisit the derivative or arithmetic.
Estimate a cube root by a nearby perfect cube
Let f(x)=x^(1/3). To estimate ∛27.3, choose a=27, where f(27)=3. The derivative is f′(x)=(1/3)x^(−2/3), so f′(27)=1/(3×9)=1/27. The change is h=0.3. The estimate is 3+0.3/27=3+1/90≈3.01111. The exact cube root is close to this value. The method works because the nearby cube 27 provides an exact reference and the function is differentiable there.
Estimate a reciprocal
For f(x)=1/x, estimate 1/4.1 using a=4. We know f(4)=1/4 and f′(x)=−1/x², so f′(4)=−1/16. The step is h=0.1. The tangent estimate is 1/4−0.1/16=0.24375. The exact value 1/4.1≈0.24390244 is slightly larger. The derivative is negative, consistent with the reciprocal decreasing as x increases. This example shows how a tangent approximation can handle a rational function without long division.
A negative input change
Estimate f(9.9) for f(x)=x² using a=10. The step is h=−0.1, not +0.1. The tangent estimate is 100+20(−0.1)=98. The exact value is 9.9²=98.01. The error is again h²=0.01. A common mistake is to use the absolute size of the change and predict an increase even though the input decreased. The sign of h carries the direction of movement along the x-axis.
A linear function gives an exact answer
For f(x)=3x+5, the derivative is the constant three. At any a and any step h, f(a+h)=3a+3h+5=f(a)+3h. The tangent approximation is therefore exact, regardless of the step size. There is no curvature and no omitted higher-order term. This special case helps explain the method: tangent approximation replaces a curved function locally by a straight line. When the original function already is a straight line, nothing is lost.
A quadratic with a negative coefficient
Take f(x)=5−2x². At a=1, f(1)=3 and f′(1)=−4. Estimate f(1.1) with h=0.1: the tangent gives 3−4(0.1)=2.6. The exact value is 5−2(1.1²)=5−2.42=2.58. Here the tangent overestimates by 0.02 because the parabola is concave downward. Expanding f(1+h)=3−4h−2h² shows that exact minus estimate=−2h². The sign of the quadratic coefficient determines the error direction.
Estimate the change in a circle’s area
A circle has area A=πr². If its radius increases from 10 cm to 10.1 cm, the derivative with respect to radius is dA/dr=2πr. At r=10, the rate is 20π cm² per cm. With Δr=0.1 cm, the estimated area increase is ΔA≈20π(0.1)=2π cm². The exact increase is π(10.1²−10²)=2.01π cm². The tangent estimate misses 0.01π cm². This is a geometric sensitivity calculation, not a measurement of a physical object.
Estimate the change in a sphere’s volume
For a sphere V=(4/3)πr³, the derivative is dV/dr=4πr². At r=5 cm, the derivative is 100π cm², which means the volume changes at about 100π cubic centimetres per centimetre of radius. If radius increases by 0.02 cm, the estimated volume increase is 100π×0.02=2π cm³. The exact change is (4/3)π(5.02³−5³), which is slightly larger. The units of the derivative times the radius change become cubic centimetres, as required.
Why the units of a derivative matter
If an output is measured in metres and the input in seconds, the derivative has units metres per second. Multiplying by a time change in seconds gives an estimated distance change in metres. If area is measured in square centimetres and radius in centimetres, dA/dr has units of centimetres. Multiplying by a radius change in centimetres gives square centimetres. Unit analysis is a useful independent check: adding a derivative directly to a function value without multiplying by an input change generally combines incompatible units.
Relative changes and percentages
Suppose y=x² with x>0. A small change dx gives dy≈2x dx. Divide by y=x² to obtain dy/y≈2 dx/x. Thus a small one-percent increase in x produces approximately a two-percent increase in y. The exact change for a one-percent increase is (1.01)²−1=0.0201, or 2.01 percent. The differential estimate of two percent is close but not exact. This is the connection between tangent approximations and the familiar squared scale factor.
Relative change for a cubic quantity
For V=Cs³ with constant C>0, differentiate to obtain dV=3Cs²ds. Dividing by V gives dV/V≈3ds/s for small finite changes. A one-percent increase in a characteristic length therefore predicts approximately a three-percent volume increase. The exact multiplier is 1.01³=1.030301, a 3.0301-percent increase. The approximation is most useful when the relative change is small; doubling a length requires the exact cube law rather than the tangent estimate.
A rate-of-change interpretation
Suppose an ideal model has height h(t)=t² metres after t seconds. At t=3, the instantaneous rate is h′(3)=6 metres per second. Over the next 0.1 second, the tangent estimate predicts an increase of 6×0.1=0.6 metre. The exact change is 3.1²−3²=0.61 metre. The model is purely illustrative; it is not a realistic description of a freely moving object. The mathematics shows how an instantaneous rate approximates a short-interval change.
Estimate an exponential change
For f(x)=e^x, the derivative equals e^x. Near x=0, f(0)=1 and f′(0)=1. Therefore e^0.02≈1+0.02=1.02. The actual value is about 1.02020134, so the tangent underestimates slightly. The graph is concave upward because its second derivative is positive. This simple estimate is the first-order approximation e^h≈1+h for small h. It should not be used without checking the size of h when greater accuracy is required.
Estimate a logarithm near one
For f(x)=ln x, the derivative is 1/x. At a=1, f(1)=0 and f′(1)=1. Thus ln(1.03)≈0+1(0.03)=0.03. The exact value is about 0.0295588, so the tangent overestimates. The logarithm is concave downward for positive x because its second derivative is −1/x². This example shows how the same method applies to a function that is neither a polynomial nor a square root.
Why a tangent approximation can fail near zero
For f(x)=√x, the derivative 1/(2√x) grows without bound as x approaches zero from the right. A small absolute change near zero can represent a large relative change, and the local slope changes rapidly. The tangent formula at x=0 is not available as a finite two-sided derivative. At a=0.01, a step of 0.01 is as large as the starting input and may produce a noticeably poor approximation. The method requires a suitable differentiable reference point and a sufficiently small step relative to local behaviour.
Why corners are different
For f(x)=|x| at x=0, the graph has a sharp corner. The left-hand slope is −1 and the right-hand slope is +1, so there is no single derivative f′(0). Consequently one tangent line cannot approximate the function from both sides using the standard differentiability formula. Away from zero, |x| is linear on each side and the tangent estimate is exact while remaining on that side. This example helps students distinguish a continuous function from a differentiable one.
Second derivatives and the direction of error
A twice-differentiable function with positive second derivative is locally concave upward, so its tangent tends to lie below the curve. With negative second derivative, the tangent tends to lie above the curve. This predicts the sign of exact minus estimated value near the reference point. It does not automatically provide the exact error size. The second derivative describes curvature, while a more detailed Taylor expansion or an exact algebraic comparison is needed to quantify the remainder.
A simple Taylor expansion connection
If f is sufficiently smooth near a, its local expansion begins f(a+h)=f(a)+f′(a)h+[f″(a)/2]h² plus higher-order terms. The tangent estimate keeps the first two terms and discards the rest. For f(x)=x², f″(x)=2 and all higher derivatives vanish, so the omitted term is exactly h². For f(x)=e^x near zero, the omitted terms begin h²/2. This is a conceptual extension explaining why errors often scale approximately with h² when the step is small.
Compare two step sizes quantitatively
Estimate 10.2² using a=10. With h=0.2, the tangent gives 100+20×0.2=104, while the exact answer is 104.04, so the error is 0.04. For h=0.1, the error was 0.01. Doubling the step quadrupled the error because the square-function remainder is h². This numerical pattern is useful for checking calculations. It also illustrates why a tangent approximation that looks excellent for a very small step may become less accurate when the same method is applied farther from the reference point.
A tangent approximation with a shifted square root
Estimate √(25.5) using f(x)=√x at a=25. We know f(25)=5 and f′(25)=1/10. The step is h=0.5, so the estimate is 5+0.05=5.05. The exact square root is approximately 5.0497525, giving an overestimate of about 0.0002475. Choosing a nearby convenient reference value simplifies the arithmetic. If the reference point were a number whose square root is not known, the tangent method would lose much of its practical advantage.
An exact check by squaring the estimate
For √101, the tangent estimate is 10.05. Squaring 10.05 gives 101.0025, slightly above 101. Since the square-root function is increasing for positive inputs, this confirms 10.05>√101. The check establishes the direction of error without needing a long decimal expansion. It also illustrates a general strategy: substitute an approximate answer into the defining equation and compare with the required value. A numerical approximation becomes more trustworthy when it can be checked independently.
Choose a good reference point
A useful reference a should be close to the target input, lie inside a differentiable region and have a value and derivative that are easy to calculate. For √101, a=100 is ideal because √100=10. For ∛27.3, a=27 works because ∛27=3. For ln1.03, a=1 is convenient because ln1=0. The choice of a can make the same approximation problem simple or cumbersome. Students should practise selecting the reference rather than treating it as always supplied.
The difference between an estimate and a bound
A tangent approximation gives a predicted value. It does not automatically provide a guaranteed interval containing the exact value. To establish a rigorous error bound, one needs additional information, such as a bound on the second derivative over the interval. For the square function, the exact error h² is known. For other functions, a statement like ‘the error is small’ should be supported by a calculation or a justified smoothness condition. This distinction matters when accuracy requirements are strict.
An error bound from a second derivative
If f has a continuous second derivative between a and a+h and |f″(x)|≤M throughout that interval, Taylor’s theorem gives |f(a+h)−[f(a)+f′(a)h]|≤(M/2)h². For f(x)=√x on [100,101], |f″(x)|=1/(4x^(3/2))≤1/4000. With h=1, the absolute tangent error is at most 1/8000=0.000125. The actual error for √101 is about 0.0001244, consistent with the bound. This is an extension beyond a basic tangent estimate, but it explains how reliability can be quantified.
A tangent approximation for a composite function
Let f(x)=(2x+1)^3. At a=1, f(1)=27. By the chain rule, f′(x)=6(2x+1)², so f′(1)=54. For h=0.01, the tangent estimate is f(1.01)≈27+54×0.01=27.54. The exact value is (3.02)³=27.543608. The error is 0.003608. This example connects the chain rule to local linear approximation: the derivative must be computed correctly before it can be used to estimate a change.
A tangent approximation for a rational function
Let f(x)=1/(x+1). At a=1, f(1)=1/2 and f′(x)=−1/(x+1)², so f′(1)=−1/4. To estimate f(1.1), take h=0.1. The tangent gives 0.5−0.025=0.475. The exact value is 1/2.1≈0.4761905. The approximation underestimates by about 0.00119. The sign of the derivative correctly predicts that the function decreases as x increases, even though the tangent does not give the exact magnitude of the finite change.
An exam-style multi-part problem
Question: For f(x)=x³, use the tangent at x=2 to estimate f(2.05), then find the exact error. First, f(2)=8 and f′(x)=3x², so f′(2)=12. With h=0.05, the estimate is 8+12(0.05)=8.6. The exact value is 2.05³=8.615125. Thus exact minus estimate=0.015125. Expanding (2+h)³ gives 8+12h+6h²+h³, so the omitted terms are 6h²+h³. At h=0.05 they equal 0.015+0.000125=0.015125, confirming the calculation.
A second exam-style problem
Question: Estimate the increase in the area of a square when its side changes from 8 cm to 8.05 cm. The area is A=s² and its derivative is 2s. At s=8, the rate is 16 cm² per cm. The side change is 0.05 cm, so ΔA≈16×0.05=0.8 cm². The exact increase is 8.05²−8²=64.8025−64=0.8025 cm². The tangent estimate underestimates by 0.0025 cm². State the units: an area change must be in square centimetres, not centimetres.
Independent practice with answers
Question A: Estimate 5.1² using a=5. Answer 25+10×0.1=26, exact 26.01. Question B: Estimate √64.8 using a=64. Answer 8+(1/16)×0.8=8.05. Question C: Estimate 1/9.1 using a=9. Answer 1/9−0.1/81, approximately 0.1098765. Question D: Estimate e^0.01 using a=0. Answer 1.01. Question E: For f(x)=4−x² at a=1 and h=0.1, estimate f(1.1) as 3−2×0.1=2.8; exact value is 2.79. In each case, identify f(a), f′(a) and h.
A harder practice set
Question F: Estimate ∛8.12 using a=8. Since f′(8)=1/(3×4)=1/12, the estimate is 2+0.12/12=2.01. Question G: For f(x)=sin x at a=0 in radians, estimate sin0.05≈0.05. Question H: For f(x)=ln x at a=1, estimate ln0.98≈−0.02. Question I: For f(x)=x⁴ at a=1 and h=0.02, estimate 1+4×0.02=1.08; exact 1.02⁴=1.08243216. Question J: Explain why f(x)=|x| has no single tangent approximation at x=0: the one-sided gradients differ.
Diagnose the first incorrect step
If a learner estimates f(10.1) for f(x)=x² as 100+20, they have omitted multiplication by h=0.1. If they use h=10.1 rather than 0.1, they have confused the target input with the change from the reference point. If they estimate f(9.9) by adding two instead of subtracting two, they have lost the negative sign of h. If they claim the tangent answer is exact for a curved function, they have confused a local approximation with the original curve. Each mistake needs a targeted correction.
A structured revision routine
Begin with a linear function, where the tangent is exact. Next use x² and calculate the exact error h². Then estimate a square root near a perfect square, followed by a reciprocal and an exponential. After that, include a composite function requiring the chain rule. Finally, ask the learner to choose the reference point independently and to state whether the tangent should overestimate or underestimate based on curvature. This sequence builds both procedural accuracy and conceptual understanding.
What parents can ask after the lesson
Ask what the known starting point is, what the input change h means and why the derivative must be multiplied by that change. Can the learner explain why the tangent is only an approximation? Can they check the estimate by substituting it into the original expression or by comparing with an exact value? Why is a nearby convenient reference point helpful? These questions focus on reasoning and error checking rather than on memorising one formula.
Frequently asked questions about tangent estimates
Is a tangent estimate always exact? Only for a linear function or special circumstances. Does a smaller step always help? Often for smooth functions near the reference point, but numerical rounding and irregular behaviour can complicate matters. Why use a derivative? It gives the local rate of change. What does a negative h mean? The target input lies below the reference input. Can I use the method at a corner? Not with a single ordinary derivative there. How can I estimate error? Compare with an exact result or use additional information about curvature.
Singapore syllabus and enrichment boundary
Differentiation, gradients and rates of change are part of the mathematical foundation of Additional Mathematics. Tangent estimates and error bounds in this guide illustrate how derivatives can be interpreted; they should not all be assumed to be separately examined applications. Consult the official SEAB G3 Additional Mathematics syllabus for the precise assessed content. The worked examples are original mathematical illustrations rather than measured engineering data. The central transferable skill is to connect the derivative to a small input change and then check the approximation.
Final checklist and worked conclusion
To estimate f(a+h), first select a differentiable reference point a near the target. Calculate f(a), then f′(a), and identify the signed step h. Form f(a)+f′(a)h and keep units consistent. Check whether the result should increase or decrease and whether curvature suggests an overestimate or underestimate. If an exact value or error bound is available, compare it with the tangent prediction. For f(x)=x² at a=10 and h=0.1, this process gives 102 against exact 102.01, with error 0.01. The approximation is useful precisely because its assumptions and limitations can be explained.
A full check of the approximation for a cubic
For f(x)=x³ at a=3, the derivative is 3x², so f′(3)=27. With h=−0.02, the tangent predicts f(2.98)≈27+27(−0.02)=26.46. The exact value is 2.98³=26.463592. Thus the estimate is lower than the exact value by 0.003592. Expanding (3+h)³ gives 27+27h+9h²+h³. The omitted terms are 9(0.0004)−0.000008=0.003592. This confirms both the sign and magnitude of the error. It also demonstrates that the higher-order terms depend on the starting point as well as the step.
Use a tangent estimate to interpret measurement uncertainty
Suppose a measured radius is r=10 cm with a small uncertainty of approximately 0.02 cm. For the area A=πr², the differential magnitude is |dA|≈2πr|dr|=20π×0.02=0.4π cm², approximately 1.257 cm². This is a first-order sensitivity estimate, not a guaranteed bound unless additional assumptions about the measurement error and curvature are supplied. It shows how a small uncertainty in an input can be translated into an approximate uncertainty in a calculated output.
How to compare absolute and relative errors
Absolute error is the magnitude of exact value minus estimated value. Relative error divides that magnitude by the magnitude of the exact value, provided the exact value is nonzero. In the x² example at 10.1, the absolute error is 0.01, while the relative error is 0.01/102.01≈0.0000980, or about 0.0098 percent. An absolute error of 0.01 may be small for an output near one hundred but substantial for an output near zero. State which error measure is relevant to the question rather than treating every decimal difference as equally important.
An estimate is not a replacement for exact algebra
If a question asks for an exact square root, an exact fraction or a proof, a tangent approximation alone is not sufficient. The approximation is useful when a numerical estimate is requested, when a derivative is being interpreted as a rate of change or when a quick reasonableness check is needed. Exact algebra remains preferable when it is available and manageable. A well-trained learner should recognise the instruction words ‘estimate’, ‘approximately’ and ‘exactly’ and select the corresponding level of precision and justification.