Advanced Performance Management · Budgetary planning and control

Quantitative Techniques in Budgeting for ACCA APM

Updated 11 October 2026 · Fact-checked

Quantitative techniques in budgeting use data to forecast activity and costs. You split costs into fixed and variable with high-low or regression, adjust labour for learning, and weight uncertain outcomes with expected values. In APM you must then explain how reliable the budget is and what limits it.

Understand Quantitative Techniques in Budgeting

A budget is only as good as the forecasts behind it. Quantitative techniques turn past data into numbers you can defend. They answer three questions: how will costs behave, how will efficiency change, and how do we deal with uncertainty?

For cost behaviour, the high-low method uses only two data points: the highest and lowest activity levels. It is quick but ignores every other observation. Linear regression uses all the data and fits a line y = a + bx, where a is fixed cost and b is variable cost per unit. It is usually more reliable, but it still assumes a linear relationship and works best inside the range of the data.

For efficiency, the learning curve says that when a task is repeated, the time per unit falls as workers gain experience. With an 80% curve, the cumulative average time per unit falls to 80% each time cumulative output doubles. This matters for new products, new processes and labour-intensive work. It does not suit stable, automated or very long-established production.

For uncertainty, an expected value is the probability-weighted average of possible outcomes. It gives a single budget figure but hides the spread of results. It also suits repeated decisions better than a one-off decision. Time series analysis (trend plus seasonal variation) is used the same way to forecast sales.

APM questions rarely stop at the calculation. You are expected to say how reliable the figure is, which assumptions it rests on, and how the budget should be used for control. That is where the professional skills marks sit.

Key rules to remember

High-low variable cost: Variable cost per unit = (Cost at high activity − Cost at low activity) ÷ (High activity − Low activity); Use the highest and lowest activity levels, not the highest and lowest costs. Then fixed cost = total cost at either point − (variable cost per unit × activity at that point).
Linear cost function: y = a + bx; y is total cost, a is fixed cost, b is variable cost per unit of activity, x is activity.
Regression slope (b): b = (nΣxy − ΣxΣy) ÷ (nΣx² − (Σx)²); n is the number of data pairs. Learn the formula and be able to apply it.
Regression intercept (a): a = (Σy − bΣx) ÷ n; Equivalent to a = ȳ − b × x̄. Calculate b first.
Correlation coefficient: r = (nΣxy − ΣxΣy) ÷ √[(nΣx² − (Σx)²)(nΣy² − (Σy)²)]; r lies between −1 and +1. Values near ±1 show a strong linear relationship.
Coefficient of determination: r² = proportion of variation in y explained by x; If r = 0.9, then r² = 0.81, so 81% of the variation in cost is explained by activity.
Learning curve (cumulative average time model): Y = aX^b, where b = log(learning rate) ÷ log 2; Y is cumulative average time per unit for X units, a is time for the first unit. For 80%, b = log 0.8 ÷ log 2 = −0.322. When X doubles, Y falls to the learning rate of its previous value.
Time for later units: Time for units m+1 to n = (n × average time for n units) − (m × average time for m units); Always work from cumulative totals, then subtract.
Expected value: EV = Σ(probability × outcome); Probabilities must add to 1. The EV may not be a value that can actually occur.
Additive time series forecast: Forecast = Trend + Seasonal variation; Seasonal variations in the additive model should sum to about zero over a full cycle.

How to solve Quantitative Techniques in Budgeting questions

Use this order for any question on forecasting, cost estimation or learning in a budget.

1Read the requirement and mark what the budget needs: a cost per unit, a total cost, labour hours, or a revenue forecast. Note the activity level you must forecast for.
2Choose the technique the data supports. Two data points or a quick estimate: high-low. A full data set: regression. Repeated manual work on a new task: learning curve. Uncertain outcomes with probabilities: expected value.
3Check the assumptions before you calculate. Is the data in the relevant range? Is the learning rate given for cumulative average time? Do the probabilities sum to 1?
4Calculate in stages and show each stage. For regression, set out n, Σx, Σy, Σxy and Σx² first. For learning curves, find cumulative totals before subtracting.
5Apply the result to the budget. Convert hours to cost with the labour rate, add fixed cost, and use the budgeted activity level, not the historic one.
6Test the answer for sense. Compare it with past figures. Check that the cost per unit is falling under a learning curve and that fixed cost is not negative without reason.
7Comment on reliability. Mention the range of the data, the fit (r²), the stability of the learning rate, and the subjective probabilities. Link to how managers should use the budget.
8If professional skills are marked, state your conclusion clearly and recommend an action, such as updating the forecast as actual data arrives.

Quickest way: Fast route under time pressure

When to use it: Use this when a question gives a clear data set and you have limited time for the calculation, so you can protect time for the written discussion.

Use the high-low method only if asked, or if the data has just two useful points. Otherwise go straight to regression.
For regression, build a five-column table (x, y, xy, x², and totals) once. Do not recalculate sums.
For learning curves, use the doubling shortcut where possible. For 8 units at 80%, the cumulative average is first-unit time × 0.8 × 0.8 × 0.8.
For units not at a doubling point, use the formula Y = aX^b with a calculator, then convert to a total.
Write one line on assumptions and one line on reliability for each technique. These lines earn marks for little time.

Common mistakes in Quantitative Techniques in Budgeting

Using the highest and lowest cost figures in the high-low method instead of the highest and lowest activity levels.
Students scan the cost column and pick the extremes by habit.
Fix: Look at the activity column only. Pick the highest and lowest activity and take the costs that go with them.
Treating the learning rate as a reduction in the time of each unit rather than in the cumulative average time.
The phrase '80% learning curve' sounds like each unit takes 80% of the previous one.
Fix: Remember that the cumulative average falls to 80% when cumulative output doubles. Find the total time from the average, then subtract to get the time for later batches.
Forgetting to subtract earlier units when the budget covers only later units, such as units 5 to 8.
Students stop after calculating total time for 8 units.
Fix: Calculate cumulative total time for both points and subtract. Write 'time for units 5 to 8 = total for 8 − total for 4'.
Extrapolating a regression line far outside the range of the data and presenting the result as reliable.
The formula produces a number, so it looks precise.
Fix: Forecast within or close to the data range. If you must go beyond it, say the relationship may not hold, for example because of step costs or capacity limits.
Using the expected value as if it were the most likely outcome, or ignoring the spread of outcomes.
A single figure is easy to put in a budget.
Fix: Explain that EV is an average. Mention the worst and best cases and that probabilities are often subjective.
Giving only calculations and no comment on limitations or use of the budget.
Students treat APM like a computational paper.
Fix: After each calculation, add two or three points: key assumptions, reliability, and what management should do with the result.

Worked examples

Example 1

A company launches a new product. The first unit needs 100 labour hours. An 80% learning curve (cumulative average time model) applies. Labour costs $20 per hour. Calculate the budgeted labour cost of units 5 to 8 and comment on one limit of the estimate.

Show the solution

Cumulative average time for 4 units: 4 is two doublings from 1 (1 → 2 → 4). So the average is 100 × 0.8 × 0.8 = 64 hours.
Total time for 4 units = 4 × 64 = 256 hours.
Cumulative average time for 8 units: three doublings. Average = 100 × 0.8 × 0.8 × 0.8 = 51.2 hours.
Total time for 8 units = 8 × 51.2 = 409.6 hours.
Time for units 5 to 8 = 409.6 − 256 = 153.6 hours.
Labour cost = 153.6 × $20 = $3,072.
Comment: the estimate assumes the 80% rate holds and that labour is stable. If workers leave or the process changes, the rate may not hold and the budget would be too low or too high.

Answer: Budgeted labour cost of units 5 to 8 = 153.6 hours × $20 = $3,072. The result depends on the assumed learning rate staying constant.

Example 2

A firm records machine hours (x, in thousands) and overhead cost (y, in $ thousands) over five periods: (1, 12), (2, 15), (3, 19), (4, 22), (5, 26). (a) Use regression to estimate the cost function and forecast overhead for 6,000 hours. (b) Compare with the high-low method. (c) Next year's hours are 5,000 (probability 0.2), 6,000 (0.5) or 7,000 (0.3). Find the expected overhead using the regression line.

Show the solution

Sums: n = 5, Σx = 15, Σy = 94, Σxy = 12 + 30 + 57 + 88 + 130 = 317, Σx² = 1 + 4 + 9 + 16 + 25 = 55.
b = (5 × 317 − 15 × 94) ÷ (5 × 55 − 15²) = (1,585 − 1,410) ÷ (275 − 225) = 175 ÷ 50 = 3.5.
a = (94 − 3.5 × 15) ÷ 5 = (94 − 52.5) ÷ 5 = 8.3. Cost function: y = 8.3 + 3.5x.
Forecast at x = 6: 8.3 + 3.5 × 6 = 8.3 + 21 = 29.3, which is $29,300.
High-low: high is x = 5, y = 26. Low is x = 1, y = 12. Variable cost = (26 − 12) ÷ (5 − 1) = 3.5 per thousand hours. Fixed cost = 26 − 3.5 × 5 = 8.5. Forecast at 6: 8.5 + 21 = 29.5, which is $29,500.
Comparison: the variable rates agree, but the fixed cost differs by $200. Regression uses all five points, so it is generally more reliable. High-low depends on two points that could be unusual.
Expected hours = (5 × 0.2) + (6 × 0.5) + (7 × 0.3) = 1.0 + 3.0 + 2.1 = 6.1 thousand hours.
Expected overhead = 8.3 + 3.5 × 6.1 = 8.3 + 21.35 = 29.65, which is $29,650. This works because the cost function is linear.
Comment: 6,000 and 7,000 hours lie beyond the data range (1 to 5), so the cost function may not hold if step costs or capacity limits appear. The probabilities are subjective.

Answer: (a) y = 8.3 + 3.5x; forecast for 6,000 hours = $29,300. (b) High-low gives $29,500, so the methods are close but regression is preferred as it uses all the data. (c) Expected overhead = $29,650, with the caveat that the forecast goes beyond the data range.

Exam tips

Show the working table for regression and the doubling steps for learning curves. Method marks are available even if the arithmetic slips.
Always link the technique to the budget. Say what activity level you are budgeting for and which cost it feeds into.
Prepare two or three limits for each technique, such as small sample size, linearity, stable learning rate, and subjective probabilities. These points score well in the discussion part.
Watch the wording of the learning curve. If the question states the learning rate applies to the cumulative average time, use the doubling method. Say which model you use.
Use the professional skills marks: give a clear recommendation, such as updating the forecast as actual results come in, and write it in a way a finance director could act on.

Practice questions from Budgetary planning and control

Quantitative Techniques in Budgeting: frequently asked questions

When should I use high-low instead of regression in APM?

Use high-low when the question asks for it or when only a few data points are given and speed matters. Use regression when you have a full data set, because it uses every observation. In your answer, say that high-low is simple but relies on two points that may be unrepresentative.

How do I use a learning curve in a budget question?

Find the cumulative average time for the output level, multiply by units to get total time, and subtract earlier cumulative time if only later units are needed. Then multiply hours by the labour rate. Add a comment on whether the learning rate is likely to hold.

What does r² tell me about a regression forecast?

It shows the proportion of the variation in cost explained by activity. A value close to 1 means the line fits the data well, so the forecast is more reliable. It does not prove that activity causes the cost, and it says nothing about forecasts outside the data range.

Is expected value a good basis for a budget?

It is a useful single figure, but it hides the range of outcomes and relies on probabilities that are often subjective. It suits repeated decisions better than one-off ones. In your answer, mention the best and worst cases as well.

Can learning curves apply to every cost in the budget?

No. They apply to labour-intensive, repetitive tasks that are new or changing. They are less relevant to automated processes or mature production where workers have already reached a steady rate. Material and fixed costs are not directly affected by the curve.