Most real costs are neither purely fixed nor purely variable. A utility bill has a standing charge and a usage charge. A maintenance budget has a base contract and a per-hour element. A delivery cost has a fleet cost and a per-drop cost. The high-low method is the simplest technique for splitting such a mixed cost into its two components, using nothing but the highest and lowest activity observations available, and it needs no software beyond arithmetic.
Arb Digital built this calculator to make the split explicit rather than assumed. Almost every budgeting, break-even and pricing exercise starts by assuming a fixed and variable split already exists, and very often nobody has actually derived it. This page derives it, shows the resulting cost equation, and forecasts the cost at any activity level you choose.
What This High-Low Method Calculator Does
Enter the highest activity level observed and its total cost, then the lowest activity level and its total cost. The hero figure is the variable cost per unit of activity. The supporting grid gives the fixed cost per period implied by that rate, the total cost forecast at whatever activity level you enter, the average cost per unit at that forecast level, and the proportion of the high-point total cost that the variable element accounts for.
The sub-line under the headline number states the resulting cost equation in full, in the form fixed cost plus variable rate times activity. That equation is the actual deliverable of a high-low exercise. It is what feeds a flexible budget, a break-even calculation or a make-or-buy decision, and it is worth writing down rather than leaving inside a spreadsheet cell.
How to Use It
- Choose a cost driver — units produced, machine hours, labour hours, deliveries, transactions — and make sure both observations measure the same one.
- Find the highest and lowest activity periods in your data. Select on activity, never on cost.
- Enter the total mixed cost for each of those two periods, taking the whole cost rather than an already-split portion of it.
- Enter the activity level you want to forecast, keeping it inside the range between your two observations.
- Press Calculate to see the variable rate, the fixed element, the cost equation and the forecast update together.
The Formula and How It Is Calculated
The method treats the two observations as two points on a straight line and solves for its slope and intercept. The slope is the variable cost per unit: Variable cost per unit = (highest activity cost − lowest activity cost) ÷ (highest activity units − lowest activity units). The intercept is the fixed cost: Fixed cost = total cost at either point − (variable cost per unit × activity at that point). Either point gives the same answer, which is a useful self-check.
Work the default figures through. The cost difference is 68,000 − 40,500 = 27,500. The activity difference is 12,000 − 5,000 = 7,000 units. The variable rate is 27,500 ÷ 7,000 = 3.9286 per unit. Taking the high point, fixed cost is 68,000 − (3.9286 × 12,000) = 68,000 − 47,142.86 = 20,857.14. Checking at the low point: 40,500 − (3.9286 × 5,000) = 40,500 − 19,642.86 = 20,857.14, identical as it must be.
The cost equation is therefore total cost = 20,857.14 + 3.9286 × units. Forecasting 9,000 units gives 20,857.14 + 35,357.14 = 56,214.29, which is an average of 6.2460 per unit. Notice how far that average sits from the variable rate of 3.9286 — the gap is the fixed cost being spread, and it is the entire reason average cost per unit falls as volume rises.
Select on Activity, Not on Cost
The most frequent error in applying this method is picking the periods with the highest and lowest total cost rather than the highest and lowest activity. In a well-behaved dataset these coincide. In a real one they often do not, because a period of moderate activity can carry an unusual cost — a one-off repair, a bonus accrual, a price increase — and selecting on cost drags that anomaly straight into the slope.
The method has no defence against this, because it uses only two points. There is no averaging, no residual, nothing to dilute a bad observation. That is the whole trade-off: the high-low method is transparent enough to do on paper and explain in a meeting, and fragile enough that one odd month can distort the answer materially. Plot the data before trusting the result, and if either endpoint looks unusual, ask why before using it.
A practical safeguard is to run the method a second time using the second-highest and second-lowest activity periods. If the variable rate barely moves, the estimate is stable. If it moves a lot, one of the endpoints is driving the answer and the split needs a better technique.
Why Regression Usually Beats It
Least-squares regression fits a line through every observation rather than two, which means an unusual month is one data point among many instead of half the input. It also produces a measure of fit, so you learn whether the cost genuinely behaves linearly with the chosen driver or whether the relationship is weak. Our linear regression calculator does exactly that, and the resulting slope and intercept are read the same way: slope is the variable rate, intercept is the fixed cost.
So when is high-low still the right tool? When you have very few observations, when you need an answer in a meeting, when the audience must be able to follow the arithmetic, or when the point is to sanity-check a regression output rather than replace it. It is also genuinely useful as a teaching device, because the two-point version makes the structure of a mixed cost obvious in a way a regression output does not.
The Relevant Range Is a Real Constraint
Cost behaviour is linear only within a band of activity, and that band is called the relevant range. Outside it, fixed costs step. Doubling output may require a second shift, another supervisor, more floor space or another machine, and none of those arrive gradually. The straight line the high-low method fits carries no information about where those steps sit, because both observations were taken inside the range.
The practical consequence is that forecasting far outside the observed range is not conservative, it is wrong. A forecast at 25,000 units built from observations at 5,000 and 12,000 assumes fixed costs that a business operating at 25,000 units almost certainly would not have. The calculator will produce the number because the arithmetic is defined, but the number carries an assumption the data never tested. Keep the forecast between the two endpoints, and treat anything beyond them as a scenario requiring separate evidence.
Variable rates are not immune either. Volume discounts, overtime premiums, learning effects and capacity constraints all bend what looks like a constant per-unit rate. Under IAS 2 Inventories, fixed production overheads are allocated to conversion cost based on normal capacity rather than actual output precisely because the relationship between volume and cost is not the simple proportion a two-point fit implies.
What You Do With the Split Once You Have It
A fixed and variable split is an input, not a conclusion. Break-even analysis needs it: our break-even calculator takes the split as given and solves for the volume at which contribution covers fixed cost, which is exactly the number the high-low method supplies. The contribution margin calculator works the same territory from the revenue side, and the margin of safety calculator measures the cushion above break-even.
The split also underpins flexible budgeting. A static budget compares actual cost at actual volume against a plan built at planned volume, which conflates a volume difference with a spending difference. A flexible budget rebuilds the plan at actual volume using the cost equation, isolating the part of the difference that is genuinely about spending. That is the same logic that separates a price variance from a quantity variance in standard costing, which our material price variance calculator covers, and the marginal cost calculator handles the incremental view.
OpenStax's Principles of Accounting, Volume 2: Managerial Accounting is a good free reference for cost behaviour, the relevant range and where the high-low method sits alongside scatter graphs and regression.
Arb Digital builds free calculators and explainers that answer a real question properly — the approach behind every tool on this site.
Content Marketing Services Talk to Arb DigitalCommon Mistakes to Avoid
- Selecting the highest and lowest cost periods instead of the highest and lowest activity periods. Select on the driver, always.
- Mixing cost drivers between the two observations — units in one period and machine hours in the other produces a slope that means nothing.
- Extrapolating outside the relevant range. Fixed costs step, and a two-point fit has no way of knowing where the steps are.
- Keeping an anomalous endpoint. A one-off repair or a bonus accrual in either observation goes directly into the variable rate, because there is nothing to average it against.
- Reading the average cost per unit as the variable cost. The average includes spread fixed cost and changes with every volume; only the slope is the variable rate.
Related Free Tools From Arb Digital
Fit a line through all your observations with the linear regression calculator, then take the split into the break-even calculator, the contribution margin calculator and the margin of safety calculator. The marginal cost calculator covers incremental cost, the material price variance calculator covers standard costing variances, and the free tools hub lists the rest.
Frequently Asked Questions
It is a technique for splitting a mixed cost into fixed and variable components using only the highest and lowest activity observations. The difference in cost divided by the difference in activity gives the variable rate per unit, and subtracting the variable element from either total gives the fixed cost.
Always the highest activity. Selecting on cost pulls anomalies into the calculation, because a moderate-activity period can carry an unusual one-off cost. Since the method uses only two points, there is nothing to dilute that distortion.
Because the two observations define a single straight line, and the fixed cost is that line's intercept. Computing it from either endpoint must give the same value, which makes it a useful arithmetic check on your working.
Usually, yes. Regression uses every observation rather than two, so a single unusual period has far less influence, and it reports how well the line actually fits. The high-low method wins on speed and on being explainable without software, which is why it survives.
It is the band of activity within which cost behaviour stays approximately linear. Outside it, fixed costs step as extra shifts, supervisors, space or machinery become necessary. A two-point fit cannot see those steps, so forecasts far outside the observed range are unreliable.
Arithmetically yes, if the higher-activity period recorded a lower total cost. That is a signal that something other than volume is driving the cost, or that one observation is contaminated. A negative variable rate should never be used in a forecast without investigating the underlying data.
Because the average includes the fixed cost spread across the units. As volume rises, the same fixed cost is divided by more units and the average falls toward the variable rate, without the variable rate itself changing at all.
This tool is provided for educational and planning use only and is not accounting or financial advice. The high-low method uses two observations and assumes linear cost behaviour within a relevant range, so its output is an estimate rather than a measured cost. A qualified accountant should confirm any figures used for reporting or decision-making.
