High-Low Method Calculator
The high-low method splits a mixed cost (one with both a fixed and a variable piece) into its fixed and variable components using only the highest and lowest activity levels observed. Use it to find the split from two data points, predict total cost at any activity level, or test the method against a full data set to see what it misses.
Background
Many real-world costs — electricity bills, maintenance, staffing — are mixed costs: part of the bill shows up no matter what (fixed), and part grows with usage (variable). The high-low method estimates that split with just two observations: the period with the highest activity and the period with the lowest. It's fast and needs no special tools, which is exactly why it's taught before regression — but that speed comes at a cost: every other data point you collected gets ignored.
How to use this calculator
- Choose Two-Point Split to turn a highest-activity period and a lowest-activity period into a fixed cost and a variable cost per unit.
- Choose Cost Predictor to use that same high/low data either direction: estimate total cost at a target activity level, or work backward from a budget to the maximum activity level it allows — and see whether the result is a safe interpolation or a risky extrapolation.
- Choose Full Data Set to paste in every period you have on record. The calculator finds the highest and lowest activity rows automatically, runs the high-low method on just those two, and then shows you how well (or badly) that line fits everything else you collected.
- Click Calculate to see the visual plus a full step-by-step explanation and a callout on what the result actually means.
How the high-low method works
Mixed costs contain a fixed piece (doesn't change with activity) and a variable piece (grows with activity). The high-low method estimates both using only the period with the highest activity and the period with the lowest.
The variable rate is the "rise over run" between those two points: the change in cost divided by the change in activity. It's exactly a slope.
Once you know the variable rate, the fixed cost falls out as whatever's left over at either point: total cost minus (rate × activity). Both points give the same answer, which is a useful check on your arithmetic.
You must pick high and low by activity level, not by cost. A period can have unusually high costs for reasons that have nothing to do with volume — picking by activity keeps the method tied to the cost driver you actually care about.
The method's biggest weakness is exactly what makes it fast: every period between the high and the low is thrown away. If those middle points don't sit near the high-low line, the split is a rough approximation, not a precise fit — that's what regression analysis is for.
Predictions made between the low and high activity levels (interpolation) are much more trustworthy than predictions made outside that range (extrapolation), since you have no evidence the cost stays linear beyond what you actually observed.
Formulas & Equations Used
Variable cost per unit: b = (Cost_high − Cost_low) / (Activity_high − Activity_low)
Fixed cost: a = Cost_high − b × Activity_high (equivalently a = Cost_low − b × Activity_low)
Cost equation: Total Cost = a + b × Activity
Solving for activity level (e.g. from a budget): Activity = (Total Cost − a) / b
Example Problems & Step-by-Step Solutions
Example 1 — The basic split
An electric bill was \$1,400 in the slowest month (400 kWh) and \$2,200 in the busiest month (1,000 kWh).
Step: b = (2,200 − 1,400) / (1,000 − 400) = 800 / 600. a = 2,200 − 1.333 × 1,000.
Result: Variable rate = \(1.333/kWh, fixed cost = \)866.67/month.
Example 2 — Predicting a new cost
Using the same bill, what would 700 kWh cost?
Step: Total = 866.67 + 1.333 × 700 = 866.67 + 933.33.
Result: Predicted cost = \$1,800.00 — and since 700 sits between 400 and 1,000, this is a safe interpolation.
Example 3 — The extrapolation trap
A maintenance line was built from 200 flight hours (\$65,000) and 500 flight hours (\(125,000): a = \)25,000, b = \$200/hour.
Step: Predicting at 5,000 hours — ten times the highest hour ever observed — gives 25,000 + 200 × 5,000.
Result: The formula returns \$1,025,000, but nothing in the data supports the line staying straight that far out. Treat it as a guess, not a forecast.
Example 4 — What the method throws away
A coffee shop logs daily cups sold vs. total cost for six days: (50,\$180), (80,\$210), (120,\$260), (150,\$290), (200,\$340), (300,\$430).
Step: Using only the low (50 cups) and high (300 cups) days: b = (430−180)/(300−50) = \(1.00/cup, a = \)130.
Result: The line predicts \(250 at 120 cups, but the shop actually spent \)260 that day — a \$10 gap the high-low method never sees, because it never looks at that row.
Example 5 — Solving backward from a budget
A cost structure of a = \(12,000 fixed, b = \)4/unit variable. How many units fit inside a \$25,000 budget?
Step: Rearrange the cost equation: X = (25,000 − 12,000) / 4 = 13,000 / 4.
Result: Up to 3,250 units before the budget is exhausted.
Example 6 — Order doesn't matter, activity does
A bakery's flour cost was \$3,200 for 5,000 loaves and \$1,400 for 1,500 loaves — entered in whichever order is convenient.
Step: The calculator sorts by activity level first (1,500 is low, 5,000 is high) rather than trusting which box each number was typed into.
Result: b ≈ \(0.514/loaf, a ≈ \)628.57 — the fixed cost that exists even on a day the ovens never turn on.
Frequently Asked Questions
Why use the highest and lowest activity, not the highest and lowest cost?
Activity level is the cost driver you're trying to measure the effect of. A period could have an unusually high cost for reasons unrelated to volume (a one-time repair, a price spike) — anchoring on activity keeps the estimate tied to the relationship you actually want.
What's the biggest weakness of the high-low method?
It only ever looks at two data points. Every other period you recorded — no matter how many — gets completely ignored, so a single unusual high or low period can distort the entire estimate. Regression analysis uses every point and is more reliable when you have the data for it.
Can the fixed cost come out negative?
Yes, and it's a red flag rather than a valid answer. A negative fixed cost usually means the relationship isn't actually linear across the range you picked, or that your high/low points aren't representative — it's worth checking the data rather than trusting the number.
Is it safe to predict costs far outside my highest and lowest observed activity?
No. The straight line is only supported by evidence between your low and high points. Outside that range you're assuming the same linear relationship continues, which is often untrue — capacity limits, bulk discounts, and overtime pay all tend to bend the line at the extremes.
How is this different from a scatter graph or regression?
A scatter graph plots every period and regression fits a line that minimizes error across all of them. The high-low method is a shortcut that skips straight to a line through just two specific points — quicker with pen and paper, but less accurate whenever the middle points don't sit near that line.
What does the fixed cost actually represent in real life?
It's the estimated cost you'd still pay at zero activity — rent, salaried staff, equipment depreciation, and similar costs that don't disappear just because volume drops to nothing. It's an estimate, not a guarantee, since some "fixed" costs do eventually step up or down at extreme volumes.
All calculators