How this instrument works
A mixed cost — a utility bill, equipment maintenance, a shipping account — carries both a fixed floor and a variable rate that scales with activity, blended into one total that arrives on the invoice without a breakdown. The high-low method separates the two using only the two most extreme observations in a cost history: whichever period ran the most activity and whichever ran the least. Because activity moved by a known amount between those two periods and cost moved with it, the change in cost divided by the change in activity isolates the variable rate; whatever cost is left over at either point once that rate is backed out is the fixed component.
The appeal is that it needs no software and no data set beyond two numbers pulled from a ledger, which is why it survives as the first technique taught in every cost accounting course and the one a controller reaches for when a flexible budget is due before a full regression can be run. A scatter-graph or least-squares regression uses every period in the history and is less swayed by any single unusual month, but both take longer to set up and neither fits on the back of an envelope. The trade the high-low method makes is speed and simplicity against precision, and that trade is only sound if the two chosen periods behaved normally.
The whole result rides on picking the right two points, and the picking rule is activity, not cost — the high period is whichever one ran the most machine hours, orders, or units, even if a different period happened to post a bigger dollar total from a one-off repair bill or a rate hike. A period distorted by overtime premiums, a rush surcharge, or an idle shutdown will drag the whole split off if it lands on either end, since the method has no way to tell a representative month from an outlier once only two are chosen.
- From your cost history, find the period with the most activity and enter its total spending into Total cost at highest activity, $ and its volume into Highest activity level (units).
- Find the period with the least activity and enter the same two figures into Total cost at lowest activity, $ and Lowest activity level (units).
- Read Variable cost per unit — the rate that cost changes by for each added unit of activity.
- Read Estimated fixed cost — the portion of the total that holds steady regardless of activity.
- Check that both chosen periods were ordinary ones before trusting the split; a repair bill, strike, or rate change in either period will carry straight into the answer.
Worked example — $80,000 against $50,000
A plant's overhead account shows two extremes across a year of records. Its busiest month ran 10,000 machine hours and posted a Total cost at highest activity, $ of 80,000. Its quietest month ran 4,000 machine hours, entered as Total cost at lowest activity, $ of 50,000 against a Lowest activity level (units) of 4,000. The gap between them is $30,000 of cost over 6,000 hours of activity, so Variable cost per unit comes out to 30,000 ÷ 6,000 = $5.00 per hour.
Backing that rate out of either period gives the fixed piece: 80,000 minus 5.00 times 10,000 leaves an Estimated fixed cost of $30,000 sitting under the account no matter how many hours the plant runs. A controller now has both pieces needed to forecast an untested month — at 7,000 hours, projected overhead is 30,000 + 5.00 × 7,000 = $65,000 — without waiting on a full regression of the whole year's data.
Questions
Why do 'high' and 'low' mean activity, not dollar cost?
Because the method is solving for how cost responds to activity, and that only works if activity is what defines the two endpoints. A period with a smaller dollar total than another can still be the correct 'high' point if it ran more machine hours, orders, or units — picking by cost instead of activity mixes up cause and effect and produces a variable rate that describes nothing real.
Who actually reaches for the high-low method?
A management or cost accountant building a flexible budget who needs a quick fixed-variable split before a full data run is ready, or a plant controller pricing a rush order who wants to know how much of the next batch's cost is truly incremental. It is also the standard first technique in a managerial accounting course, taught before scatter-graph and regression methods that need more than two data points.
How is this different from a regression or scatter-graph estimate?
Regression fits a line through every period in the cost history, so one strange month gets averaged against the rest and barely moves the answer. The high-low method uses exactly two periods and ignores everything between them, which makes it fast enough for a scratch pad but leaves the result fully exposed if either chosen period was not typical.
What if my highest-activity period wasn't a normal one?
The split inherits whatever distortion that period carried. A busiest month padded with overtime premiums or a rush surcharge inflates the variable rate and understates the fixed piece, since the formula has no way to separate a genuine cost driver from a one-off charge once only two points are in play. Rerun the calculation using the next most extreme normal period instead.
Can the estimated fixed cost come out negative?
Yes, if the underlying cost is not truly linear across the range between the two chosen periods or if one of them was unusually cheap. A negative fixed cost signals the two-point split has broken down, not that the account genuinely earns money by sitting idle — treat it as a flag to check the data or switch to a method using more observations.
Does this replace a full cost accounting system?
No. It produces one estimate from two data points to unblock a budget or a pricing decision quickly, not an audited cost breakdown. A business with accounting software and a full period's data available should run a regression across the whole history instead, and treat the high-low figure as a rough check against that more complete result.
References
Read this first: This instrument shows arithmetic, not advice. Real offers add fees, taxes and terms that vary by lender and place — verify the figures against your actual paperwork before deciding anything.