Check the two months behind a high-low cost estimate

The high-low method estimates a variable cost rate from the difference between the highest and lowest activity observations. It then treats the remaining cost as fixed within the modeled range. The arithmetic is quick, but a one-time charge in either selected period can materially change both estimates.

Select the highest and lowest activity levels, not the highest and lowest costs. The method is trying to associate a cost difference with an activity difference. Choosing observations because their costs are extreme changes that question and can select periods dominated by events unrelated to volume.

OpenStax's managerial accounting text explains mixed-cost equations, the high-low method and the need to remain within a relevant operating range. Use the calculation as an initial model to inspect. It does not establish the physical cost behavior of every account included in the total.

Suppose a service operation records 1,000 orders and $18,000 of monthly support cost at its lowest activity, and 2,500 orders and $36,000 at its highest. The cost difference is $18,000 over 1,500 additional orders, giving a variable estimate of $12 per order.

Subtracting $12,000 from the low month's $18,000 gives a fixed estimate of $6,000. The resulting model is monthly cost = $6,000 + $12 × orders. At 1,800 orders, it predicts $27,600. Substitution into either endpoint reproduces that endpoint by construction, so that check alone does not validate the model.

Two middle observations show $22,000 at 1,500 orders and $26,000 at 2,000. The high-low line predicts $24,000 and $30,000 respectively, above both. Those differences provide a reason to investigate before using the model to approve the next operating budget.

Assume the highest-activity month includes a documented $6,000 one-time repair unrelated to order volume. Removing that charge from the recurring-cost dataset reduces the high endpoint to $30,000. The revised difference is $12,000 over 1,500 orders, giving $8 per order and $10,000 of fixed cost.

The revised prediction at 1,800 orders is $24,400, which is $3,200 below the original estimate. It also matches the two middle observations in this constructed example. Real observations rarely align so neatly, but the example shows how one endpoint charge can inflate the slope and depress the estimated fixed component simultaneously.

Retain the repair in the accounting records and in any cash period where it is payable. The adjustment changes the recurring-cost estimation dataset, not the fact that the business incurred the expense. Document why the event is nonrecurring and assess whether some maintenance allowance still belongs in future planning.

If activity barely changes across the observations, the estimated slope can be unstable because a small denominator magnifies unrelated cost differences. With identical high and low activity, the calculation is undefined. Do not manufacture a variable rate by selecting different cost observations at the same volume. Use contract terms, operational estimates or a better-designed data population to understand the cost, and label the uncertainty until the activity range provides useful evidence.

An order count may not explain support effort if some orders contain one item and others require extensive configuration. Changes in complexity or customer mix can look like a change in variable cost. Review the operational driver with the team doing the work, and consider separate populations where the work differs materially.

Match the cost period to the activity period. A supplier invoice covering two months cannot be compared directly with one month's orders without a consistent allocation or accrual. Payroll cutoffs, annual invoices and late adjustments can create apparent jumps unrelated to the underlying service.

Keep the measurement unit stable. A change from orders to order lines, from completed cases to opened cases, or from whole currency units to thousands can produce an incorrect slope without generating a spreadsheet error. Preserve the raw activity source and the definition used in every selected period.

A second supervisor, additional premises or a software tier can create a step in fixed spending after a threshold. If the lowest observation is below that threshold and the highest above it, the high-low slope spreads the step across every extra order. The resulting rate may overstate incremental cost below the step and understate the commitment required to cross it.

Build separate ranges where the operating plan supports them, or model the step explicitly. Ask when the commitment begins and whether it can reverse if volume falls. A capacity addition may remain payable for a contract term even after activity returns to its earlier range.

Changes in supplier prices also matter. Comparing a low-volume month at last year's rates with a high-volume month at current rates combines price and activity effects. Use supported adjustments or separate periods when possible. Do not erase actual inflation merely to force the points onto a cleaner line.

Plot or tabulate all available comparable observations, then inspect the largest unexplained differences. A regression can use more of the data, but it still needs a suitable driver, consistent periods and a defensible relationship. More sophisticated fitting does not fix a mixed population or a missing operating step.

Check the proposed forecast within the conditions represented by the data. The estimated $10,000 intercept in the adjusted example is not proof that cost would fall to $10,000 if orders stopped entirely. Zero activity may trigger a different operating state with its own staffing and contractual consequences.

Keep the selected endpoints, adjustment evidence, calculated rate, range and alternative estimate in the working papers. Give the budget owner a cost forecast with its assumptions, not a claim that fixed and variable cost have been permanently discovered. Update the estimate when a concrete change in the process or price structure makes the previous relationship unsuitable.