Mixed Cost Estimation: Three Methods

Table of Contents

    Mixed cost estimation helps a manager turn past bills into a usable planning equation. If machinery utilities and maintenance rise when a workshop runs longer, how much of next month’s cost should change with machine hours, and how much should remain in the monthly base? A total bill alone does not answer that question.

    This guide compares scattergraph inspection, the high-low method and ordinary least squares regression using one hypothetical six-month dataset. It then carries the estimated components into a break-even calculation. All business figures are invented U.S. dollars, and the examples assume a stable operating arrangement unless a limitation is stated.

    What Is Mixed Cost Estimation?

    A mixed cost contains a fixed component and an activity-related component. Estimation uses observed activity and cost to infer those components. This differs from applying a known contract, such as a stated monthly service charge plus a stated usage rate. Managerial accounting explains that cost-behavior classification; here, the components are initially unknown.[5]

    Estimated Mixed-Cost Equation
    Estimated total cost = a + b × x
    a is the estimated fixed component for the period; b is the estimated variable rate per activity unit; x is activity in that period.

    For a monthly machinery-cost pool, a is dollars per month, b is dollars per machine hour and x is that month’s machine hours. The rate multiplied by hours gives the month’s activity-related dollars. Actual cost can differ from the fitted equation because the simple model leaves other influences unexplained.[1][2]

    An Estimate Is Not a Known Tariff

    A fitted base and rate summarize the chosen observations. They do not establish a contractual minimum bill, prove which costs can be avoided, or guarantee next month’s spending.

    Start With Comparable Monthly Data

    Consider a fictional workshop whose manager records machinery utilities and maintenance as one cost pool. Each row pairs costs incurred during a month with machine hours used during that same month. Assume the workshop initially treats the equipment, tariff and included cost categories as comparable across the six months.

    Hypothetical monthly machinery utilities and maintenance
    Month Machine hours Cost incurred (USD)
    January 1,000 $4,000
    February 2,000 $6,000
    March 3,000 $7,500
    April 4,000 $10,000
    May 5,000 $13,500
    June 6,000 $13,000

    The highest bill occurs in May, while the highest activity occurs in June. Keep that distinction: it determines which observations belong in the high-low calculation. Six invented months keep the arithmetic visible; they are not a recommendation about how much data a real forecast needs.

    Before estimating, reconcile the cost pool to its records. Match service periods rather than pairing an old invoice’s payment date with this month’s hours. Check for changed tariffs, equipment additions, unusual repairs and inconsistent classifications. Machine hours should also have an operating connection to the costs: a single driver may explain machine wear better than weather-sensitive building utilities.[5]

    Inspect the Scattergraph Before Fitting a Line

    A scattergraph places machine hours on the horizontal axis and monthly cost in USD on the vertical axis. Each month’s pair becomes one point. Look for a roughly straight pattern, a curve, separate groups, changing spread or an observation far from the rest. These are questions about whether one straight-line cost equation is a useful starting point.[4]

    The six pairs above show a broadly upward relationship, but May costs more than June despite using fewer hours. That reversal deserves investigation; it does not automatically disqualify a linear approximation. A hand-drawn trend line can suggest a base and rate, but different analysts can draw different lines. Use it as a diagnostic and rough estimate, then apply a stated numerical method.[1]

    An upward pattern also does not prove that every increase in cost was caused by machine hours. Seasonal energy prices or a maintenance schedule could move alongside activity. The separate guide to correlation explains why association needs interpretation.

    High-Low Method: Use the Activity Extremes

    The high-low method fits a line through the observations with the highest and lowest activity, using their associated costs. Select June’s 6,000 hours and $13,000 cost, and January’s 1,000 hours and $4,000 cost. Do not substitute May merely because its bill is largest.[1]

    1. Estimate the variable rate

    High-Low Variable Rate
    b = (Cost at high activity − Cost at low activity) / (High activity − Low activity)
    The change in cost divided by the change in the activity driving the model.

    For the workshop, b = ($13,000 − $4,000) / (6,000 − 1,000) = $9,000 / 5,000 = $1.80 per machine hour.

    2. Estimate the fixed component

    Subtract the estimated variable portion from either selected total:

    Recover the Same Base From Either Endpoint

    Using January: a = $4,000 − ($1.80 × 1,000) = $2,200 per month.

    Using June: a = $13,000 − ($1.80 × 6,000) = $2,200 per month.

    The high-low equation is therefore Estimated monthly cost = $2,200 + ($1.80 × machine hours). It exactly matches those two observations by construction. That agreement does not test how well it describes the other four months.

    Regression: Estimate the Base and Rate From All Six Months

    Ordinary least squares (OLS) chooses the intercept and slope that minimize the sum of squared vertical residuals across all included observations. A residual is observed cost minus fitted cost at the same activity. Squaring prevents positive and negative misses from canceling and gives larger misses more weight.[2]

    For the straight-line model with an intercept, let x̄ and ȳ denote the sample means:

    OLS Cost Rate and Intercept
    b = Σ[(xi − x̄) × (yi − ȳ)] / Σ[(xi − x̄)2]
    a = ȳ − b × x̄
    Calculate both sums over the same paired observations. The activity values must vary so the denominator is positive.

    In our six months, x̄ = 21,000 / 6 = 3,500 hours and ȳ = $54,000 / 6 = $9,000. January contributes (1,000 − 3,500) × ($4,000 − $9,000) = 12,500,000 to the numerator and (1,000 − 3,500)2 = 6,250,000 to the denominator.

    Adding the six centered cross-products gives 35,000,000; adding the six squared activity deviations gives 17,500,000. Thus b = 35,000,000 / 17,500,000 = $2.00 per machine hour, and a = $9,000 − ($2.00 × 3,500) = $2,000 per month.

    OLS Equation for the Hypothetical Workshop
    Estimated monthly cost = $2,000 + ($2.00 × machine hours)
    The fitted slope and intercept describe this sample; they are not known contractual charges.

    The numerator has dollar-hours units and the denominator has hours-squared units, leaving dollars per hour. The intercept returns to monthly dollars. For the broader statistical framework and assumptions behind coefficient interpretation, see simple linear regression.

    Check the misses, not just the coefficients

    Keep all six original observations and compare the two fitted equations. For May, OLS predicts $2,000 + ($2.00 × 5,000) = $12,000, leaving a residual of $13,500 − $12,000 = +$1,500. High-low predicts $2,200 + ($1.80 × 5,000) = $11,200, leaving $13,500 − $11,200 = +$2,300.

    Observed minus fitted cost: hypothetical monthly residuals (USD)
    Month High-low residual OLS residual
    January $0 $0
    February +$200 $0
    March −$100 −$500
    April +$600 $0
    May +$2,300 +$1,500
    June $0 −$1,000

    OLS has a squared-residual total of 5002 + 1,5002 + 1,0002 = 3,500,000 USD2. High-low’s total is 2002 + 1002 + 6002 + 2,3002 = 5,700,000 USD2. OLS fits better by this all-sample criterion even though high-low fits June exactly and OLS misses it by $1,000.

    At a planned 4,500 hours, high-low estimates $2,200 + ($1.80 × 4,500) = $10,300, while OLS estimates $2,000 + ($2.00 × 4,500) = $11,000. The $700 difference shows that choosing an estimator changes the budget. It does not tell us which estimate will be closer to an unknown future bill.

    Scattergraph vs High-Low vs Regression

    What each mixed-cost estimation method contributes
    Method Main use Main caution
    Scattergraph inspection See shape, unusual observations and whether one line is plausible. A hand-fitted line is subjective; a visible relationship does not prove cause.
    High-low Obtain a quick, checkable rate and base from activity extremes. Uses only two observations; unusual endpoints can dominate.
    OLS regression Use all included observations to minimize squared residuals. Can be distorted by outliers, omitted drivers or a changing process.

    These methods can complement one another: inspect first, use high-low as a rough comparison, and evaluate a regression fit and its residuals. OLS’s smaller in-sample squared error is a mathematical property, not a guarantee of better forecasting or a substitute for understanding the operation.[2][3]

    Limitations: Relevant Range, Intercepts and Outliers

    The observed activity spans 1,000 to 6,000 machine hours per month. Treat that as the data’s coverage, not automatic proof that one operating regime applies everywhere inside it. Check when extra shifts, new equipment, overtime or tariff changes would alter the relationship. Costs of production explains why broader cost curves need not be straight lines.[5]

    The Intercept Is Not a Verified Shutdown Bill

    OLS’s $2,000 intercept is the fitted line’s value at zero hours. Zero lies outside these observations. Calling it an estimated fixed component within the planning model does not establish what the workshop would actually pay during a shutdown. Contract terms, standby costs and canceled services require separate evidence.

    Investigate unusual observations. May’s +$1,500 OLS residual could prompt a review of repair records, energy use and invoice periods. It does not prove that a one-time repair occurred. Correct an error when supported; separate a genuinely different cost when the model’s purpose requires it; otherwise retain valid observations and explain any sensitivity analysis. Never remove a month merely to improve the fit.

    More data do not repair a mismatched model. A longer sample may mix old and new equipment or pricing. Monthly errors can also persist through time or grow with activity, affecting statistical uncertainty. A second driver or a different functional form may be needed. NIST identifies outlier sensitivity and extrapolation as key limitations of linear least squares.[3]

    Feed the Estimated Costs Into Break-Even

    Continue with the same fictional workshop. For this planning extension, assume it sells one standard product requiring exactly one machine hour per unit, sells everything produced, and maintains a $20 unit price. Other variable costs are $12 per unit and other fixed costs are $10,000 per month. Both exclude the entire utilities-and-maintenance pool already estimated.

    Using OLS, convert the pool’s activity rate to a product rate: $2 per hour × 1 hour per unit = $2 per unit. Total variable cost is $12 + $2 = $14 per unit, and total fixed cost is $10,000 + $2,000 = $12,000 per month.

    Conditional Break-Even Units
    Break-even units = Total fixed costs / (Price per unit − Total variable cost per unit)
    Use positive contribution, matched periods and a cost model valid at the resulting volume.

    The result is $12,000 / ($20 − $14) = 2,000 units, corresponding to 2,000 machine hours within the observed activity span. The check is (2,000 × $20) − (2,000 × $14) − $12,000 = $0 operating profit.[6]

    High-low instead gives total fixed costs of $12,200 and variable cost of $13.80 per unit: $12,200 / ($20 − $13.80) = 1,967.74 units, rounded up to 1,968 whole units to avoid a modeled loss. That volume produces 1,968 × $6.20 − $12,200 = $1.60 of modeled operating profit. Different cost estimates therefore produce different sales thresholds.

    Do not add the full estimated mixed-cost bill again after incorporating its base and rate. If each product required a different number of hours, convert the rate accordingly before combining costs. See break-even analysis for contribution, target profit and sales-mix considerations; this estimation exercise does not establish achievable demand or cash sufficiency.

    How to Use a Mixed-Cost Estimate

    1. Define the decision and cost boundary. Identify the period, included costs, activity measure and operating conditions.
    2. Inspect the observations. Reconcile records, examine the activity-cost pattern and investigate unusual months.
    3. Fit and compare. Calculate high-low from activity extremes, then evaluate an all-data regression and its residuals.
    4. Check the intended volume. Confirm the forecast fits the operating range and update rates or capacity assumptions when necessary.
    5. Monitor subsequent results. Compare actual spending with the allowance at actual activity. Variance analysis helps investigate differences without confusing lower volume with better cost control.

    Common Mistakes in Mixed Cost Estimation

    • Selecting the largest and smallest bills. High-low selects the highest and lowest activity and their corresponding costs.
    • Dividing total cost by hours and calling it the variable rate. That average includes the fixed component spread across activity.
    • Comparing different data samples without saying so. Keep observations consistent when attributing fit differences to methods.
    • Equating the intercept with avoidable spending. A fitted fixed component does not identify what disappears after a decision.
    • Deleting inconvenient points. Investigate and document errors or model-boundary changes rather than choosing the preferred answer.
    • Mixing machine hours with units sold. Convert the activity rate using the product’s resource requirement before applying a unit contribution formula.

    Frequently Asked Questions

    It estimates the change in cost across the widest observed activity interval. In this example, June has the most hours even though May has the largest bill. Choosing May because of its cost would use different endpoints from the high-low method specified here.

    OLS minimizes total squared residuals for the included sample and straight-line specification with an intercept. That does not guarantee the smallest error on every observation or the best future forecast. Data quality, unusual points, driver choice and operating changes still matter.

    Not necessarily. It is the fitted line’s extension to zero activity, which may be outside the observed range. Review the driver, cost boundary, unusual points and functional form before assigning an economic meaning. Do not silently replace the intercept with zero; a constrained fit is a different estimation problem.

    Only after matching units. Multiply dollars per machine hour by machine hours per product to obtain dollars per product, then add other variable costs without double counting. Also add other fixed costs to the estimated pool base and confirm the cost assumptions apply at the resulting sales volume.

    Sources

    Primary sources accessed November 14, 2026. All workshop observations, fitted results and planning assumptions are hypothetical. The article concerns internal cost planning, not a prescribed tax or financial-reporting treatment.

    1. OpenStax, Estimate a Variable and Fixed Cost Equation and Predict Future Costs, Principles of Accounting, Volume 2: Managerial Accounting, section 2.3; cost equations, scattergraphs and activity-based high-low. Book published February 14, 2019.
    2. NIST/SEMATECH, Least Squares, e-Handbook of Statistical Methods, section 4.4.3.1; objective, straight-line slope/intercept and estimated versus true parameters.
    3. NIST/SEMATECH, Linear Least Squares Regression, section 4.1.4.1; limitations, extrapolation and outlier sensitivity.
    4. NIST/SEMATECH, Scatter Plot, section 1.3.3.26; diagnostic questions and association versus causation.
    5. OpenStax, Identify and Apply Basic Cost Behavior Patterns, section 2.2; cost drivers, mixed costs and stepped costs.
    6. OpenStax, Calculate a Break-Even Point in Units and Dollars, section 3.2; contribution, fixed-cost recovery and CVP assumptions.
    Disclaimer

    This article is for educational and informational purposes and does not constitute accounting, tax or investment advice. Its hypothetical observations simplify cost behavior and do not establish future spending, profitability or cash needs. Use reliable business records, appropriate operating assumptions and qualified professional advice for a specific decision.