Lena Voss had a problem she kept describing as a sales problem. Redstone Machine, her 14-person shop in Erie, Pennsylvania, was winning nearly every CNC bracket job it bid on and losing nearly every weldment assembly job. Her estimator assumed the competition had gotten aggressive on fabrication work.
The shop was busier than it had ever been. But net income had fallen for three straight quarters.
Nothing was wrong with the sales team. The quotes were being built on a costing method that had quietly stopped matching how the shop actually operated, and it was pushing every price in the wrong direction at once.
Overhead has to land somewhere
Direct materials and direct labor are easy. Steel gets bought for a specific job, and a machinist's hours get logged against a work order. Both trace cleanly to the product.
Manufacturing overhead does not. Shop rent, equipment depreciation, utilities, maintenance, the supervisor's salary, indirect supplies, and shop insurance are all incurred to keep the shop running, and none of them can be assigned to a single job. Redstone's overhead pool runs $52,000 a month.
This money still has to reach the product because, under GAAP, manufacturing overhead is a product cost that gets absorbed into inventory and then into cost of goods sold. Companies do this with a predetermined overhead rate calculated before the year starts.
Predetermined overhead rate = Estimated total overhead ÷ Estimated total units of the allocation base
Applied overhead = Predetermined rate × Actual units of the base used by the job
Total job cost = Direct materials + Direct labor + Applied overhead
The allocation base is the driver you assume causes overhead to be consumed. Redstone used direct labor hours, the most common base in American job shops, because that is what the shop's first bookkeeper set up in 1998.
At 2,600 direct labor hours a month, the rate is $52,000 ÷ 2,600, or $20 per direct labor hour.
Labor does not depreciate
That single plantwide rate assumes that a job consuming 10% of the shop's labor also consumes 10% of the shop's rent, depreciation, and maintenance. Redstone bought two CNC machining centers in 2023. Those machines carry most of the depreciation and most of the maintenance in the overhead pool, and they replaced people.
So the CNC work now runs enormous machine hours against very few labor hours. The weldment work runs the opposite way, using older equipment that is nearly fully depreciated plus a large number of hands. Under a labor-hour base, the labor-heavy jobs absorb the depreciation on machines they barely touch.
Splitting the pool by what actually drives each cost fixes it. Redstone's $31,000 of depreciation, maintenance, and machine utilities is driven by machine hours (1,900 a month). The remaining $21,000 of supervision, indirect labor, and shop supplies is driven by labor hours.
Machine-related rate = $31,000 ÷ 1,900 = $16.32 per machine hour
Labor-related rate = $21,000 ÷ 2,600 = $8.08 per direct labor hour
The same two jobs, costed both ways
Redstone is quoted at cost plus 25%. The CNC job went out at $4,950 against a true cost of $6,540, losing $1,590 every time it ran. The weldment job went out at $10,050 against a true cost of $6,854, which is why competitors kept taking it.
Lena was winning the losing work and losing the winning work, and the harder her estimator sold, the faster margin drained.
Why bad allocation also changes behavior
Suppose four people are splitting a restaurant check evenly. The person who ordered a salad subsidizes the person who ordered the steak, and once everyone understands the rule, everyone orders the steak.
Cost allocation works the same way inside a company. When a department or a product line gets charged for a slice of a pool it barely touches, the cheap-looking option gets ordered more, and the pool grows.
Redstone saw this with programming time. CNC jobs required hours of setup from a programmer whose salary sat in the general overhead pool, so the sales team treated programming as free and quoted short-run CNC work aggressively. Charging that time out by the number of setups ended the practice within a quarter.
The markup rule of thumb that hides the damage
Plenty of contractors and shops skip allocation entirely and add a flat 10% for overhead and 10% for profit on top of direct costs. That shortcut assumes overhead scales with the size of a job, and it treats a simple repetitive order and a custom low-volume order as if they burden the office identically.
They rarely do. A custom job pulls engineering hours, extra purchase orders, more inspection, more scheduling attention, and more phone calls, while a repeat production run mostly leaves the office alone. A flat percentage overcharges the simple work, which loses bids that would have been profitable, and undercharges the complex work, which wins bids that quietly drain cash.
A single plantwide rate is fine in two situations: overhead is small relative to direct costs, or every job consumes resources in roughly the same proportions. Once automation pushed overhead above half of total product cost in most manufacturing operations, neither condition held for long.
Two rules for keeping this honest
Allocation never changes total company profit. It only changes which product appears profitable, which is precisely why it changes every pricing decision downstream.
Rates get set on estimates, so the applied amount rarely matches the actual amount spent. At period end, compare the two:
Over- or underapplied overhead = Applied overhead − Actual overhead incurred
Applying less than was spent means jobs were undercosted and probably underpriced. Redstone had been underapplying by roughly $4,000 a month, a number that had been sitting in the closing entry for two years without anyone asking what it meant.
Fixing this issue does not require rebuilding a costing system. Take the three largest overhead pools, name the single best driver for each, and rerun last quarter's jobs with those drivers. If one base is currently carrying all three pools, a more accurate answer is already sitting in the data. Ask which pools are large, what genuinely drives each one, and whether the rate has been reviewed since the equipment changed. Lena rebid the same customer list at corrected costs, lost most of the CNC work, and won back the fabrication jobs. Revenue fell 9%. Gross profit rose 22%.
