The Variance Lines That Matter
A fabrication job has predictable places where actuals diverge from estimates. Compare each, not just the total, so you know what to fix.
- Material: sheets consumed versus quoted utilization, and price paid versus quoted price
- Laser: actual beam-on and handling minutes versus estimated
- Press brake: setup and run hours versus estimate, especially on small lots
- Welding and finishing labor, which is the most variable and least measured
- Outside processing cost and the rework or rejects behind it
- Scrap, remakes and engineering changes after release
Why Labor Data Is Often the Weak Link
Material is easy to measure because it comes from purchase orders and issues. Labor is harder. If operators log time at the end of the day, or a job moves between work centers without a scan, the actual hours are approximate.
Welding and manual finishing are the worst offenders, which is a problem because they often carry the biggest estimating risk. Improve capture with simple start and stop at each operation, shop floor terminals or tablets, and a rule that time is entered on the job, not on a general bucket. Perfect data is not required, but consistent data is, or variance reports will point at the wrong problems.
Costing Method Affects What You See
Standard costing shows variances against fixed standards, while actual costing posts what really happened to each job. Many job shops use actual or estimate-based job costing since each job is unique. Whichever method you use, decide how overhead is applied: by labor hour, machine hour or a percentage of direct cost.
A shop where overhead is allocated by direct labor hours will show laser-heavy jobs as artificially profitable and weld-heavy jobs as expensive. Review the allocation once a year against your actual cost structure, particularly after buying automation that shifts cost from labor to machines.
Try it here, free
Sheet Metal Part Cost Calculator
Calculates the price of a cut and bent sheet metal part from material weight, utilization, cutting and bending time, finishing, setup and markup, and shows price per square foot.
From Variance Report to Pricing Decisions
The purpose of the report is to change future quotes. Review it monthly. Look at estimate accuracy by process, by customer and by part family, and update rates, utilization and setup assumptions where error is systematic.
Note the customers whose jobs consistently run over estimate, because that may be a pricing, engineering-change or documentation problem rather than a shop floor problem. Do not reprice everything based on a single bad job. Look for patterns across at least a quarter of data before adjusting, otherwise you chase noise.
AI as a Margin Analyst
An AI layer on the ERP can answer questions like which customers had the largest gap between quoted and actual margin last quarter, or which part families regularly exceed their weld hours, without anyone building a custom report. It shows the underlying query and works from your own data under existing permissions. The limit is data quality: if time and material postings are unreliable, answers will be too.
If your reporting needs are modest and an analyst can produce the monthly variance report in an hour, an AI layer is a convenience, not a necessity.
Frequently asked questions
How do I calculate job margin for a fabrication job?
Subtract actual material, direct labor, machine cost, outside processing and applied overhead from the invoiced revenue, then divide by revenue. Compare each cost component with the estimate so you can see where the difference came from.
What is a normal gross margin for a sheet metal fabricator?
It varies widely by mix, equipment and customer base. Rather than chase an industry figure, track your own margin by job type and customer and set targets from that history.
How often should I review actual versus estimated cost?
Monthly for trends and immediately for large or unusual jobs. Adjust rates and assumptions quarterly, once there is enough data to separate pattern from noise.
Ask Netray to show estimated versus actual margin by customer and process from your own ERP data, privately and read-only, before you decide whether a larger project makes sense.