Manufacturing KPIs That Matter for Job Shops

The Dashboard Problem

Walk into any job shop that recently implemented an ERP or MES and you'll find a dashboard with 30+ KPIs. Utilization. Scrap rate. Cycle time variance. Setup time ratio. Inventory turns. Most of them are green. The shop is still losing money on 20% of its jobs.

The issue isn't measurement — it's signal-to-noise ratio. Enterprise KPI frameworks (designed for plants running 10,000+ units of the same part) don't translate to job shops where every week brings different parts, different materials, and different quantities. A shop running 200 unique part numbers per month needs metrics that work across that variability.

After looking at what actually separates profitable job shops from struggling ones, seven metrics consistently show up. These aren't theoretical — they're the numbers that shop owners who've been in the business for 20+ years check first every Monday morning.

1. Quote Win Rate (and Why the Number Alone Is Misleading)

The headline number: what percentage of quotes you send turn into purchase orders. Industry average for job shops is 20-30%. Top performers hit 35-45%.

But win rate alone is dangerous. A shop that quotes everything and wins 40% might be leaving margin on the table — they're probably underpricing. A shop that carefully selects which RFQs to quote and wins 25% at healthy margins is likely more profitable.

The metric that matters is win rate segmented by job size. If you're winning 50% of jobs under $5,000 and 10% of jobs over $50,000, that tells you something specific about your estimating accuracy at different scales. Large jobs often get lost because the quote took too long, the scope wasn't fully captured, or the pricing didn't reflect the actual work breakdown.

This is where tooling matters. Manually estimating a complex RFQ package — reading drawings, extracting quantities, building a work breakdown — takes 2-6 hours. Tools like ForgeAI Workshop's Takeoff can extract scope and quantities from RFQ packages in minutes, which means you can quote more jobs without adding estimating headcount. More quotes out the door at consistent quality directly impacts win rate.

2. On-Time Delivery (OTD)

The single most customer-visible metric. OTD directly determines whether you get repeat business and referrals. Industry benchmark for job shops is 85-90%. Best-in-class shops hit 95%+.

Measure it simply: jobs delivered on or before the promised date, divided by total jobs delivered, over a rolling 90-day window. Don't use a 30-day window — it's too noisy for low-volume shops. Don't use calendar year — it hides seasonal patterns.

The hidden killer of OTD in job shops isn't machine breakdowns or labor shortages. It's front-end delays: waiting for drawing clarifications, incomplete RFQ packages, missing material certs, ambiguous specs. By the time the job hits the shop floor, it's already behind schedule because the quoting and planning phase consumed too much of the lead time.

Tracking OTD against quoted lead time (not customer-requested date) reveals whether the problem is in planning or execution. If you're consistently hitting your quoted dates but customers are unhappy, your quoted lead times are too long. If you're missing your own quoted dates, the problem is downstream.

3. First Pass Yield (FPY)

Percentage of parts that pass inspection the first time without rework, repair, or scrap. This is the single best proxy for process capability in a high-mix environment.

Job shop benchmark: 92-96%. Below 90% and your rework costs are eating your margins. Above 98% and you might be over-inspecting or running excessively conservative processes (which costs money in cycle time).

Track FPY by operation type, not by part number. In a job shop, you'll never run enough of any single part to get statistical significance. But you will run enough turning operations, milling operations, and welding operations to see meaningful trends. If your FPY on 5-axis milling is 88% while turning is 97%, that tells you exactly where to focus process improvement.

4. Estimated vs. Actual Hours

This is the metric that directly measures estimating accuracy — and estimating accuracy is what determines whether a job makes money. Track the ratio of estimated labor hours to actual labor hours for every completed job.

A ratio of 1.0 means your estimates are perfect. Below 1.0 means jobs are taking longer than estimated (you're losing money). Above 1.0 means you're padding estimates (potentially losing bids).

Healthy range: 0.90-1.10. Consistently below 0.85 on a job category means your estimating process for that work type needs calibration. This often happens when estimates are based on gut feel rather than structured work breakdowns.

The fix isn't more experienced estimators — it's better data flowing into the estimate. When your quote is built from an actual work breakdown with shop rates applied to each operation, the estimate-to-actual variance tightens naturally. Workshop's Quote Builder structures this process: work breakdown in, dollar quote with shop rates out. The structured output also makes it easier to do the post-job comparison and identify which operations were consistently over or under-estimated.

5. Revenue per Employee

Total annual revenue divided by total headcount (including office staff). This is the simplest measure of overall shop productivity and the easiest to benchmark.

Shop Size	Below Average	Average	Top Quartile
10-25 employees	<$120K	$150-180K	$220K+
25-50 employees	<$140K	$170-210K	$250K+
50-100 employees	<$160K	$190-240K	$280K+

Revenue per employee increases with shop size due to economies of scale, but the spread within each category is wide. A 20-person shop at $220K/employee is outperforming a 60-person shop at $180K/employee in terms of operational efficiency.

This metric is most useful as a year-over-year trend line. If it's flat or declining while you're adding people, you're scaling headcount faster than revenue — a common trap when shops win a big contract and hire aggressively before the revenue stabilizes.

6. Quoting Throughput

Number of quotes sent per week (or per estimator per week). This is the leading indicator that most shops ignore until it's too late.

If your quoting throughput drops, your pipeline dries up 30-60 days later. If it spikes beyond capacity, quality suffers and win rate drops. The goal is consistent throughput that matches your capacity.

Benchmark: a single experienced estimator should be able to produce 15-25 quotes per week for typical job shop work. If your estimator is producing 8-10, the bottleneck is usually in the data gathering phase — reading drawings, extracting dimensions, looking up material costs, building the work breakdown. Shops that have reduced this overhead (through better templates, structured extraction, or tools that automate scope capture) consistently hit the higher end of the range.

7. Job Margin Variance

The difference between quoted margin and actual margin on completed jobs. This is the metric that ties everything together.

Average job margin for healthy job shops is 25-40%, depending on complexity and industry. But the variance matters more than the average. A shop averaging 32% margin with a standard deviation of 5% is in much better shape than one averaging 35% with a standard deviation of 15%.

High variance means some jobs are very profitable and others are underwater. That's a sign of inconsistent estimating, scope creep, or jobs that weren't fully understood at the quoting stage. When an RFQ comes in and the team doesn't catch that it requires a special material cert, a secondary process, or a tight-tolerance feature buried on page 14 of the drawing package, the quote goes out low and the actual cost blows past it.

Reducing margin variance starts at intake. Tools that extract and structure RFQ data — pulling parts lists from drawings, flagging spec requirements, scoring package completeness — catch the scope items that manual review misses. ForgeAI Workshop handles several of these extraction tasks (Drawing Extraction for parts lists, Readiness Check for package completeness scoring), which helps standardize the front-end process that feeds into estimates.

Implementing Without Overhead

The biggest objection to KPI tracking in small shops is time. Nobody has a data analyst. The shop manager is also the lead estimator, the HR department, and half the sales team.

Here's what works:

• Start with three metrics, not seven. OTD, estimated vs. actual hours, and quote win rate give you the most signal per unit of tracking effort. Add the others once these three are habitual.
• Weekly cadence, not daily. Review metrics every Monday morning. Daily tracking creates noise and anxiety without actionable insight in a job shop environment.
• Spreadsheet is fine. You don't need a BI tool. A shared Google Sheet with one row per completed job, tracking quoted hours, actual hours, quoted margin, actual margin, and delivery date is enough to calculate everything above.
• Automate the inputs, not the dashboard. The time cost isn't in calculating the KPI — it's in gathering the data that feeds it. If your quoting process produces structured output (line-item work breakdowns, material lists, labor hours by operation), the post-job comparison takes 5 minutes instead of 30.

What to Do When the Numbers Are Bad

The point of measuring isn't to have pretty charts. It's to know where to act. A quick decision tree:

• Win rate below 20%: You're either quoting too slow (speed), quoting too high (pricing), or quoting the wrong jobs (targeting). Check quote turnaround time first — it's the most common culprit.
• OTD below 85%: Track where jobs lose time. If it's before the shop floor, fix your quoting and planning intake. If it's on the floor, look at scheduling and setup time.
• Estimated vs. actual below 0.85: Your estimating process needs more structure. Build quotes from operation-level work breakdowns, not job-level gut estimates.
• Margin variance above 12% standard deviation: Jobs are getting quoted without full scope understanding. Improve your RFQ intake process to catch missing information before the quote goes out.

Every one of these failure modes traces back to the same root: the gap between the information in the RFQ package and the information in the estimate. Close that gap and most of these metrics improve as a side effect.