TLDR: Your real cost per design is the designer's loaded hourly rate multiplied by the hours a case actually consumes, plus software and hardware amortised per case, plus the cost of rework. Most labs only count the first term, which is why in-house design usually looks cheaper on paper than it is.

Why is the outsourced number so much easier to see?

Outsourced design arrives as an invoice with a per-arch price on it. In-house design is paid in salary, which is a fixed monthly number that does not move when volume moves. One is a unit cost you can read. The other has to be calculated, and mostly is not.

This asymmetry quietly distorts a lot of lab decisions. A lab owner can tell you to the dollar what an outside design service charges per arch, because the bill says so. Ask the same owner what an in-house design costs and you usually get a shrug and a guess, and the guess is nearly always low.

Industry rates for outsourced denture design run up to about $40 an arch. That is a clean number. The question is what the same work costs when you do it yourself, honestly measured.

What goes into a real cost per design?

Four components: loaded labor for the hours the case actually takes, software and hardware amortised across monthly volume, the cost of rework and remakes traceable to design, and the opportunity cost of design capacity you could not sell because it was busy.

Take them one at a time.

Loaded labor. Not the salary. The salary plus payroll taxes, benefits, paid time off, and the share of supervision the role consumes. A common rule of thumb puts loaded cost 25 to 40 percent above base pay, and your accountant can give you the real multiplier for your lab. The U.S. Bureau of Labor Statistics reports median pay for dental laboratory technicians at $49,610 a year as of 2025, which is a floor for the calculation rather than the number for a skilled CAD designer.

Hours the case actually takes. This is the term labs get wrong most often, because the estimate is usually the time at the workstation with the case open. It should also include setup, the interruption cost of switching between cases, and the second pass when something comes back. Manual CAD design of a full-arch denture commonly takes 30 minutes to over an hour of focused work.

Software and hardware, per case. Add annual CAD licence fees, workstation depreciation and the support contract, then divide by cases designed in the year. In a low-volume lab this term is larger than people expect. At 1,000 cases a month it gets small quickly.

Rework. The cases that come back. Not every remake is a design fault, so attribute honestly, but design-attributable rework belongs in the design cost and it rarely appears in anyone's model.

A worked method you can run this week

1. Pick a month that was typical. Not your busiest, not your slowest.
2. Count full-arch denture designs completed in that month. Call it V.
3. Total the loaded cost of everyone who spent time designing, for the share of their time that went to design. Call it L.
4. Total design software licences, workstation depreciation and support for the month. Call it S.
5. Count design-attributable remakes and rework in the month, and cost them at materials plus the labor to redo them. Call it R.
6. Your cost per design is (L + S + R) divided by V.

Run the same calculation for a month six months earlier and compare. If the number is drifting up while volume is flat, you have a capacity problem that is showing up as overtime or as rework rather than as a hiring requisition.

How should AI-assisted design change the calculation?

It changes the hours term and leaves the others roughly alone. If a design proposal arrives in about two minutes and the technician's job becomes review and approval rather than building from scratch, the labor minutes per case fall while the review minutes stay. Model both, not just the first.

The honest way to model it is to split the hours term in two. There is generation time, which is where the large reduction sits, and there is review time, which is human and does not disappear. A model that assumes review time goes to zero is describing a product nobody is selling and, in this category, nobody is permitted to sell.

SmileShape reports labs reducing CAD design cost by up to 60 percent against outsourced rates of up to about $40 an arch. Treat that as the claim to test against your own number rather than as your number. The comparison that matters is the one you calculate.

What to do with the answer

• If in-house is well below outsourced and your designers have slack, you are fine. Revisit when volume grows.
• If in-house is below outsourced but your designers are saturated, your real constraint is capacity, not unit cost, and the choice is between hiring, outsourcing the overflow, or adding design capacity to the team you have.
• If in-house is above outsourced, find out why before you act. Low volume spreading fixed costs thinly is a different problem from high rework, and they have opposite fixes.
• If you cannot calculate it at all, that is the finding. Fix the measurement first. Every decision downstream depends on it.

Related reading

• what 30 cases an hour actually means
• hire, outsource, or add AI-assisted design
• why faster design alone will not fix turnaround

Sources

1. U.S. Bureau of Labor Statistics, Occupational Outlook Handbook, Dental and Ophthalmic Laboratory Technicians and Medical Appliance Technicians