DigitalChiselCo First-party data · 365 days to 2026-09-22
The State of CNC Relief Carving
2026
Everybody in this trade argues about what sells. Almost nobody has the numbers, because the shops that hold them do not publish them. This report does: 5,344 real sales across 1,226 designs over a full year, with the method written out so you can judge for yourself how much of it to believe.
- Published
- 22 September 2026
- Author
- Jolly, DigitalChiselCo
- Window
- 365 days
- Licence
- Free to quote with a link
Key findings
- 28% of designs never sold once, even after a full year on sale. The top 114 designs, 9.3% of the catalogue, produce half of all sales.
- Pricing low did not buy volume. Designs at $15 to $20 averaged 18.2 sales each. Designs at $5 to $8 averaged 3.8, while making up 82% of the catalogue.
- Deer and elk is the most productive subject per design at 8.1 sales each. Religious work is the largest by volume at 16.8% of sales and carries the highest average price, $11.19.
- A finding was removed before publication. Controlling for how long a design had been on sale destroyed an apparent result that long titles sell nothing. See the method section.
1 Most designs sell nothing, and that is normal
Of 1,226 designs on sale for a full year, 342 never sold a single copy. The top 114 designs account for half of all sales, and the top 10% for 51.9%.
This should change how a designer judges their own catalogue. Measured by hit rate, a shop like this looks like it is failing most of the time. It is not. A 28% miss rate is the cost of finding the 114 designs that carry everything. It is also uncomfortable for anyone selling a pack of "proven winners": the winners are only visible afterwards.
2 The expensive designs sell more, not fewer
This is the finding that surprised us most, and the one we expect to be argued with hardest. Designs priced at $15 to $20 average 18.2 sales each. Designs priced at $5 to $8 average 3.8: roughly a fifth as often per design, at about a third of the price.
Read this one carefully
Price is not causing the sales. The expensive designs are the large, complex, multi-figure scenes that took longest to make, and the cheap tier is where nearly the whole catalogue sits (1009 of 1,226 designs). The defensible claim is narrower than "raise your prices": pricing low did not buy volume. The cheap tier holds 56.3% of revenue while holding 82% of the catalogue. The $15 to $20 band is 20 designs, so treat it as a direction, not a decimal.
3 What people actually carve
Subjects are matched on listing titles, and a design can belong to more than one, so these describe demand for a theme rather than slices of a pie. They are sorted by sales per design, which is the number that matters when deciding what to make next, rather than by total sales, which mostly records what a shop happened to make a lot of.
Religious work is the largest single block by volume at 16.8% of all sales, and carries the highest average price at $11.19. But deer and elk is the most productive thing in the catalogue per design, and nautical has the highest hit rate of any subject at 88.5%: almost everything in it sells something.
The quiet lesson sits in the gap between volume and rate. Floral and botanical has 90 designs producing 3.01 sales each. Deer and elk has 67 producing 8.1. Making more of the first category feels productive and is not.
4 The same picture by shop category
Categories are the shop's own taxonomy rather than title matching, so this is a second and independent cut of the same year. It agrees with Figure 2, which is the useful part.
5 Method, and what we threw away
The data
Every figure comes from one shop's own sales records over 365 days to 2026-09-22. Counts are sales per design. No buyer, order or personal detail is used anywhere in this report, and nothing below the aggregate level is published.
The maturity control, which is the important part
776 designs were excluded, almost all because they were added in the last few months and have not had a year in which to sell. Including them would have produced a far more dramatic report: a 55% failure rate instead of 28%, and a striking result that listing titles over 90 characters sell essentially nothing.
Both are artefacts. The long titles belong to the new designs, which have not had time to sell at any length. Restricted to designs listed before 2026-07-01, the title-length effect disappears entirely: 72.0% against 72.7%. It is not reported here because it is not real.
What this is not
One shop, selling digital files, mostly to buyers in English-speaking countries, on one marketplace. It describes that demand well. It is not the world market. Subject buckets overlap by design, and bands with few designs carry their counts so you can discount them yourself.
Corrections
If you find an error, say so and it will be fixed and the change noted here. A report nobody can check is marketing, not data.
6 How to cite this
Free to reproduce, including the figures, for any purpose including commercial, with a link back. No permission needed and no email required.
The State of CNC Relief Carving 2026. DigitalChiselCo, 22 September 2026.
https://digitalchiselco.com/reports/state-of-cnc-relief-carving-2026
Questions about the method, or after a cut of the data that is not published here? Ask. If you run a shop and would pool anonymised numbers for next year's edition so it stops being one shop's view, that is exactly what we want to do.