What 152 Australian coffee roasters actually look like under the hood

We read the public product data of 152 Australian coffee roasters and counted the things that quietly cost money. Here is the whole result, including the parts that are less dramatic than they sound.

Scanned August 2026. No store is named. Method at the bottom.

85/100

Median store health score. The full range was 66 to 99, so the spread is wide. Median catalogue size was 48 products; the largest was 2,000.

How common each problem is

Percentage of the 152 stores with at least one instance.

Products with fewer than 2 images
93%
Variants with no SKU
89%
Product images with no alt text
88%
Products with descriptions under 200 characters
87%
Products live but completely sold out
73%
Products whose 'was' price is BELOW the current price
28%
Products sharing a title with another product
21%
Sampled pages with no Product structured data
12%
Sampled pages with no meta description
7%

The same data with counts

The percentage tells you how widespread something is. The median count tells you how bad it is at a store that has it, which is usually the more useful number.

ProblemStores affectedMedian count, affected stores
Products with fewer than 2 images93%18
Variants with no SKU89%20
Product images with no alt text88%20
Products with descriptions under 200 characters87%8
Products live but completely sold out73%7
Products whose 'was' price is BELOW the current price28%2
Products sharing a title with another product21%5
Sampled pages with no Product structured data12%5
Sampled pages with no meta description7%2
Scan your own store against this data Paste your address, get your numbers in about five seconds. No app install, no login to your store.

Method, so you can argue with it

Every Shopify store serves its full catalogue at /products.json, publicly, with no key and no login. We read that, plus a small sample of product pages, at one request per second with an identifying user agent. Nothing behind a login, no orders, no customer data, no traffic figures.

Two honest caveats. Catalogues were capped at 2,000 products, so a handful of counts are floors rather than totals. And roughly one store in ten disables the public feed, so those are missing from the sample entirely.