We Scored 23 DTC Product Pages. Machine Readability Is Almost Binary.

A Distribution With No Middle

As shopping increasingly happens through agents and answer engines, a product page has a second audience that never sees it render: a machine reading the markup with no JavaScript and no patience.

The PlatformDTC research team scored 23 direct-to-consumer product pages on how well that second reader can understand them. The median score was 100 out of 100 — and there was effectively no middle of the distribution at all.

Pages either emit schema.org Product markup and land at the top, or emit none and sit in a narrow band below. A quarter of the sample had no Product markup whatsoever, including brands most readers would recognise.

Why Binary Is Good News

A bimodal result is a more actionable finding than a smooth curve would be. It means machine readability is not a long programme of incremental tuning that a team chips away at over quarters.

It is closer to a switch. Emit correct structured data for the product — name, price, availability, identifiers — and a page moves from the bottom cluster to the top one. Most of the distance is covered by one decision rather than by a hundred small ones.

The Checker, and Two Ways We Got It Wrong

The scoring was done with a checker built to read a page the way a crawler does, with no JavaScript executed. It has been published as a free tool.

Building it surfaced two of our own bugs, both published alongside the results. The one worth repeating: an early version passed a 113-character bot-challenge page as healthy. A page that never reached the real content scored as though it had.

That failure mode matters beyond our tool. Any audit that measures what a crawler receives has to distinguish between a page that is readable and a page that was never served, and the two are easy to confuse when you are counting bytes rather than reading them.

Read the Study

The scoring method, the full distribution and the checker are published at platformdtc.com/blog.