We read every product page back against the code. It was wrong in both directions

On 26 September we read every product page on rutba.io back against the code: the catalogue listings, the comparison pages, the pages beside them on our other sites, and the posts on this blog that describe an app. Each claim was checked against the app it describes.

We expected to find pages promising too much, and we did. We also found the opposite, and more of it than we would like.

Pages that said less than we had built

Procurement, Dispatch, Track, Quality and Maintenance were listed as coming soon. The Fleet app was described as not built. The Workforce rota and absence, Talent’s courses, skills and succession, and the first live bank feed in Rutba Books were all named as things still to come. Every one of them is built.

The worse part is when. Most of these were built before the post that called them unbuilt was published. The bank feed was built four days before the Books post said it was not. Quality, OEE and the finite schedule were built the day before the manufacturing post listed them as outstanding. Procurement had ten screens two days before the stock post said it did not exist, adding that we would rather lose the sale than disappoint somebody in month two. It was losing a sale for nothing.

The comparison pages were the plainest case. They said our Recruit app had no features and advised readers to keep their applicant tracking system until it did. Recruit had eleven screens the day after that was written, and its own listing was corrected on 3 September. The comparison pages went on saying it until 26 September.

Pages that said more than the product does

• The CRM claimed quotes, lead scoring and company records. It has none of the three. Quotes live in Rutba Books, beside the invoices they become.
• Marketing claimed drip journeys, a shared social calendar, an approval step and conversion tracking. It sends a campaign once or on a repeat, and records opens, clicks, bounces and unsubscribes for each recipient.
• The storefront plans listed themes, price lists and abandoned-cart recovery. None of them is built.
• The order pages said counter sales land in the order queue. They settle at the till, against the same stock.
• Rutba Mail claimed a customer’s balance beside each message. It files a message against the customer, contact or order it concerns.
• Drive claimed full-text search. It has none.
• Studio claimed designs bound to the live catalogue. It renders one template once per row of a CSV or JSON dataset.
• The Sign API claimed an SDK generated from its routes. The client is written by hand and held to the API’s OpenAPI document by a test.
• The Relay’s data-deletion page told owners to press Close account in the dashboard. There is no such button; an owner emails us instead.
• Rutba Office pages said a release arrived most days. It was 34 releases on 8 of the 19 days since 8 September: it ships in bursts.

Even a picture carried one: the drawing on the CRM post had a box labelled Quote.

How a page drifts

None of it was written to mislead. A product page is prose about a live product. It is written once, on the day the thing is true or planned, and then nobody re-reads it while the product moves underneath. A label is worse, because it is short: "coming soon" gets remembered rather than checked, and copied from one page to the next.

Neither direction is harmless. An under-claim loses a buyer who needed the thing and was told it did not exist. An over-claim loses something harder to get back: the trust of a customer who bought on the strength of the page and then went looking for the feature.

An under-claim costs a sale. An over-claim costs the customer who believed the page.

What we changed

• Sixteen posts corrected, each dated. Where a post was wrong on the day it was published, the text is corrected in place and shows when it was updated. Where later work overtook a claim that was true when written — Fleet, Dispatch and Track, the asset register’s work engine — the original stays and a dated note says what changed. A silent rewrite would make the blog look as if it had always been right.
• Five pictures redrawn, with their alt text, because they drew the old claims.
• The listings name what shipped, as well as what did not: procurement’s orders, price lists and spend, the dispatch exceptions queue, payroll’s bank file and emailed payslips.

The one that stung was a check we already had. The catalogue carries a script that compares every listing with the apps the suite actually ships: an app no listing sells, a key no listing grants, an unbuilt app labelled as available. Since a directory move on 13 September it had been looking for the apps in the wrong place, finding nothing, and passing by skipping. A check that passes because it cannot see is worse than no check, because it is trusted.

It runs again now, and it failed at once on one real gap: the Affiliate Program was sold nowhere. Its app was in no listing and its key was granted by none, so nobody could have bought it. It is now listed under Marketing and labelled coming soon, because applications, review and agreements are built, and coupons, earnings and payouts are not.

What is still labelled, on purpose

Not every label moved. Early access and coming soon still mean what they say, and they stay where they are true: Facilities is early access, with the statutory register, room bookings and health and safety not built; the Affiliate Program is coming soon; the Calls apps say what each one shows today, and that dialling and routing are not built. Those labels were read in the same pass, and the ones left are true.