Capability pages a buyer can compare: specifications in dated HTML tables

What a comparison asks of each supplier's page

Comparing suppliers means filling in the same row for each of them: largest part, tightest routine tolerance, materials, certificates, batch sizes, lead time. A buyer might do it in a spreadsheet, on the back of an RFQ, or by asking an assistant to line three shops up. We have not measured which, and nothing on this page depends on it. What can be said is what a page gives any reader who is filling in that row.

A table with a unit in every cell gives the row its values. A paragraph that calls the shop "high precision" and the machines "large-format" gives it nothing to write down, and a figure without a date gives a value nobody can place in time. With web search on, an AI assistant can state a figure for your firm correctly only if a page it fetched, or text it learned from, states that figure.

The machine list, with envelopes and capacity in units

Start with the machines, because the other figures depend on them. One row per machine type: how many you run, the travel on each axis or the turning diameter and length, the heaviest part it takes, and the detail a buyer checks next, such as the number of axes or the bar capacity. Give millimeters and inches, kilograms and pounds, when you sell into markets that measure differently.

The date goes in the caption, and it changes the week a machine arrives or leaves; that is what freshness means on a capability page. Capacity sits under the table in words: shifts per day, days per week, and the batch sizes each cell is set up for, every one as your own bracketed figure with the month it was checked. A buyer reading "[n] shifts, [n] days a week, batches of [n] to [n]" knows what kind of work you are built for.

Tolerances, stated the way your inspection proves them

One line saying the shop holds [tolerance] everywhere is the tolerance statement a buyer can do least with, because a tolerance depends on the process, the feature, the material and the size of the part. State tolerances per process and feature: what you hold routinely, what you offer on request, on which materials, and how the result is inspected.

Keep the routine column for what you hold as a matter of course; a tolerance you offer on request has a column of its own. If you publish a quality figure such as first-pass yield, publish it as your own bracketed, dated number with its method: "[yield] across [n] lots, [period], counted as [method]". A quality figure with no period and no method gives a buyer nothing to check it against.

Materials, certificates and volumes, in the same plain form

List materials by process and by grade, the way a drawing names them, and say what you do not take: "we do not machine [material]" answers a question as surely as the list does. Where a grade calls for mill certificates or a heat treatment, say whether you source and document them.

State each certificate as it was issued: the standard's name as printed on the certificate, the issuing body, the certificate number, the scope in the certificate's own words, the dates it is valid from and until, and a link to the issuer's public lookup where one exists. Keep the scope exactly as wide as the certificate. One that covers machining at one plant says nothing about assembly at another. What the standard requires is between the standard and your auditor; the page says what you hold.

Volumes and lead times follow the same pattern: prototype, low-volume and series batches as [n] to [n] parts, annual capacity as [n] parts, a quote in [n] working days and a first article in [n] weeks from an approved drawing, each with the month it was checked.

Datasheets: an HTML page first, the PDF beside it

For a standard machine or a catalog component, the datasheet belongs on the product page as a table: the figures, their units, a caption naming the model and the revision, and header cells that say what each column holds. The HTML standard's table element has the parts for it: a caption, and header cells scoped to their column or row.

Keep the PDF, because engineers print and file datasheets, and link it under the table as a download that says what it is: "Datasheet, [model], revision [n], [month year], PDF, same figures as this page". Produce both from one source, so a changed figure changes in both on the same day. The PDF trap covers what a fetcher receives when the file is the only copy. In our audit of 9,483 small business websites, 2.1% kept their real content in a PDF, a figure for the whole sample with no engineering cut.

A configurator that computes its figures in the browser is the same problem in another form: the HTML an AI crawler receives holds the controls and none of the results. Google renders JavaScript as a separate, deferred step, and the assistants' crawler documentation says nothing about rendering, so do not count on their fetchers running scripts. Publish the standard configurations as static tables and keep the configurator for the rest.

Markup that repeats the tables, and nothing more

Once the tables are on the page, the structured data can repeat them in JSON-LD. A standard machine or component is a Product: name, model, mpn, material, width, height, depth and weight, with additionalProperty holding each figure schema.org has no property for, such as an axis travel or a cycle time. schema.org's PropertyValue carries it with a name, a value or a minValue and maxValue, and a unit: unitCode for a UN/CEFACT Common Code of three characters, or unitText for the unit as written.

A process you sell is a Service under your company's node, with serviceType, provider, areaServed and category, and hasCertification pointing to the Certification that covers it, stated no wider than the certificate. A template for one machine, with brackets for your own values:

{
  "@context": "https://schema.org",
  "@type": "Product",
  "@id": "https://www.example.com/machines/[model]#product",
  "name": "[Machine name] [model]",
  "model": "[model]",
  "mpn": "[part number]",
  "manufacturer": { "@id": "https://www.example.com/#organization" },
  "material": "[frame material]",
  "width": { "@type": "QuantitativeValue", "value": "[width]", "unitText": "mm" },
  "depth": { "@type": "QuantitativeValue", "value": "[depth]", "unitText": "mm" },
  "height": { "@type": "QuantitativeValue", "value": "[height]", "unitText": "mm" },
  "weight": { "@type": "QuantitativeValue", "value": "[weight]", "unitText": "kg" },
  "additionalProperty": [
    { "@type": "PropertyValue", "name": "Cycle time per part", "value": "[cycle time]", "unitText": "s" },
    { "@type": "PropertyValue", "name": "Part length handled", "minValue": "[min]", "maxValue": "[max]", "unitText": "mm" }
  ]
}

Every value in the markup is a value in a visible table: Google's introduction to structured data says not to mark up information a visitor cannot see. Change the two together, since Google's general structured data guidelines ask for up-to-date information, and expect nothing guaranteed in return: they say correct markup enables a feature without guaranteeing it will appear. Which types describe the business, and which describe only the website, is the subject of the article on structured data.

Checking the page before it goes live

1. Open the page with scripts turned off and find each kind of figure in the text: an envelope, a tolerance, a material, a certificate number, a lead time.
2. Check that every figure has a unit and every table a date in its caption.
3. Open the PDF beside each datasheet and compare five figures with the page.
4. Compare the values in the Product schema with the tables line by line, and delete any value the page does not show.
5. Link the capability page to your RFQ page, and say there which drawings and details a quote needs.
6. Write the change and its date in your site log, so that a later reading of the answers can be set beside it, as the monthly routine describes.

Then put the question the page was built for to an assistant, in a buyer's words, and read which figures the answer gives for your firm; the record for that reading has a column for each one. The overview for machine builders and machine shops lists the other pages a supplier's site needs.