PRDCT-704: describe and type 28 marketing-sales extractor pages

[Iamfle4ka]

One more sub-issue under PRDCT-696. This PR is cut from main and does not depend on #1189, #1190, #1192, #1193 or any other open PR. The marketing-sales index and the mixpanel pages are in #1183 and left out here.

Open questions

I left two pages untyped (mixed).

• google-analytics: concept prose sits next to numbered steps. :27 "How Data Loading Works" and :29 "There are two distinct "incremental" concepts" come before :48 "1. Name your output table in Storage" and :56 "5. Click Test query to preview results". Should Features and How Data Loading Works (:18 to :34) become an explanation page, or shrink to one line next to the incremental load step of the how-to?
• facebook-ads: a UI walkthrough (:15 to :38) is followed by about 70 lines of per-option reference (:44 to :112) and the output data model (:114 to :144). Should the page split into a how-to for authorizing an account and creating a query, plus a reference page for the query options and the output tables?

I typed two pages how-to after a close call.

• customerio has a 16-row column table, but it is one sample record that sits inside the step that selects the data.
• bing-ads has long parameter lists, but they sit inside the steps that set them, and its column reference already lives on the report presets page.

Evidence I want the next pass to see:

• The two Adform list pages name neither the page nor the connector in their lede, so a reader landing on one from search cannot tell what the list belongs to.
• The ledes on google-cm360, google-dv360 and pinterest-ads describe the product, and say nothing about what the page does.
• D14 title vocabulary: "Customer IO" against Customer.io, no AdWords on google-ads, CM360 and DV360 where readers type the long names, "Google Business Profile" against "My Business", "Shoptet Permalink" against Shoptet, "Zoho" against Zoho CRM, and no "Meta Ads" on facebook-ads, which says only Facebook.
• linkedin-pages :16 carries a create-configuration step on a page I typed reference.

What changed

All 28 pages now have a description, and each carries the type I read from the page. bing-ads/report-presets-columns-and-pk also got one lede, and no page got a Next footer, because the connector catalogs are exempt from the Next rule (Nikita, 2026-10-13). The diff is additions only (+82, 0 deletions).

Type markers:

• how-to (22): every connector walkthrough except the two untyped pages and LinkedIn Pages.
• reference (4): both Adform list pages, Bing Ads report presets and LinkedIn Pages.
• untyped (2): google-analytics and facebook-ads.

For the owners (not changed here)

• google-ads :28 links the Query Builder at API version v17, which Google has probably retired. Unverified.
• customerio :111 links developers.keboola.com/extend/component/processors/, and that domain is being retired. The page is also built on the "Customer IO Beta API" (:9), which may have left beta.
• gooddata-reports :12 links an old help.gooddata.com/display/API/ path, and :32 reads the legacy /gdc/md/ report URIs, so the connector may be tied to a retired platform.
• adform-dsp-reports :14 links the MasterData connector on components.keboola.com. The two list pages are missing from the nav, and their slugs use hyphens where the file names use underscores.
• salesforce :17 asks for "Salesforce V2" while the title says Salesforce, and the body has em dashes at :35, :46 and :61.
• Several pages use -- or en dashes as list separators: babelforce :20, mailchimp :36 to :42, pipedrive :27 to :28, customerio :31 to :32, adform :57 to :58, zoho :36 to :37.
• zoho :28 mentions "the Self Client" without saying what it is, and :11 links Bulk Read API v2 docs that may be outdated.
• facebook-ads :25 names the app to revoke as Keboola Ads Data Source. Worth a check against the current app name.
• shoptet :19 uses a sample export link with a realistic-looking hash.

Checks

• Diátaxis script on the 28 pages: 0 FAIL, 38 WARN (22 D7, 10 D6, 2 D1 and 2 D12 on the Adform lists, 2 D2 on the untyped pages).
• npm run build: clean.
• Link and redirect check against main: 0 blockers, 0 warnings.
• Dash grep over every added line: clean.
• A reviewer and the diataxis-checker agent read the pages. All 26 types and both untyped verdicts held, and their other findings are the evidence list above.

🤖 Generated with Claude Code

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
connection-docs	Ready	Preview	Oct 14, 2026 2:25pm UTC

[linear-code]

PRDCT-704