The pool is filled from the client's sitemap and from the pages Pinterest already rewards, then filtered on cooldown, topic coverage and board assignment. An empty candidate list almost always means the pool was never imported.

Control: P4.1.1 · The URL pool on this cycle, above.

Seasonal URLs surface automatically when their peak falls six to ten weeks out. The peak comes from the URL, or from its primary keyword when the URL has none of its own — set it on the keyword and every URL using that term inherits it.

Ask about launches and new blog posts, then add each one under URLs so it enters the candidate pool. Brand-new URLs are rewarded heavily.

Happens in: The client

The selection arrives proposed: what has won here first, then what the season is opening, with the reason on each row and a counter against the store's monthly frequency. Untick what the client does not want pushed — that is the one thing the ranking cannot know.

Control: P4.1.4 · This month's URLs on this cycle, above.

Five keywords come pre-picked with the reason each — gridded terms first, because the design brief can only set format and colour from a keyword that has a grid row. Confirm or swap.

Control: 1 · Setup on this cycle, above.

Five boards come pre-picked, ranked by what has already won on this account, then topic, then the approved Steal List. Check them: swimwear does not belong on a strapless bra board even though both are lingerie.

Control: 1 · Setup on this cycle, above.

Three to five long-tail terms become the text overlay on the click pin. Pre-picked longest-first from the keywords you just confirmed — check each one reads as a phrase somebody sees on an image, not as a label.

Control: 1 · Setup on this cycle, above.

Search the primary keyword and record what page one rewards. Without a grid row the design brief falls back to the 80/20 default.

Happens in: Pinterest search (incognito) or PinClicks

Direct where the client has usable lifestyle material, AI where they do not. The brief and the image prompt both branch on it.

Control: 2 · Design brief on this cycle, above.

Builds from the grid, the brand book, the taste graph and what has already won on this account. Underneath it: which format each of the four designs should be, and what the competition is doing for this keyword.

Four visually distinct designs: three SAVE at 2:3 and one CLICK at 9:16. Upload them per design — Canva, a shoot, the client’s own material — and the file is renamed to the SOP name on the way in, because Pinterest reads it out of the URL. Generating is the fallback for an account with no usable material: one image per design, each from its own prompt built from the visual worlds, the palette and the grid, and it takes a couple of minutes.

Copy variant A keeps the original; B, C and D each take 96% of the frame from a different corner and scale back. That makes all four pins off one design read as four images while sharing one copy set — which is why four copy sets per URL is right and sixteen would be waste.

File names are generated lowercase, hyphenated and keyword-bearing. Check they survived the export — design tools rename on download.

Colours right, overlay rule respected, four genuinely different designs, file name correct. A rejection needs a reason — it is what the next generation gets corrected from.

Four copy sets, drafted from the brand book, the tone of voice and this account's own research. You approve them afterwards.

Title length and keyword position, description 250 to 300, no exclamation marks, hashtags or dashes, and the brand book's banned words. A failure names the rule it broke.

Only what the validator cannot judge: does it sound like the brand, does it match the image, does the landing page deliver what the copy promises, are the four sets genuinely different.

Sixteen pins, dates and board rotation. Design 1 goes to boards 1-2-3-4, design 2 to 2-3-4-1, so every board gets every design.

Check the spread on the calendar first — nothing goes to Pinterest until you approve it. Queueing hands the pins to the cron, which publishes each one on its own date. Pins whose board is not on Pinterest yet are held back by name; press again once those boards exist.

Queues the sixteen pins; the cron posts each one on its own date. It does not publish now on purpose — the dates are spread over weeks, and posting them together would collapse the waterfall into a single day. Standard pins over the API, never the simplified or idea format, which are barely distributed.

What went live, what is still waiting, and what is in the way. A rate limit re-queues itself and needs nobody; an expired token never resolves on its own and takes the next cycle down with it.