Product Feed Rules: Clean & transform catalog data without code

Your store's product data was written for product pages, by people who were never told it would someday feed four ad platforms. Titles run long, brand columns sit empty, categories follow your logic instead of the platforms' taxonomy - and none of it is anyone's fault, because the store is the wrong place to fix it.

Feed rules are the right place: transformations that run between your store and each platform's output, correcting and enriching data on the way through. The store stays untouched, the platform teams get spec-compliant feeds, and the fixes apply to every product at once - including ones you haven't added yet.

This guide covers how rules work and, more usefully, the ten recipes that solve the problems every catalog actually has. The workflow context is in product feed management.

How feed rules work

A feed rule is a condition plus a transformation, applied per product at feed generation:

• If a field is empty → fill it from another field or a default

• If a value matches a pattern → rewrite, truncate, or normalize it

• If product data meets a condition → set a label, category, or availability value

Ten rules worth copying

1. The title formula

Brand + Product Name + Key Attribute, truncated at 65 characters so Meta doesn't cut it mid-word. Front-loads what buyers scan for, standardizes the whole catalog in one pass. Keep the full title for platforms that display more.

2. The brand backfill

If brand is empty → fill from your vendor field; if that's empty → default to your store brand. An empty brand column is a Meta rejection and a matching penalty everywhere else; this rule retires the problem permanently.

3. The category mapper

Map your product_type values to google_product_category - the taxonomy Meta, TikTok, and Pinterest all use. Build the mapping once per category, 2-3 levels deep; every product inherits it.

4. The availability threshold

in stock only when inventory is above a buffer (say, 3 units), not merely above zero. Stops ads sending five buyers to the last unit, and absorbs inventory-sync lag before it becomes an oversell.

5. The sale price guard

Populate sale_price only when it is genuinely below price. Add sale_price_effective_date when the promotion has a scheduled window; use an open-ended sale only where the destination permits it and the landing page shows the same price.

6. The text sanitizer

Strip HTML tags, emojis, control characters, and ALL-CAPS from titles and descriptions. Meta rejects HTML and shouting; TikTok rejects emojis outright. One rule, every text field, zero rejections from copy-paste artifacts.

7. The margin labeler

Compute margin band from price and cost, write it to internal_label (on Meta - volatile labels belong there, not in custom labels). This is the rule that unlocks margin-tiered product sets and profit-aware bidding.

8. The bestseller flag

Top N products by sales rank get a bestseller label, refreshed on every generation. Prospecting sets built on it stay current without anyone curating a list.

9. The season tagger

Category- or attribute-driven season labels (summer, holiday, evergreen). Come campaign time, the seasonal set and its creative treatment are already addressable.

10. The image fallback

If additional_image_link is empty → leave it empty and route the product to an enrichment queue; duplicating the main image adds no new buyer information. If the main image fails your size check → route the product to a fix list instead of the feed. Do not publish an asset that fails the destination's image requirements.

Rules to refuse

Two anti-patterns worth naming:

• Rules that lie - inflating availability, masking prices, "optimizing" titles into keyword salads. Platforms crawl landing pages; every dishonest rule is a mismatch disapproval on a timer.

• Rules that hide broken sources - if half your catalog needs a brand backfill, the backfill is triage, and the store data deserves a real fix. Rules patch the last mile; they shouldn't become load-bearing fiction.

Testing and maintaining rules

• Preview before publish: run the rule, diff the output against the previous feed, check the changed-row count matches your expectation

• One change per deploy: stacked rule edits make regressions unattributable

Putting feed rules into the publishing pipeline

Feed rules are most useful when they live in the pipeline rather than in a spreadsheet someone reruns by hand. Flow's catalog engine provides the mapping and validation layer between source data and channel output, while the creative studio can reuse supported product fields inside templates and variations.

Implement only the rules your current workspace supports, preview the changed rows, and diff the output before publishing. A rule should make the catalog more truthful and maintainable; it should never hide a broken source.

FAQ

Final thoughts

Feed rules are leverage in its purest form: a condition written once, applied to every product forever. The ten recipes here cover most of what any catalog needs - clean titles, filled fields, honest availability, self-managing sales, and labels that make segmentation real.

Write them, preview them, and let the pipeline enforce what used to be a wiki page nobody read.