Nobody Searches While Scrolling:

Discovery Is Not Search

A shopping engine matches product data to a query. Social commerce has no query. Meta, TikTok, Pinterest, and Snapchat decide which products to show, to whom, in which moment — and every input to that decision comes from two places: the behavioral signal the platform already owns, and the product data the merchant supplies. You control exactly one of those.

That changes what the catalog is for. In a search auction, data answers a question the shopper asked. In a discovery feed, data has to give the algorithm enough to make the match on its own — clean identity for event matching, distinct variants for delivery, images that survive placement rules, categories that route the product to the right audiences. The catalog isn't describing products to people. It's describing products to a delivery system, which then chooses the people.

The Connection You Already Have

Every platform ships a native sync. Shopify pushes to Meta's Commerce Manager and TikTok in a few clicks; Pinterest and Snapchat ingest the same catalog formats. Transport is solved here the way it's solved everywhere — which is why two merchants with identical connections run catalog ads with completely different results. The pipe is the same. The payload isn't.

A native sync moves storefront data as-is: the storefront title, the storefront image set, whatever variant structure the platform export produces, and nothing the export can't see — no segmentation logic, no margin encoding, no per-platform image selection. Everything in this guide is about the difference between data that arrives and data that performs once the delivery algorithm starts reading it.

Key Insight: Catalog ads are the automation story of paid social — Advantage+ and GMV Max run targeting, bidding, and creative assembly. The merchant's remaining input is the catalog they assemble from. Automated campaigns didn't reduce the data's importance; they made it the last lever.

The ID Match That Makes or Breaks Retargeting

Dynamic retargeting has one load-bearing joint: the product ID in the behavioral event has to match the product ID in the catalog. When a shopper views a product or adds it to cart, the Pixel or Conversions API fires an event carrying content_ids. The platform looks those IDs up in the catalog to know which product to retarget with. If the event says SKU-1042-NAVY and the catalog says 1042_navy, the lookup fails — and it fails silently.

Nothing errors. The catalog looks healthy, events flow, campaigns spend. What breaks is invisible: view- and cart-based audiences build thin or empty, dynamic ads fall back to broad delivery, and the platform's optimization learns from a fraction of the signal it should have. The symptom reads as "retargeting doesn't work for us" or "Advantage+ went cold" — a campaign diagnosis for a data-plumbing failure. ID mismatch is the single most common structural defect in social catalog setups, and the most common source is the obvious one: the event integration and the catalog were configured at different times, by different tools, from different ID fields.

Silent Failure: ID mismatch produces no disapproval, no error state, no alert. The only way to catch it is to test it: fire a view event for a known product on each platform and confirm the event's content ID resolves to a live catalog item. Run this check after every replatform, catalog rebuild, or tracking migration — the three moments it breaks.

Images Are Attributes

In social placements, the catalog is the creative. The image the ad renders comes from the feed, which makes image data a performance attribute, not an asset-library concern. Three failure modes dominate. Policy rejections: text overlays, watermarks, promotional badges, and borders get catalog items disapproved, and they tend to arrive in batches because the same template produced every image. Variant mismatch: a shopper who viewed the navy colorway gets retargeted with the parent's default image — a relevance loss caused by children missing their own image_link. And placement fit: catalog placements are built around square imagery; a catalog of uncropped landscape shots renders badly everywhere it appears.

Pinterest raises the bar in its own direction — a planning surface where lifestyle imagery and alternate angles carry product Pins, which makes additional_image_link a real field, not a nice-to-have. The operational point across all four platforms: image selection and per-platform image assignment belong in the feed logic, so each surface receives the image set built for it rather than whatever the storefront exported first.

Titles and Descriptions in a Feed That Scrolls

Social surfaces truncate hard, render mobile-first, and give a title one glance to register. The shopping-engine title strategy — attribute-dense, built for query matching — doesn't transfer intact, because there's no query to match. What transfers is the front-loading principle: product identity in the first 40 characters, because that's often all a placement shows. "Halcyon Supply Men's Weekend Tee — Navy" survives truncation. A title that opens with brand boilerplate or SEO padding shows a shopper three words that identify nothing.

Descriptions do quieter work: product detail surfaces, Shops listings, and the platform's own category matching all read them. Storefront SEO copy pasted into the description field is signal written for the wrong reader. Two sentences that say what the product is and who it's for outperform four paragraphs of keyword scaffolding — and this is precisely the field where per-channel transformation earns its keep, because the same source data needs a different shape here than it needs on Google.

Key Insight: One catalog, per-channel payloads. The title that wins a Google auction and the title that survives a TikTok placement are different renderings of the same product record — built at the feed layer, not by rewriting the storefront.

Product Sets as Segmentation

Catalog campaigns default to "All Products," and an unsegmented catalog hands the delivery algorithm the same instruction an empty custom label hands Google: spend wherever conversion comes easiest. The segmentation lever on social is the product set — a filtered slice of the catalog a campaign can target, built from catalog fields.

Which means the sets are only as good as the fields. Segmenting by margin tier, seasonality, price band, bestseller status, or clearance requires those distinctions to exist in the data — as custom labels or structured fields the set filters can read. Merchants who encode them run flagship and clearance as separate campaigns with separate budgets and separate ROAS expectations. Merchants who don't run one campaign where clearance cannibalizes flagship delivery, and read the blended result as platform underperformance. Same algorithm, different inputs.

TikTok, Pinterest, Snapchat

TikTok: two connections, not one

TikTok runs two distinct product-data surfaces and merchants routinely conflate them. Catalog ads — Video Shopping Ads, GMV Max — read a catalog through Ads Manager and behave like the Meta model: ID matching, variant structure, image compliance. TikTok Shop is a commerce surface with its own listing data, category-specific requirements, and content-paired selling. The same product needs different data shapes for each, and treating a Shop listing export as an ads catalog — or the reverse — imports each surface's assumptions into the other's failure modes.

Pinterest: the planning surface

Pinterest shoppers arrive earlier in intent and stay longer — planning purchases, not scrolling past them. Product Pins live or die on imagery depth and accurate categorization, and catalog freshness matters because a Pin saved to a board keeps surfacing for weeks; a price that changed since the save is a trust break at exactly the moment intent matures.

Snapchat: the low-contest catalog

Snap's dynamic ads read the same catalog architecture — item IDs matched to Snap Pixel events, variant grouping, square-first creative. For a catalog already engineered for Meta, Snap is marginal effort and thinner competition: the same payload discipline, pointed at an auction fewer of your competitors bother to enter.

The Catalog Checklist

The checklist below compresses the guide into three working tools: the four platforms at a glance, the catalog audit, and the failure triage — enough to pressure-test a social catalog setup in under an hour, starting with the ID match check most teams have never run.

The four platforms at a glance

• Meta — Reads: Commerce Manager catalog. ID matching key: content_ids in Pixel/CAPI must match catalog id. Where the catalog shows up: Advantage+ catalog ads, DPA retargeting, Shops.
• TikTok — Reads: Ads Manager catalog; separate listing data for Shop. ID matching key: sku_id / catalog id matched to event content. Where the catalog shows up: Video Shopping Ads, GMV Max, Shop listings.
• Pinterest — Reads: Catalog feed → product Pins. ID matching key: item id; item_group_id groups variants. Where the catalog shows up: Shopping ads, organic product Pins, boards.
• Snapchat — Reads: Catalog via Snap Ads Manager. ID matching key: item id matched to Snap Pixel events. Where the catalog shows up: Dynamic ads, collection ads.

Catalog audit — the fields that decide delivery, matching, and retargeting

• id ↔ event content_ids (Critical) — Retargeting exists only where event IDs match catalog IDs — mismatch breaks DPA silently. Self-check: Test event per platform resolves to a live catalog item
• item_group_id (Critical) — Variant grouping — which color/size renders, and whether variants compete. Self-check: Every child carries the group ID plus its own color, size, and image
• image_link (Critical) — The catalog is the creative — quality and compliance decide delivery. Self-check: Per-variant images; no overlays, watermarks, or borders; 1:1 available
• additional_image_link (Rec.) — Lifestyle and alternate angles — Pinterest performance. Self-check: Lifestyle imagery present; not the same shot five times
• title (Critical) — First words carry the placement — surfaces truncate hard. Self-check: Product identity in the first 40 characters
• description (Rec.) — Detail surfaces and Shops; read by category matching. Self-check: Written for a shopper, not pasted SEO copy
• price / sale_price / availability (Critical) — Trust and checkout integrity — drift kills quality scores. Self-check: Promos and stock-outs never show stale in an ad
• google_product_category / platform taxonomy (Rec.) — Category matching; TikTok category-specific listing rules. Self-check: Most specific node; Shop rules checked per listing
• product sets / segmentation (Rec.) — Campaign segmentation — margin tiers, seasonal, clearance. Self-check: Sets built to a documented strategy, not 'All Products'
• link (Critical) — Landing consistency — ad product = page product. Self-check: Deep links resolve to the exact variant

Failure triage — five breakages, fix direction

• Retargeting audiences tiny or empty — Check event-to-catalog ID matching first. If content_ids don't resolve to catalog items, view/cart events build nothing.
• Wrong variant or image in the ad — Audit item_group_id integrity and per-variant image links; children showing parent data means the grouping is broken.
• Catalog items rejected on image policy — Strip overlays, watermarks, borders, and collaged shots; re-run the affected set, not the whole catalog.
• Stale price or stock in live ads — Tighten sync cadence; audit sale_price windows — promo start/end drift is the most common source.
• Delivery concentrating on a few SKUs — Segment with product sets and separate campaigns by margin or priority; an unsegmented catalog spends on the easiest conversions.

A checklist finds the breakages. Fixing them at catalog scale — IDs reconciled across event integrations, variants structured with their own images, titles and descriptions rendered per platform, segments encoded into the fields product sets filter on — is rules-engine work, and it's the work GoDataFeed does across every social platform from one source catalog. The fastest way to see where yours stands is to look at what you're actually sending.

1. Meta Commerce Manager catalog and Advantage+ catalog ads documentation
2. TikTok Ads Manager catalog and TikTok Shop seller documentation
3. Pinterest catalogs and shopping documentation
4. Snapchat dynamic ads documentation
5. GoDataFeed internal error-pattern and compliance history