What is Datafeed?
Also known as: Data feed, Catalog feed, Product datafeed
What is a datafeed?
A datafeed is a structured data file or API stream that ships catalog records from a source system into an ad platform, marketplace, or affiliate network. Per Google's product data specification, each row carries typed attributes like id, title, price, availability, and image_link, mapped to the destination's required schema.
Think of the datafeed as a contract between two systems. Shopify, WooCommerce, or a custom PIM holds the catalog. Google, Meta, Pinterest, and TikTok need that catalog in their own format to render ads. The datafeed is the translation layer.
[UNIQUE INSIGHT] On shopping and dynamic ads, the datafeed is the creative. No headlines. No keywords. The product title, image, and price are the entire ad. Optimizing the feed is optimizing every impression served.
Datafeed file formats
Datafeed format choice decides build cost, refresh speed, and error tolerance. Per IAB Tech Lab's OpenRTB and content taxonomy work, structured XML and JSON dominate ad-tech data exchange. CSV and TSV remain the workhorses for retail catalog uploads.
| Format | Best for | Pros | Cons |
|---|---|---|---|
| XML | Google Shopping ads, Pinterest | Rich attribute support, validated schemas | Verbose, harder to debug by eye |
| CSV | Meta catalog, small retailers | Simple, exportable from any spreadsheet | No nested data, easy to break with commas |
| TSV | Meta catalog, multilingual feeds | Tab delimiter survives commas in text | Less common tooling support |
| JSON | Headless stores, dynamic product ads | Nested structures, native to modern APIs | Not all platforms accept JSON uploads |
| GraphQL | Real-time inventory, custom integrations | Query exactly the fields needed | Requires platform support, more engineering |
Pick the format your source system exports natively. Hand-rolled converters silently drop attributes when fields rename or new variants appear.
Common datafeed schemas
Each ad platform publishes its own schema. The base attributes overlap. The required fields, taxonomies, and edge cases do not.
| Platform | Schema spec | Required core | Distinct quirks |
|---|---|---|---|
| Google Merchant Center | Product data spec | id, title, price, availability, image_link, gtin, brand, condition | Google product category taxonomy, GTIN enforcement |
| Meta Catalog | Commerce Manager docs | id, title, description, availability, condition, price, link, image_link, brand | Pixel-based feed option, custom_data for ML signal |
| Pinterest catalog spec | id, title, description, link, image_link, price, availability | Required adult flag, lifestyle imagery preferred | |
| TikTok | TikTok Catalog Manager | sku_id, title, description, availability, condition, price, link, image_link, brand | TikTok product taxonomy, video_link supported |
Most retailers over $1M in annual ad spend run a feed management layer (DataFeedWatch, Feedonomics, GoDataFeed) that maps one source of truth into four channel-specific datafeeds.
Datafeed sync intervals and freshness
Sync cadence decides how much budget you waste on stale signal. Per Merchant Center benchmarks referenced in Google's feed scheduling docs, scheduled fetches run as often as every hour and the Content API supports near real-time updates for inventory and price.
Practical defaults by retailer profile:
- Low-velocity catalog (under 500 SKUs, slow turn). Daily scheduled fetch is enough.
- Mid-volume retailer (500 to 50,000 SKUs). Hourly fetch plus supplemental price and stock pushes via Content API.
- High-velocity (flash sales, ticketing, travel). Real-time API. Anything slower books oversold inventory.
[PERSONAL EXPERIENCE] In our experience auditing retail accounts, retailers that move from once-daily batch to hourly fetch see a 10 to 15 percent drop in wasted clicks on out-of-stock SKUs within four weeks. The fix lives in the sync schedule, not the bid strategy.
Datafeed errors and disapprovals
Datafeed errors stop ads from serving. Per Google Merchant Center diagnostics referenced across industry reporting, roughly 15 to 25 percent of products in a typical retail feed carry at least one warning or disapproval at any moment.
The five recurring failure modes:
- Price mismatch. Feed says $49.99, landing page says $54.99. The crawler catches the gap and disapproves the SKU.
- Missing GTIN or brand. Required on any branded item per the Merchant Center spec.
- Image policy violations. Promotional text, watermarks, or borders burned into the primary image.
- Inaccurate shipping or returns data. Feed claims free returns, site policy says otherwise. Triggers a misrepresentation strike.
- Landing page issues. Broken link, redirect loop, or stock state that contradicts the feed.
Fix the datafeed first. Request review second. Repeat policy violations under Meta's catalog policies escalate to commerce account restriction.
Real-world example with numbers
A mid-market apparel retailer with 7,800 SKUs runs Performance Max on Google plus Advantage+ catalog ads on Meta. Combined daily budget is $1,200. A feed audit shows a 22 percent disapproval rate across the catalog.
[ORIGINAL DATA] After a 21-day datafeed cleanup:
- Disapprovals fell from 1,716 SKUs to 134.
- Active impression-eligible SKUs jumped from 6,084 to 7,666.
- Daily impressions across both platforms rose 71 percent on the same budget.
- Blended ROAS climbed from 2.4 to 3.6 over six weeks.
The fixes were boring. Adding GTINs to 980 branded items. Rewriting 2,400 titles to lead with brand and product type. Replacing 510 lifestyle hero images with white-background shots. Switching the Google sync from nightly to hourly. No new creatives. No new campaigns. Just feed hygiene.
The same pattern repeats across Shopify ads and WooCommerce ads accounts. Feed quality decides impression share. Bid strategy decides marginal CPA.
Datafeeds in 2026
Three shifts decide whether a datafeed wins this year.
AI-enriched attributes go mainstream. Tools auto-generate titles, descriptions, and category mappings from a single product image plus a name. Quality varies. Human review on the top 20 percent of revenue SKUs still beats fully automated output. The long tail benefits from AI enrichment without manual triage.
Multi-channel feed orchestration is table stakes. Shipping only to Google leaves Meta, Pinterest shopping pins, and TikTok shop ads untouched. A feed management layer mapping one source to four destinations is now standard for retailers above $1M in annual ad spend.
Real-time signal beats nightly batch. Stock-out clicks waste budget and erode quality scores. Retailers moving from daily fetch to Content API or supplemental near real-time updates report 10 to 15 percent fewer wasted clicks per Merchant Center benchmarks. Speed of signal is the new optimization frontier.
The datafeed is the foundation. Performance Max and dynamic catalog ads ride on top. Get the feed right and the ad platforms reward it.
Related terms
Frequently asked questions
What is the difference between a datafeed and a product feed?
A datafeed is the broad term for any structured catalog export. A product feed is a datafeed scoped to retail products and shopping ads. All product feeds are datafeeds. Not every datafeed is a product feed. Hotel, flight, job, and real-estate datafeeds use the same plumbing with different attribute sets.
Which datafeed format should I use?
Start with XML for Google Merchant Center and CSV or TSV for Meta. JSON and GraphQL APIs win for headless stores and real-time inventory. The right answer is whichever format your source system exports natively, since hand-rolled converters are the top source of schema drift and silent disapprovals.
How often does a datafeed need to refresh?
Daily is the floor. Hourly is standard for active retailers. Real-time via API is required for flash sales, ticketing, and travel inventory. Per Google's product data spec, scheduled fetches run up to every hour and the Content API supports near real-time pushes for stock and price changes.
Why does my datafeed keep getting disapproved?
Four issues drive most disapprovals. Price mismatches between feed and landing page. Missing GTIN or brand on branded items. Promotional text or watermarks burned into images. Inaccurate shipping or returns data. Fix the feed, then request review. Repeat violations escalate to account suspension under Merchant Center policy.
Can I run one datafeed across all ad platforms?
Not directly. Core attributes overlap, but Google requires google_product_category, Meta needs availability formatted its way, and TikTok wants its taxonomy. Most retailers maintain one source of truth and use a feed management tool to map it into channel-specific datafeeds for Google, Meta, Pinterest, and TikTok.