AppLovin feed for Shopify

AppLovin's Catalog manager offers three data sources: an inferred catalog built from pixel events, a file import and a direct Shopify connection. The file route is a custom CSV: you put the catalog at a publicly accessible location and AppLovin downloads it on a daily basis, syncing the catalog stream once per day. AppLovin's own words for why you would choose it are that CSV stream catalogs give you finer control over which products are in your product catalog.

Feedflock hosts that CSV on a stable public URL and regenerates it as often as every hour, so the daily sync always reads current prices and availability. CSV is the only format AppLovin names for the file, and it publishes no item or size cap. It also publishes no validator or diagnostics page, so the practical check after connecting is the catalog's own item count and preview in Catalog manager once the first sync has run.

AppLovin has no catalogue of its own to pair your products with. It joins the feed to shopper behaviour, on two ids at once: id must match the item_variant_id in your pixel, and itemId must match the item_id. If they do not match, the file still imports and the ads still serve, but purchases cannot be attributed to products. This is the one thing to verify before anything else.

Feedflock's defaults are Shopify's numeric variant ID for id and the numeric product ID for itemId, the choice that is unique by construction. Compare them with what your pixel actually sends in its events, and if your pixel sends something else, such as the SKU, switch the id mapping to match. Every SKU is its own row, and rows sharing an itemId are treated as variants of one item. The webUrl is pinned to the variant, and AppLovin requires that page to have the pixel running and fire a view item event.

Most CSV dialects protect a comma inside a value by quoting the cell. AppLovin's does the opposite and bans the characters: commas may only separate fields, and no field may contain a comma, a newline or a carriage return, because they cause processing issues. Feedflock therefore writes the file without any quoting and replaces a comma or line break inside a value with a space, so a description with a comma can never shift a column.

Where a field holds several values, they sit in one cell separated by semicolons: that is how additional image URLs are sent. AppLovin's own Catalog CSV template also settles the details the prose leaves open, such as the column order and the uppercase TRUE and FALSE for Booleans, and the template follows it. Valid image extensions are .jpg, .jpeg and .png, and AppLovin's sample image URLs carry a query string after the extension, so Shopify's CDN addresses fit the pattern.

The catalog has no currency column: price is a bare number, and currency lives on the pixel's events instead. The sale price is the trap. AppLovin documents that a salePrice is not active when both of its dates are empty, and that if you include salePrice, both salePriceStartDate and salePriceEndDate are mandatory. Shopify's compare-at price has no dates, so the familiar habit of sending the full price plus a sale price would produce a sale that is documented as inactive, and an ad showing the higher price. Feedflock's default therefore puts the price you charge now in price and leaves all three sale columns empty. If you keep a real sale window, map all three together, and never let salePrice exceed price.

The rest of the vocabulary is leaner than a Google-style feed. The category is categoryId, an integer holding Google's category ID, so the feed's category mapping should send the ID form. There is no GTIN or any other barcode field, no condition and no shipping element of any kind. Fields Shopify has no source for, such as videos, ratings, age group, gender, rewards and labels, are in the template without a default, and a fresh feed simply does not write them until you map them.