What is order-anchored attribution? Which post sold the bag, step by step

Order-anchored attribution credits sales starting from the order your store has already recorded. The store sends each order to the attribution system, server to server.

The system then looks for identifiers on that order that also appear on earlier activity. A visit, a tap on a customer video, an add-to-cart from a gallery. If it finds no link, the order still counts, as unattributed.

So the difference is in what can go missing. In a pixel model, an order the pixel never saw does not exist. In an order-anchored model every order exists. The open question is how much of it the content can claim.

The gap pixels leave

A checkout pixel is JavaScript that runs in the shopper's browser at the end of checkout. On Shopify it has three blind spots.

• It needs the newer checkout. App pixels run on Shopify's current checkout. Orders outside it are invisible.
• Consent can switch it off. Shopify gates pixels on the shopper's privacy choice. Decline tracking, and the pixel stays silent.
• Some orders skip the browser. Point-of-sale orders and staff-made draft orders never touch a browser checkout.

The store in the opening had plenty of orders and zero pixel checkout events. No gallery change could have fixed that report. It was built on the wrong end of the sale.

How an order finds its way back to a post

Here is the path from a gallery view to a credited order in Idukki, on Shopify.

The visitor id is the strongest bridge, but not the only one. Idukki tries them in this order:

What "proven" means

An order is proven when the same visitor added a purchased product from a customer post. That add-to-cart must fall in the 30 days before the order.

The visitor id alone is not enough. The widget writes it on impression, so it only says the gallery loaded for that shopper.

The same rule applies whether the pixel or the order webhook reports the sale. A pixel checkout with no matching add-to-cart is filed as a store order.

Proven, influenced and unattributed: why the columns stay apart

Adding every kind of credit into one headline makes attribution meaningless. Idukki keeps the buckets apart, with fixed definitions.

Conversion rate follows the same rule. It is proven orders divided by engaged visitors, the people who interacted with customer content.

Our July 2026 case studies report checkout click-throughs, and say so. The order join was not live when those numbers were pulled.

One order, traced end to end

1. 1Day 1, 8:40 pm. Maya opens a canvas tote bag page on her phone. The gallery loads and writes her visitor id into the cart.
2. 2Day 1, 8:42 pm. She watches a customer video of the bag packed for a weekend. She taps Add to cart on the video's product card.
3. 3Day 1, 8:43 pm. She declines the cookie banner and closes the tab.
4. 4Day 3, 7:15 am. She comes back from an email and buys the tote and a keyring. The checkout pixel stays silent.
5. 5Day 3, 7:15 am. Shopify sends the order to Idukki. Her visitor id is on it.
6. 6Result. Same visitor, tote added from a customer post two days earlier. The order is proven.

Now change one thing at a time. Here is where the same order would land.

Under a pixel-only report, every row in that table reads the same: nothing. She declined tracking, so the order was never seen.

What you see in the dashboard

• Proven orders and revenue in their own tile.
• Influenced orders by tier: click, content and view.
• Store orders seen, so you can check the total against Shopify.
• Unattributed orders, shown, never hidden.
• Estimated matches flagged wherever the fingerprint bridge was used.

What it cannot do

It cannot prove cause on its own. A shopper who watched a customer video may have bought anyway. Only a controlled test answers that, which is why proven and influenced stay apart.

History has a reach limit too. Shopify lets an app re-fetch about 60 days of orders. Older orders Idukki already stores are re-attributed from its own copies. An order with no bridge stays unattributed. The model does not guess a shopper into existence.

Questions to ask any attribution report

• Where do the orders come from? The store, or a script in the shopper's browser?
• Are unattributed orders shown? If not, you cannot tell how much it missed.
• Are proven and influenced kept apart? One blended figure hides the difference.
• What counts as proven? A page view, a click, or an add-to-cart?
• Which matches are estimates? Fingerprint matches should be labelled.
• What is the lookback window? Idukki uses 30 days.