How Idukki decides a sale came from UGC

Any tool can put a revenue figure on a dashboard. What matters is the test an order had to pass to get there. If the test is loose, the figure grows with traffic. Sales can stay flat.

So here is our test, rule by rule. It is the same test on every report, email and admin screen. Our KPI definitions fix it in one place.

Rule one: start from the order

On Shopify, every new order fires the orders/create webhook. It goes from Shopify's servers to ours. The shopper's browser plays no part.

That matters for three kinds of order. A shopper who declines the consent banner silences checkout pixels. Point-of-sale and staff draft orders never touch a browser checkout. The webhook still reports all three.

Checkout pixels still run as a second signal. Their reports are checked against the webhook rows in both directions.

Rule two: find the shopper

An order carries clues that also appear on earlier visits. Idukki tries four, strongest first.

Finding the shopper proves nothing on its own. The widget writes the visitor id when a gallery loads. So it only shows that the shopper saw a page with a gallery.

Rule three: proven means a widget add-to-cart

An order is proven when the same shopper added a product they bought from the widget. The add must fall in the 30 days before the order. The pixel path and the webhook path use this one rule.

A pixel checkout with no matching widget add-to-cart is filed as a store order. It can still land in an assisted tier.

One narrow extension came in on 18 October 2026: same style. Say a shopper adds the blue shirt from a post, then buys the green one. The titles match once colour words are removed, with at least 3 words left. That order counts as proven too. It applies on the webhook path only and carries its own flag.

Add-to-carts from the Guided Shopping Quiz are stored apart. A quiz pick involves no customer post, so it never makes an order proven.

Rule four: assisted orders keep their own column

Most orders that touch UGC never pass the proven test. They land in one of four assisted tiers, strongest first. In the dashboard these add up to influenced orders, with proven.

• Click. The shopper clicked a post for a product they bought. They never added it from the widget.
• Cart. They added a product from the widget, then bought a different one.
• Content. They clicked customer content for some other product.
• View. The gallery showed them a post tagged with a product they bought.

A view also counts when the impression was on the bought product's own page. A gallery that loaded no posts earns nothing. A fingerprint match never makes an order cart-assisted.

Rule five: show the rest, and add nothing together

Conversion rate uses proven orders only. It is proven orders divided by engaged visitors, the people who interacted with customer posts.

Add-to-cart value from the widget is shown as intent. It is never added to any revenue figure. Every definition is on how we measure.

Three orders, traced

In the third order the shopper never added anything from the widget. So the till sale lands as assisted, in its own column.

Fixing the past, in the open

Until 30 September 2026, one pixel path treated the visitor id as proof. That counted orders from shoppers who had only seen a gallery. We changed the rule and relabelled those orders as store orders. They now go through the assisted tiers like any store order.

Shopify lets an app re-fetch about 60 days of orders. Older orders we already store are re-attributed from our own copies, daily. An order with no bridge stays unattributed. Nothing guesses a shopper into existence.

What proven does not prove

Proven shows the post was on the path to the sale. It does not show the shopper would have walked away without it. Only a holdout test answers that. Our guide to holdout tests for UGC shows how to run one.