Order attribution
Recommendation widgets
Item-level attribution Retail Rocket
An order is attributed to a recommendation widget if it contains at least one product with which the website or mobile app visitor interacted through the widget within 24 hours.
Interaction is understood as a click or adding to the cart directly from the product card in the recommendation widget.
For grouped products, interaction with any of the products in the group is taken into account. For example, a user can click on a recommended t-shirt of one color, and order another — this product will still be considered sold with the participation of the recommendation widget.
Main characteristics:
- Attribution window: 24 hours;
- What is attributed: Only products with which the user interacted through the widget;
- Attribution retention: Attribution is retained even if the user left the site and returned via a direct link or from an external source (e.g., a search engine).
- Goal: A "strict" model that provides an accurate measurement of the real contribution of recommendations to a purchase.
Full-check attribution
Full-check attribution considers the entire order if the user interacted with any product through the recommendation widget within 24 hours, even if that product was not included in the final order.
This model simplifies the comparison of the effectiveness of Retail Rocket algorithms with other marketing tools that use the same model.
Main characteristics:
- Attribution window: 24 hours;
- What is attributed: The entire order, regardless of whether it contains a product that was interacted with;
- Attribution retention and reset:
- Attribution is retained when returning via a direct link or from payment service pages;
- Attribution is reset when returning from an external source (e.g., a search engine);
A buyer clicked on a phone case in the recommendation block and left the store. Returned within 24 hours after the click and placed an order for a screen protector and headphones. The phone case is not in the order, but we will account for the screen protector and headphones in this attribution model.
Differences between models
| Item-level attribution | Full-check attribution | |
|---|---|---|
| Attribution window | 24 hours | 24 hours |
| What is attributed | Only products with which the user interacted through the widget If the buyer clicked on a phone case in the recommendation block and, in addition to it, added a screen protector and headphones to the cart, only the case is attributed. | The entire order, even if the recommended product is not in it If the buyer clicked on a phone case in the recommendation block and left the store. Returned within 24 hours after the click and placed an order for a screen protector and headphones. The phone case is not in the order. In the new attribution model, we will account for the entire order amount: screen protector and headphones. |
| Attribution reset | After 24 hours | After 24 hours or upon returning from an external source |
| Goal | Accurate measurement of the impact of recommendations | Simplification of comparison with other marketing tools |
Email newsletters and web push notifications
Main concepts
Retail Rocket can identify the same visitor on different devices. This allows taking into account cross-device scenarios: for example, a user could open an email on a phone and place an order on a computer; or buy a product in an offline store after viewing an email. Thanks to this approach, the system links the order to the original mailing even when switching between devices.
Below are the main principles of attributing orders that came through email newsletters and web-push notifications:
- Attribution window: 6 months from the moment the user interacts (clicks) with a push notification or email. That is, the 6-month attribution window works as follows: if an order is placed within six months after clicking from an email or push notification and there were no transitions from other channels (other mailings, pushes, advertising, etc.) between the click and the order, then the entire order is attributed to this email.
- Binding by visitorId (CDV): Retail Rocket links all user actions by a unique visitorId, combining them into a
Cross Device Visitor(CDV). This allows tracking the buyer on different devices. - Attribution retention: Attribution is retained if the user returns to the site directly (via a direct link) or from a payment page. Such scenarios are not considered a channel change.
- Attribution reset: If the user switched to the site through another channel (e.g., from search, social networks, or another email) before placing an order, the initial attribution is canceled. This means that the new source "overwrites" the previous one.
- Automatic statistics: All data for such orders are automatically calculated in the Retail Rocket system based on the combined profile (CDV), provided that the necessary data is available.
Case examples
1. ROPO (Research Online, Purchase Offline):
When loading sales data from offline stores, Retail Rocket determines how much recommendations and communications contributed to offline purchases.
Example: A client receives a personalized email newsletter with recommendations, clicks through to the website from it, and then browses products on the site. Several hours or days later, they visit an offline store and buy a product (e.g., a TV), providing their email or using a bonus card. Retail Rocket links this purchase to the previously viewed email by the user's common visitorId.
The order is attributed to the email channel because the user interacted with the mailing before the purchase.
For a marketer, this is important because it shows the contribution of email campaigns to offline sales (ROPO), allowing them to evaluate the effectiveness of mailings and take into account offline purchases when analyzing channels.
2. Cross-device: viewing from one device, ordering from another
The system links user actions on different devices, providing a holistic view of the path to purchase. This is especially important for modern shoppers who start their journey on one device and complete it on another.
Example: a user receives a mailing and views it on a mobile device (e.g., opens the email and clicks on a product). Later (e.g., the next day) they visit the site from another device (e.g., a computer) and place an order. Retail Rocket combines user actions on different devices by their visitorId (CDV), so even if the purchase is made on another device, it is attributed to the original email newsletter.
The order is attributed to the email channel thanks to cross-device identification (CDV).
This is important for a marketer because it allows for a correct assessment of the effectiveness of mailings without losing data when the user uses different devices.
3. Built-in browsers of email clients (Gmail, Mail.net)
Retail Rocket attribution handles the peculiarities of email clients like Mail.net, Yandex, and Gmail, which use built-in browsers when opening links in emails.
Example: a user opens an email in the mobile Gmail or Mail.net email app and clicks on a product link. The transition occurs in the email client's built-in browser, but the Retail Rocket system records this event as a transition from an email. Then the buyer places an order on the site. Despite the browser being "internal", the transition is counted as coming from the email channel.
The order is attributed to the email newsletter because Retail Rocket registered a transition from the email in the built-in browser. This is important because many users open newsletters through mobile email applications — correct attribution takes into account such transitions and allows evaluating the contribution of newsletters to sales.