Calculating Attributables

Overview

Algonomy defines an attributable sale as a sale that follows a click on Algonomy recommendation within a specific period called a lookback window. This period, a set time interval between the click and the purchase. Lookback windows are configured on individual sites, defined by a number of hours in which the sale must follow the click to trigger attribution. This setting can be found in Site Configurations of the Omnichannel Personalization dashboard.

The configuration of a lookback window is either:

  • A positive number of hours prior to the purchase that a click is eligible for attribution

    • OR

  • A value of -1, which means that attribution can only occur in the same day, in the time zone of the site.  For example, a click at 1AM followed by a purchase at 2AM counts as attributable, but a click at 11:59PM one day followed by a purchase at 12:01AM the next day is not attributable.

Attributable Calculation Data Flow

To calculate attributables, we process a day's purchase data in a batch along with clicks for the last 15 days. The batch is scheduled to execute twice: once early in the morning for the prior day’s data to compute preliminary results, and an additional calculation in the evening for the final results, in the event that some data was delivered late and not included in the first batch.

Attributable Calculation Logic

  1. Purchases for the day are de-duplicated based on the OrderID provided. If we receive purchase data for the same order more than once, we keep only the first order.

  2. Purchases are joined (compared and aligned) with clicks based on the following identifiers:

    1. External User ID: The provided user ID from the merchant's instrumentation
    2. Session ID: The provided session ID from the merchant's instrumentation
    3. UserGUID:  An ID that Algonomy stores in the cusomter's browser cookie
  3. For each of these identifiers, purchases are compared to clicks where the identifier and the product ID match. For any one of the identifiers, if a match is found where a click on a given product ID occurs prior to a purchase of the product ID and the corresponding click is within the configured lookback window, we track the order and product purchased as an attributable sale. 
  4. Use the matched up data to calculate various things:
    1. Total attributable product purchase dollar amounts and counts by product
    2. Total attributable order dollar amounts and counts
    3. Strategies, page types, and placement location 'shares' of attributables based on clicks. For example, if a user clicks on a product twice in two different strategies and buys the product, each strategy gets half credit.
    4. Other internal metrics sliced by genre, catalog, strategy, page type, etc.  For example, attributables sliced by MVT.

This logic is used to invoice customers that have an Attributable Pricing Model.

The following reports on the Omnichannel Personalization dashboard display attributables: