How are the Insights created?

Gal El Al
Gal El Al
  • Updated

For each customer, we collect data from the target eCommerce platforms and product types (product categories) in 3 parallel methods:

  1. The public product pages (PDP)
  2. Search results using predefined keywords for each product  type
  3. If applicable, get the customer’s catalog and sales info from the eCommerce platform vendors’ portal

These data sources are used in various ways to create different insights:

 

Sales and Market Share:

For each eCommerce platform, we identify public data “signals” from the product page that correlate with the sales volume, inventory level, best seller ranking, # of reviews added, search results ranking, and other indicators. 

For each customer and his relevant product categories, the machine learning algorithm takes these signals and, using the details of the actual sale (“Ground Truth”) from the vendor portal and builds the model for each category; This model then calculates the sales details for the products in the category using the public data signals.

 

IMPORTANT: customer’s data from the vendor portal is used only for calibrating the model for that customer and is not used or shared, directly or indirectly, with other customers

 

Category Insights and Digital Shelf:

Public information from the product page, as well as search results ranking are integrated with the enriched sales and market share information from the ML model to provide full, 360, overview on all relevant product categories


Catalog tracker (SKU tracker):
Catalog tracker focuses on the customer SKUs, in all eCommerce platform they are sold, the Insights in this module are created by combining actual sales information, and margins from the customer’s vendor portals access and the public data collected from the product pages

Coming SOON: 
In-store

Similar to the online stores, sales information from the brick and mortar stores is calculated using ML algorithm, based on the public information collected from the various stores' websites, using zip code to scan all applicable locations. The calibration of the model is done using the actual sales information (“Ground Truth”) from the customer account in the vendor portal