1. Home
  2. Our Data
  3. Methodology


How do we project sales?

To create our dataset, we use state-of-the-art machine-learning models which combine consumer panel data with the world’s largest truth set of online sales data.

Consumer panels help us identify traffic patterns on retail domains across the home, category, and product pages. We apply conversion rate models based on product-level sales data from more than 200K retailers to predict sales for each product. These bottom-up (from SKU-level) sales prediction models back into predictions of total retailer sales.


Step 1
We have developed several productivity tools, such as Retailer Benchmarking, which collectively create the world’s largest set of online sales data. Additionally, we use public data captured from hundred of millions of products and product category URLs as well as licensed data from leading data providers.
Step 2
To prepare for our modeling phase, we classify page types (home, category and product), categorize hundreds of millions of products and then clean and homogenize the data.
Step 3
We predict sales from the SKU-level upwards by estimating traffic to product and category pages for each URL and applying conversion rates as well as the retailers’ price points. These models are continuously calibrated against our set of live retailer data.

Accuracy rate

Overall accuracy rate within Grips for all countries and all categories:

Conversion Rate Average Order Value Sessions Transactions Revenue
95% 96% 70% 71% 72%

Truth Data Source

Here, we present a detailed overview of the total coverage we possess for the selected countries. With our extensive truth data in these regions, we ensure a robust foundation for calculating accuracies and providing valuable insights to our clients.

Country Reported total ecommerce market [2021] Grips tracked revenue Grips tracked revenue on sku level
United States  $960,444,000,000 10.08% 2.12%
United Kingdom  $177,465,300,000 17.70% 6.18%
Germany  $127,500,000,000 28.50% 11.29%
France  $92,710,000,000 11.06% 3.92%
Netherlands  $30,600,000,000 30.73% 23.94%
Poland  $28,900,000,000 33.30% 25.72%
Brazil  $26,100,000,000 74.28% 47.75%

This represents the largest available, global sku-level dataset to date.

Updated on 2023-08-04

Was this article helpful?

Related Articles