eBag: how we counted, and how to set it up
For the analyst or merchandiser who wants the counting rules behind the eBag numbers, and the four placements and few rules that produced them.
- Every definition behind the numbers on the case study
- The setup in the store, step by step
- What to check in your own data
Open the method and setup notes
Enter your work email and the notes open on this page. We also send you the link, so you can come back to it.
Thanks. The notes are open below.
We use your email to know who reads our research. Privacy policy
What was measured
Period and population
eBag, an online grocery in Bulgaria on a custom platform, a client since 2019, using one part of Releva: recommendations on its website. No email, no push, no ads. The main window is 1 June to 28 September 2026: 693,666 orders. The trend uses June 2024, June 2025 and August 2026. Influenced traffic is January to August 2026. In September 2026 the store took about 180,000 orders a month, about 13 items each.
What counts as a result
An order with a recommended item: 210,119 of 693,666 orders, 30%. Orders per 100 clicks on each placement. Share of recommendation revenue by placement. The share of online revenue attributed to recommendations, 4.0% for 2025.
How an order is credited
An item counts when the shopper clicked it in a recommendation and then bought it. One order counts once in the 30%, however many recommended items it holds. Revenue by placement is the value of items clicked in that placement and bought.
Controlled or observed
Attributed, not measured lift. eBag has no holdout group. Influenced visitors opened at least one page through Releva and chose to click, so part of the conversion gap, 40 to 53% against 2 to 3%, is intent. The trend across three years is a trend, not a controlled result: other things changed too.
What the numbers do not say
April and May 2026 are left out: a counting error credited 5.7 items per order instead of the usual 2.2. Conversion is the share of a month's visitors who bought that month, not a per-visit rate.
The terms on the case study
| Term | How it is defined here |
|---|---|
| Order with a recommended item | A completed order holding at least one product the shopper clicked in a recommendation. Counted once per order. |
| Orders per 100 clicks | Orders that followed clicks on a placement, per 100 clicks: 20.7 on the cart page, 14.8 on themed rows, 15.8 on category pages, 5.2 on product pages. |
| Share of recommendation revenue | The share of all revenue attributed to recommendations that came through one placement: 68% cart page, 13% home-page rows, 12% category, 8% product pages. |
| Influenced traffic | Visitors who opened at least one page through Releva: 10% of visitors and 69% of buyers, January to August 2026. |
| Conversion | The share of a month's visitors who bought that month: 40 to 53% for influenced visitors, 2 to 3% overall. |
| Themed row | A home-page row the team names and filters once; the agent picks the products inside it per shopper. 372 launched since January 2025, 16 live. |
What was switched on, in order
The catalog and the orders connected
Products with availability and language, and every order, so the agent knows what each shopper buys and what is in stock.
Cart page: your usual items
Before checkout, items from the shopper's own past orders that are not in today's cart. eBag's busiest placement.
Home page: themed rows
The team names a theme and sets one filter; the agent picks the products for each shopper. 206 rows ran in the window.
Category and product pages
"Popular in this category" and "People also bought".
Few rules
Only available products, the shopper's language and one theme filter. No manual boosts on price or brand.
To check it in your own store
Pull your orders for a window and flag each one that holds a product the shopper clicked in a recommendation. The share of flagged orders and the orders per 100 clicks per placement are the two numbers to compare with eBag's.
If you want lift rather than attribution, hold back part of your visitors from the cart block and compare their orders with the rest. Releva sets that holdout up automatically.
Back to the result, or see it in your store
See which of last week's visitors you missed
Thirty minutes, with your store and ad accounts open. Then 30 days free. A holdout group decides: if the agent doesn't add orders in 30 days, you don't pay.
Book 30 minutes
You'll talk to one of us.