Product network
Product network is a map of your products. Products that people often buy in the same order are joined by lines, and products that tend to go together are grouped by colour. Below the map, Starting products shows which products bring the buyers who spend the most later on.
Open My Shop → Analytics → Product network. There are two tabs: Network (this page) and Product analysis, which looks at one product at a time. The map uses the last 180 days of orders and shows up to 200 products.
How to read the map
| What you see | What it means |
|---|---|
| A photo | One product. |
| Bigger photo | More connections to other products. Under Size by you can size photos by Product orders instead. |
| Same-colour ring | A group of products that are often bought together. These groups are called communities. A grey ring means the product has no connections on the map. |
| An arrow from A to B | Of the orders that had product A, how many also had product B. This percentage is called the attach rate. Thicker arrows mean a higher rate. |
Arrows, with an example
Say there's an arrow from tripod to phone holder marked 40%. That means 4 in 10 orders with a tripod also had a phone holder.
The two directions are often very different. The arrow back from phone holder to tripod might be only 5%: most people buying a phone holder don't want a tripod. So the tripod is the product to suggest the holder with, and not the other way round.
Each product shows its 3 strongest arrows to start with. Change this under Outgoing per product (3, 5 or 10). To see plain lines without direction, open Filters and switch Edges to Same-basket overlap. (On some shops this switch sits in the toolbar instead, labelled Directed and Co-purchase.)
Finding your way around
- Find a product searches by name.
- Click a Community button to zoom in on one group.
- Drag to move around, and scroll or pinch to zoom. Fit network shows everything again.
- Hover over a product or a line for a second to preview it, and click to keep its details open. Press Esc to close.
- Statistical evidence hides pairs that could just be coincidence. Significant keeps pairs we're fairly sure about. Strong evidence is stricter and keeps fewer.
- The Minimum attach rate slider hides arrows below the percentage you set.
The details panel
When you click a product, you see its orders, its community, and the other products bought with it, strongest first. Attach rate next to each one is how many of this product's orders also had that product. If your filters hide some connections, Show all brings them back. Product analysis opens this product's buyer journey.
When you click an arrow or line, you see both directions side by side, for example tripod → phone holder 40% and phone holder → tripod 5%. You also see Affinity vs. expected. Popular products end up in the same order now and then just because they're popular. 1× means a pair turns up together about as often as popularity alone would explain. 3× means three times as often, which points to a real link between them.
Starting products
This section looks at each buyer's first order with you, and everything they bought afterwards, order by order. DataGlass rebuilds this record every Monday, and the section appears once the first one is ready. The Period at the top shows which order dates it covers.
A product start is a first-time buyer who had this product in their first order.
Contribution means profit: the price minus product cost, platform fees and the product's own ads, at today's prices and costs.
| Section | What it tells you |
|---|---|
| Where do buyers start? | The 10 products people most often buy on their very first order. |
| Highest contribution LTV | Profit per buyer in their first 180 days, later orders included, for each first product. (LTV stands for lifetime value: what a buyer is worth to you.) |
| Long-run repeat activity | Which products your returning buyers buy most over many future orders, whatever they bought first. (Bought this again on Product rebuy is different: it only counts first-time buyers buying the same product again within 90 days.) |
| Where retention matters | First products where many buyers don't come back within 180 days, and where winning back a few more would earn you the most. |
| Product reach & time | How likely, and how soon, buyers who start with one product go on to buy another. |
| Compare starting products | All starting products in one table. Click a column, such as LTV · 180d, to rank by it. |
A † next to a number means some product costs are missing, so it only counts what's known. — means there's no number. If you don't have the Sensitive financials permission, money columns are hidden.
- Line thickness in the direction-free view is orders with both products divided by orders with either one.
- Statistical evidence uses q-values, which correct for testing many pairs at once: Significant is q ≤ 0.05 and Strong evidence is q ≤ 0.01. Each pair also needs a minimum number of shared orders, shown on the page.
- Size by → PageRank sizes products by how central they are in the whole network.
- Communities are found from the full network (Louvain method) and don't change when you filter.
- Network diagnostics, below the map, summarise how quickly repeat buying fades across the whole shop. Spectral radius is how much repeat buying shrinks from one round of purchases to the next, and the half-life line marks the round by which it has halved. You don't need these to use the page.
- Highest contribution LTV only ranks products with at least 30 starts and costs known for 95% of their sales.
- Contribution leaves out VAT and shop-wide ads that aren't tied to a product.
Use cases
- Build bundles and add-ons from real orders. Set Statistical evidence to Significant, then look for arrows that are much thicker one way than the other. The product the arrow starts from is the main item, and the one it points to is the add-on. Set them up as an add-on deal or a bundle deal.
- Choose what to show next to a product. In a product's details panel, the products with the highest Attach rate are the ones to feature on its page or in your shop decoration.
- Plan collections and campaigns. Each community is a natural set of products for a shop category, a campaign theme or a matching set.
- Pick products to win valuable buyers. In Compare starting products, rank by LTV · 180d. Products near the top are worth putting ad spend and new-buyer vouchers behind, even if their first order is small.