How does the clustering work? What even does Bidbrain?

If you are around Bidbrain, chances are that you have heard or read the word ''clusters or clustering'' but what is it?

Many customers ask: what is clustering and what attributes are Bidbrain using to create these clusters.

To give a quick answer to ‘’What is clustering’’:

Clustering is when we group different products together. Products are getting matched based on shared attributes. We like to call them ‘’sibling products’’. Once Bidbrain has managed to find enough sibling products to group together, then a cluster is created.

Why are we creating these clusters?

Quick answer: to find potential in bad performing products and elevate good performing products.

Since launching Bidbrain we started to notice that almost all our customers focus only on their good performing products. And therefore left a huge part ‘’untouched’’.

By finding similarities in products, grouping them together and ultimately looking for profitability in their combined historical data, we have seen both elevation in good products, but also activation on so-called forgotten products.

This has led us to increase conversion value and lower ROAS at the same time, in one case we increased the conversion value by 400+ % in just two months.

To give you a clear path of action:

Bidbrain receives your feed
Bidbrain scans your product feed after related or similar products
After finding different groups with similar products the Bidbrain looks into your combined historical data after what we call ‘’ profit max’’ or ‘’Sweet spot’’, this is to find a bidding level.
Once that is done we live the bids.
Once live Bidbrain then change the clusters according to performance, in some cases even deletes clusters. Or sometimes just moves products around. (This is part of our calibration period)

What attributes are you using while clustering?

Basically your entire feed. But the most important attributes are:

Article ID

Name

Price

Product_type

Regular price

Brand