First step is the creation of new pricing rules that are dedicated to Price Genius only. The minimum number of Price Genius pricing rules is 80, but if you have more free pricing rules in your GAM we usually create more in order to increase the accuracy. Price Genius rules are divided into two groups:

Exploration UPRs (koszyk)

These UPRs are targeted at key values yb_ab=a0,a1,…,a9, and their purpose is to allow to test how particular ad units react to price changes and provide the main data for modelling

These rules are named e.g. koszyk_samp_12.56_a8

Production UPRs (grid)

These UPRs are used for setting the prices predicted by Price Genius on the majority of the traffic.

These rules are named e.g. CO_grid_1.24

In order to avoid overlapping of Publisher’s rules with Price Genius exploration at this stage all existing UPRs created by the publisher get additional targeting: yb_ab<>a0,a1,…,a9

Remember, that each time a new pricing rule is added by the Publisher it must have yb_ab<>a0, a1, …, a9, b targeting applied! Otherwise it will interfere with Price Genius and will negatively impact its performance.

Exploration is the core of Price Genius data gathering process for modelling. It comprises of 40-50 Unified Pricing Rules that are targeted at key values yb_ab = a0, a1, … , a9. You can learn mora about the Exploration process in the article How does Price Genius work?

This part is purely backend - we set up our database and a script that updates the Exploration setup and start gathering data.

It is recommended that the Exploration process is going on for at least 7 days before we can launch Price Genius. The reason for that is that we want to grasp the weekly seasonality of the inventory. Also, the Exploration setup will change in time - Price Genius automatically changes it so as to get the highest possible data accuracy for modelling. See also: When will I be able to see positive impact of Price Genius?