Chopping Cost per new Customer with statistical campaign clustering

SUMMARY

Treatwell, the leader in Europe for booking beauty treatments with more than 45K  salons and 100M yearly managed reservations, engaged Booster Box with the task of finding an innovative solution to reduce the Cost per New Customer Acquisition.

The UK and German accounts, managed with Search Ads 360, were hard to control and optimize because of their complex structure. The team came up with a statistically proven clustering method utilising an internally built Python script. As they wanted to take into consideration both volumes and performances, they chose Impressions and Cost per Conversion for new users as variables. The clusters were plotted according to this data and the script returned statistically sound tiers of campaigns with similar CPAs and volumes. There were huge gains in terms of conversions and efficiency: -19.6% CPA and +236% conversions for the British account, -10% CPA and +235% conversions for the German account.

THE CHALLENGE:

To improve budget management smoothly and reduce drastically the Cost per New Customer Acquisition for the British and German markets .

THE SOLUTION:

We took the last 30 days campaign data, taking into consideration Impressions and Cost per Newbie user.

The next step was dividing the campaigns into different clusters based on volumes and impressions, so as to have balanced clusters where the team could apply the same tCPAs.

This allowed allocating the budget among the top performing campaigns belonging to different categories and accounts.