Using IBM Watson, Xaxis extracted actionable insights from Volvo’s historical site traffic. These insights allowed us to more accurately identify an audience with a higher chance of conversion. Xaxis then created bespoke algorithms for Volvo, designed to optimise all relevant client outcomes, and used mPlatform Co-Pilot to apply these algorithms in AppNexus. This use of machine learning allowed Volvo to optimise faster and to reduce the need for manual adjustment of campaign variables. In the end, it yielded a cost-per-action (CPA) that was 29% lower than the target set by the client.

Together with Mindshare Norway, Xaxis decided to build a custom algorithm that used Google Analytics data as its starting point. This data included all traffic generated for www.volvocars.com/no in the last year. By discovering conversion patterns using this data, we were able to create and deploy a custom algorithm in Co-Pilot, GroupM’s machine-learning platform. This took year-to-date data from Google Analytics and, using IBM Watson found the variables that had the biggest impact on conversion rates.

These variables were then implemented as a custom algorithm written using the Bonsai framework. Using Co-Pilot, the Xaxis team applied the new algorithm to AppNexus, using the AppNexus Programmable Bidder (APB) function in order to find the right inventory through machine learning.

At the same time, the team also used the EMEA Analytics Algorithm to take a different approach, analysing historical performance of the active Norwegian campaign based on AppNexus data. The team ran a predictive model to estimate optimal bidding values for each placement and frequency. We also optimised the campaign for maximum reach. This second algorithm was also implemented through Co-Pilot as a Bonsai tree using the APB function.

The targets for the campaign were 60-65% viewability and a CPA of 400 Norwegian Krone or less.

For the main campaign, the results were:

• 82.4% Viewability

• 283 NOK CPA

The Google Analytics based Algorithm built by the Norwegian team:

• 83.5% Viewability

• 220 NOK CPA

The AppNexus historic data-based Algorithm build by the EMEA Analytics team:

• 59.4% Viewability

• 170 NOK CPA