Resources // Blog

Aligning Machine Learning with Business Rules (for Real Outcomes)

By Kevin van der Straat, Director Global Product, EMEA

Within today’s programmatic advertising universe accountability and outcomes are increasing in priority for advertisers. As part of our mission to provide Real Outcomes, Light Reaction, Xaxis’ performance business, leverages our proprietary Optimization Engine to deliver this mission for our brands. 

On a daily basis, the Optimization Engine receives a vast amount of data signals, and uses Machine Learning, and touches aspects of Artificial Intelligence, to process, value and assess data sets to ultimately determine the value for advertisers. In essence, it allows Traders to build comprehensive and adaptive optimization algorithms for clients with clear outcome goals. (More on Data’s evolving role in marketing and the rise of Artificial Intelligence by Damien Healy, Xaxis COO)

As a Trader and Strategist this has become an ever more important aspect in my role when planning and building guaranteed outcome campaigns for clients. The automation of leveraging client data sets, campaign learnings and behavioral user logic has enabled me to deliver more relevant and optimized advertising.  However, as Machine Learning greatly improves the media buying process, it’s also becoming increasingly more important to align this with the client’s business rules.

What do I mean with business rules? Business rules are the broader setting or ecosystem in which the client practices programmatic buying, inclusive of tactics such as multi-channel (push and pull), attribution and creative proposition. As a Trader the focus lies on defining this setting, together with the client, and activating relevant Machine Learning capabilities. The fundamental alignment of business rules and Machine Learning is unfortunately, not emphasized enough and thought about in the client’s planning phase and challenges arise as a result. So how do we, as Traders and Clients optimize these settings to maximize the potential offered through Machine Learning? Creative proposition and communication:

As a start, with Light Reaction’s Optimization Engine, Clients can instead focus on the fundamentals of creating value through creative variation and proposition rather than buying and optimization. Not saving in this department eventually benefits the outcome as more variation testing and personalized addressable creative messaging can be realized on scale. There are plenty of cases where creative assets are last-minute efforts without any personalization or variation, eventually leading to limited optimization opportunities for Traders and Optimizers.

Channel collaboration:

As Prospecting has a direct influence on the size and scale of a Client’s Retargeting user pool it makes sense to fully align these two outcome-drivers. In cases where Retargeting is managed as a completely separate channel it’s challenging to enable dynamic budget allocation between these correlating efforts. By combining these efforts through Light Reaction’s Protargeting, Clients benefit through our automated allocation and pacing of budget. 

Attribution and maximizing channel contribution:

One of the bigger returning challenges for Traders, and as well for Clients, is attributing the right value to each channel or even impression. Within an accountable and outcome-focused market this, however makes the difference between being successful (reaching optimum conversion volumes and values) or not. This doesn’t mean that all Protargeting conversions require 100% attribution but it does mean that in order to successfully maximize Prospecting and Retargeting efforts, in alignment with other channels, the discussion needs to be alive between the Trader and the Client.

On this topic it’s also fundamental to match practicalities and settings that influence conversion attribution and contribution. These include adapting post-click and post-view look-back windows and cookie life-times. This will foremost allow equal conversion attribution and identify true impact of Prospecting and Retargeting for example.


Although Machine Learning, and adaptable algorithms as part of it, are hyped and sold as the holy grail, the potential uplift in outcome it could offer Clients is evident. It however will be a growing focus for Traders to coordinate with the Client’s Business rules to maximize outcome results through Machine Learning.

This website uses cookies, some of which are necessary for the operation of the website and some of which are designed to improve your experience. You can review the cookies we use here.
If you are happy with the use of these cookies please continue to browse our website. You may adjust your cookie settings via your chosen browser tools.