performance marketing : using machine learning for programmatic

We all thought that programmatic advertising would solve all of marketings problems – personalization, engagement and boom (!)- conversion. Apart from the many problems its faces in execution (transparency, data quality, ad fraud etc) the issue seems to be the lack of focus – more talk-time is spent on targeting and viewability (these are milestone not metrics) while programmatic is meant to improve purchase intent/ aid conversion.
 
Given that programmatic spends will grow to $50 billion in 2017, improving quality in programmatic has to be an imperative if marketing is to benefit from the efficiencies in automation @ scale that it promises.
 
The fault is not with programmatic advertising itself, its in what rules we preset for programmatic – rules that are less focused on contextual logic/ conversion but more blandly robotic in executing using remarketing principles. As simple as it sounds ‘right person, right time, right content’ is anything but.
 
How should we choose the preset rules ? Using machine learning might help us create these rules by keeping conversion as the key goal and allowing for machine learning to match time/ audience demographics/ contextual content / advertising content to goals and determining the patterns that work and those that don’t.
 
2 patterns that machine learning will map for you :
1. Identify your high value audiences : using past behaviors, expressed current signals, probability of conversion and the average Customer Lifetime Value (CLV).
2. Identify your high conversion content-context : the golden combinations of context + content that leads to conversions with the high value audience.
 
Patterns are not static and change over time of year/ influenced by trends and events as well as the entry of new audiences/ exit of current customers. Focusing on conversion oriented goals in programmatic and using machine learning to develop strong conversion models will help sharpen the efforts of targeting-viewability-content creation.  
 
One innovation that is allowing transparency and goal conversion to become the focus is ‘post trade programmatic’ – where the buyer can read 300-400 signals of the delivered impression to determine the accuracy-efficacy of the buy, including targeting and viewability. This then allows for machine learning to kick in, analyse the delivery Vs goals to determine success/ failure paths and optimise the buy every time.
 
Using machine learning to help improve programmatic performance is not the only solution to achieving better quality programmatic but it certainly begins to immediately improve user experience, targetting accuracy, transparency and content choice.

REFERENCE