Marketing Decisions has been developing machine learning/AI prediction solutions for business outcomes for our Eloqua clients since 2014. Our first prediction model was predicting likelihood that a web enquiry would result in an appointment booking at a specialist medical practice. Those most likely to make an appointment, according to the predictor, were prioritised for outbound call follow up.
Marketing automation provides an ideal base platform from which to start predicting business outcomes. This is because it tends to ingest a range of data from demographic, behavioural and integrated data from other systems such as CRM. It also provides data automation and data transformation possibilities. Whilst marketing automation vendors are focussed on AI to optimise email deliverability for example with send time and subject line optimisation and fatigue analysis, currently there is no possibility to focus AI on predicting business outcomes. Some CDP solutions offer AI prediction modelling and behavioural intelligence across data ingested in the CDP, but these tend to be very high-cost solutions, still requiring substantial custom configuration effort.
This is where the Marketing Decisions’ AI Predictor steps in. Our data experts have fine-tuned how to define, gather and transform data in marketing automation to feed into the Predictor. These same consultants understand the mathematical models of the Predictor and know how to tune the model for optimised predictions. Today we have models working across a range of industries e.g. predicting likelihood that a web enquiry will convert to a software sale for a software company, predicting best channel to send messages through for a specific conversion event for a charity, predicting which students will be most likely to accept an offer if made for a university. We currently only take on prediction projects where we feel confident that our modelling will safely predict 3 out of 4 predictions correctly – this is our current minimum success metric.
So why build out prediction? It enables an organisation to best focus resources e.g. to convert prospects and it also gives forward visibility into business insights. We can provide analytics on a database of leads with their prediction scores in marketing automation and derive a sense of future business volume. The Predictor is self-learning as more actual outcomes are fed into it. The model will adjust to take account of actual outcomes vs predicted. This is a major advance on current marketing automation lead scoring which is static and generally based on “gut-feel”.