Predictive analytics works by training a machine learning model to identify the outcome you desire.
At MadKudu, we do this with several inputs from you.
Leveraging your historical transactions, win or loss - the model needs to know who to target
Leveraging historical behaviors/activity for leads that convert - the model needs to know when your lead is heating up, or cooling off.
The above brings the volume of data and historical "facts" to the machine so that it can learn. But, we also need to give it guidance on how you want it to score, i.e. who and what are you trying to predict? This differs between customer to customer, but some of the more common conversions to predict are:
Inbound leads highly likely to become a New Business Open Opportunity
Inbound leads highly likely to become a New Business Closed Won Opportunity
Inbound leads highly likely to become a New Business Closed Won Opportunity with a minimum expected spend of $X and New Business
Additionally, many clients have their CRM for many years and as we all know, data clean-up projects are nearly always running. For example, if you had a major data cleansing and harmonization project complete in the recent past, perhaps from that time onwards your CRM entries are most reliable. In this case, you will need to advise us if there is a date range from which we should focus on or ignore.