Many businesses are good at looking at their historical data to get a sense for what happened (and potentially why).  But it’s harder to analyse that data in a way that makes it useful and actionable moving forward.

Predictive modelling is designed to help organisations make use of their historical data in predicting what might happen in the future.  This could be sales forecasting, or estimating how many products to stock, or predicting who is about to churn, or knowing who best to reach out to about a cross or up-sell.  Datamine has been creating predictive and recommendation models for over two decades across a broad swathe of industries and business challenges.  Here is the general process for building such a model:

• Discussion to agree on the desired outcome for the model – what is the model trying to predict within what period of time?  
• A workshop to determine how the client will implement and use the model results in order to make sure they have everything in place to implement (this might be resource, a suitable technical environment to host and run the model or understanding whether any changes to existing processes and workflows are required)
• Exploratory analysis before the model build – this step helps the client understand what’s going on with customers as it relates to say, churn.  This is also the point at which we start to determine what data will be used in the model and is likely to be predictive
• Creation of the model using available data – there are a lot of things to get right, and it’s important here to choose the best variables to consider in the model.  The Datamine team brings their expertise to the table here, helping choose variables that the client has some control over or that matter to their business
• Model validation – making sure the model works by going back in time see if it correctly predicts what actually happened.  This will be an important step in getting the model across the line internally, as clients will have to prove to the organisation that it delivers
• Drive the model to implementation – show the value using control groups, allowing you to measure the impact and calculate ROI