In discussions with clients about Predictive Analytics, the most common question we receive is, “Can you guarantee we will see results?” This is understandable as (1) it is a hard sell internally if there are no results, (2) with limited time and resources, clients want to ensure they are putting their effort into initiatives that will produce results.
When it comes to Predictive Analytics, there is no such thing as “guaranteed results”, but there are steps that can be taken to increase the likelihood of positive results.
- Understand the Business Problem – Not all business problems can be effectively solved using Predictive Analytics. Before starting, you should understand (i) the business problem, (ii) that the business problem is predictive in nature, (iii) whether and a clear path of action exists to solve the problem.
- Understand the Data Requirements – It is important to have the domain knowledge to understand the required and relevant data for the predictive model. Considerations include: (i) number of different types of data and (ii) quantity of data required. Each predictive scenario would be slightly different but in general the more relevant data you have to feed the model, the better chance of success.
- Understand the Modelling Techniques – Predictive Analytics is a very rich domain that allows business questions to be approached from many different angles with respect to predictive modeling. Not all predictive models are relevant to each business scenario, so choosing the correct predictive model is key.
- Agile Execution – Predictive Analytics projects should be executed in an agile fashion. The predictive model should be proven first before building out the full architecture required for your solution. A small initial investment can be used to increase the confidence level of the overall solution.
When executing your next Predictive Analytics project, consider the above steps to be able to see value immediately and increase your likelihood of success.