How Often to Update Predictive Models
Everyday new information is being created in your business. Your customers are buying more, subscribing or unsubscribing, and before you know it your customers today are seemingly different than the customer you had the day before.
As these new patterns emerage its important to periodically take time to investigate your data, update your predictive models, and challenge the assumptions about your business going forward. But how often should you do this? To answer that question, consider the following:
- How often is my data changing?
- How often do I plan on making decisions with the data?
To understand how often your data is changing, its necessary to understand what type of data you are gathering and how often. The average number of customer transactions per year can be a good place to start. For example, if on average your customers purchase from you once a month or once a quarter, it may be possible to identify new patterns in their behavior on an annual basis, which would only require you to update your predictive models and assumptions annually. Predictive models primarily based on customer demographics are good examples of this because a customer’s demographic profile tends to change little over time, often times not even annually.
Contrastingly, if your customers have more frequent transactions (i.e. daily, weekly, bi-weekly), customer data is growing much quicker in size and complexity. The patterns created as a result of this new data are also more dynamic, which means you need to update your predictive models and your assumptions more often.
Knowing how often you plan to make decisions with your data also bounds some of the uncertainty around updating your predictive models. For example, if it only makes sense to focus marketing efforts or strategic initiatives quarterly, there may not be a need to do more frequent updates beyond those time intervals. However, not updating your predictive models in preparation for those initiatives can be dentrimental as well because of the likelihood of having obsolete assumptions.
Finding a happy medium between these two points is what we look for when making suggestions to our clients about updating predictive models, but regardless its really up to your discretion.