How Predictive Decisioning Transforms Marketing StrategyJun 4, 2015 - by Neil Capel
Download our Definitive Guide to Predictive Marketing to discover an in-depth look at the 8 predictions changing the game for revenue generation and customer experience.
Predictive intelligence and decisioning are not new concepts, but thanks to advances in data science, they have more applications, and more accuracy, than ever.
Today’s marketer can use data science to generate and apply true predictions that apply to individual users. That’s a big leap forward from traditional methods, in which marketers have created segments of users, hoping that their past behavior will provide some clues as to what they’ll be up to tomorrow.
What’s the difference? Well, it’s the difference between knowing which group of people recently purchased, how many times, for how much, versus knowing that a specific customer is 99% likely to make a purchase within a week (seriously). Even though both sets of information are called predictions, it’s the difference between relying on information about past habits and truly being able to predict yet-to-be-performed actions.
Imagine how your marketing strategy, customer retention, messaging and budget allocations would change if you had this information at your fingertips!
The Science of Predictive Technology, or, How to See the Future
The most effective and popular approach to making predictions has been RFM modeling. To use RFM modeling, marketers try to isolate their best customers by looking at three metrics: purchase recency, frequency of purchase, and monetary value. By applying linear regression analysis to this data, marketers have tried to pinpoint the likelihood that a segment or cohort of customers would take a specific action (such as making a purchase).
If that sounds like a lot of work, it’s because it is. Worse, it’s inefficient in terms of cost, time and accuracy. Stitching together the data means exporting and normalizing. That alone can be expensive and slow. By the time marketers get their results, they’ve often already been superseded by new data coming in. In the end, marketers end up with only a best guess as to which actions a segment of customers might take. But because human action is unpredictable, and because predictions are being made by looking at previous behaviors, the whole thing becomes a bit of a fool’s errand.
Modern Predictive Technology Differs Greatly
The raw materials used by today’s predictive intelligence tools to create an accurate prediction are multiple streams of time series data – successive measurements made over specific time intervals. Think of data streams representing the time the tide comes in each day, the daily close of the S&P 500, or the daily ridership of New York’s subway system.
Using time series data, as opposed to individual data points (i.e. purchase recency, frequency and value), allows data to exist within context and cadence, which enables a better understanding of customer actions. A simple example: a cohort of customers new to your brand may behave very differently than a cohort of high-spending, loyal customers.
Depending on the predictions you are looking to make, these time series may be attached to any of a variety of variables, such as the number of messages opened per day, clicks and even sign-up times. To be accurate, a predictive algorithm needs to look at dozens of these variables – again, over time – amassing hundreds of data points per user.
The guts of a software’s predictive strength lies in its modeling capabilities. The most advanced predictive tools automatically build millions of models every day, and then test those models for accuracy using recent data and actual behaviors.
For example, Sailthru’s predictive decisioning tool excludes the most recent month of customer data, and then examines how well a given model, if fed data up until the most recent month, would have predicted those customer behaviors. When successful, that model is used to make predictions on that given day. In the end, one model per client per prediction is selected as the most accurate—for that particular day. Every 24 hours, the entire process is repeated.
Why does the model have such a short lifespan? Because the world changes. In an extreme case, the model that best predicts user behavior on Thanksgiving will probably not be the one that has the most explanatory power on Black Friday. Other, subtler forms of seasonality may also come into play. Or brands may change strategy or push out discounts. For a prediction to remain relevant, the model needs to adapt.
A data scientist would tell you that in reality it’s much more complicated. That’s absolutely true, but the basics remain the same. Which is why you should ask any potential provider of predictive intelligence the following key questions:
• What type of data is used? Time series or single data points?
• How does the predictive tool differ from existing reporting methodologies, such as RFM modeling?
• What types of models are used and how are they tested for accuracy?
Predictions have the potential to become a powerful new addition to a marketer’s toolkit – as long as we understand what those predictions are really measuring, and how they can best be used.