Right now, with widespread recognition of the importance of relevant communications, the ‘360 degree customer view,’ is a frequently heard piece of marketing speak. It’s also a misleading piece of s*** if all you want to do is predict the buying behavior, risk, and product interests of each customer so you can properly target campaigns or personalize product offers and recommendations.

The idea behind ‘360’ is that with all the data companies are collecting about their customers, it’s now possible to build customer profiles that will identify which customers are good prospects to buy a given product.

The process goes like this: Pick a product and compile a list of all the customers who have bought that product. Then, from their customer profiles, assemble a list of the attributes these customers have in common. Use logistic regression to decide which attributes are most important. Finally, find other customers with these attributes at the necessary strength and pitch them the product. In theory these customers should be ‘just like’ the customers who already purchased the product, and so should want to purchase it themselves.

In practice, there are many problems. This approach requires a massive data integration project.  The time, labor, and costs are enormous because an approach like this needs data sourcing, access, cleansing, enrichment, transformation, schema modeling, documentation, storage, maintenance and on-going governance.

Then if you were able to gather the necessary data, you run headlong into an even bigger obstacle – this process doesn’t scale. You need to build a new regression model for each product for which you want to find appropriate customers. You might build ten or even forty models, but if you have thousands of products, this is not a feasible approach. And that’s assuming the models have some longevity, which they often don’t.  They need to be rebuilt after some weeks.

To make this process work, companies are using bigger computers and hiring statisticians and data scientists. Yet despite these efforts (and costs!) they are not delivering relevant communications at scale. There are too many regression models to build and too much data to integrate and maintain.

Fortunately there is a better way to go – Loyalty Builders. Instead of building regression models to handle the most popular products, we can predict the buying behavior of each of your customers. We use the years of transaction data for all customers, which you already own, to calculate a myriad of powerful and accurate predictions about when any given customer will buy, what they will buy, or if they are at risk. This transaction history is an under-exploited goldmine for predicting the future purchase behavior of each and every customer. You actually don’t need anything else to make very accurate predictions.

Purchase predictions based on transaction data are much more accurate than those based on demographics, other behaviors, or other customer attributes. Customers vote with their wallet, not their zip code. With this “Little Data” approach of getting the most predictive value out of just a few (but most important) fields of transaction data, the data integration and modeling problems disappear (along with the need for all those IT and data science projects).

Further, Loyalty Builders predicts the propensity of every customer to buy every product you offer, even if you have tens of thousands of products. You are not limited to offering only your most popular products (the ones for which you’ve built regression models). Accurate predictions and relevant communications can span your entire customer population and total product inventory, all at a significantly lower total cost of ownership.

We know this can be hard to believe for companies that are not Loyalty Builders customers. Skeptics are everywhere but among our customers. That’s why we grind our teeth at night when we hear marketers talk about ‘customer profiles’ and ‘a 360 degree view of customer,’ at least when it comes to predicting loyalty and product purchases. We know it’s just not required for that.

It’s not the 360 degree view you need, folks; it’s a 180 degree turn, from product-centric regression models to customer-centric predictions based primarily on transaction data.

Tags :