Predictive analytics is a hot topic and we are often asked how specifically marketers can use predictions to develop more profitable relations with their customers.
In this post, I’ll give you an overview of 13 predictive models you could use to increase revenues and delight your customers.
There are three types of predictive models marketers should know about:
- Clustering models (segments)
- Propensity models (predictions)
- Collaborative filtering (recommendations)
I’ll go through each and give you a definition, as well as a total of 13 examples:
A. Clustering models
Clustering is the predictive analytics term for customer segmentation. With clustering you let the algorithms, rather than the marketers, create customer segments. Think of clustering as auto-segmentation. Algorithms are able to segment customers based on many more variables than a human being ever could. It’s not unusual for two clusters to be different on 30 customer dimensions or more. We call these dimensions the cluster DNA. See below for an example of some of the factors that could make up a cluster’s DNA.
The most used clustering algorithms are behavioral clustering, product based clustering (also called category based clustering) and brand based clustering.
Predictive model 1: Behavioral clustering
Behavioral clustering informs you how people behave while purchasing: do they use the web site or the call center? Are they discount addicts? How frequently do they buy? How much do they spend? How much time will go buy before they purchase again? This algorithm helps set the right tone while contacting the customer. For instance, customers that buy frequently but with low sized orders might react well to offers like 'Earn double rewards points when you spend $100 or more.
Predictive model 2: Product based clustering (also called category based clustering)
Product based clustering algorithms discover what different groupings of products people buy from. See the example below of a category (or product) based segment or cluster. You can see people in one customer segment ONLY buy sweaters, whereas those in another customer segment buy different types of active wear products, such as outerwear, sportswear, swimwear and watches – but never kids’ clothes, intimates or jewelry. This is useful information when deciding which product offers or email content to send to each of these customer segments.
Predictive model 3: Brand based clustering
Brand based clusters tell you what brands people like. Now you know what specific brands to pitch to certain customers. When a brand releases new products - you know who is likely to be interested. See the example below of brand based clusters. As you can tell, the algorithm has discovered that customers who like Tahari also tend to like Calvin Klein and Nine West, but would not be interested at all in Desigual or 6126.
B. Propensity models
Propensity models are what most people think of when they hear “predictive analytics”. Propensity models make true predictions about a customer’s future behavior. With propensity models you can truly anticipate a customers’ future behavior.
Model 4: Predicted lifetime value
Algorithms can predict how much a customer will spend with you long before customers themselves realizes this. At the moment a customer makes their first purchase you may know a lot more than just their initial transaction record: you may have email and web engagement data for example, as well as demographic and geographic information. By comparing a customer to many others who came before him (or her) you can predict with a high degree of accuracy their future lifetime value. This information is extremely valuable as it allows you to make value based marketing decisions. For example, it makes sense to invest more in those acquisition channels and campaigns that produce customers with the highest predicted lifetime value.
Model 5: Predicted share of wallet
With predicted share of wallet models you can estimate what percentage of a person’s category spend you currently have achieved. For example if a customer spends $100 with you on groceries, is this 10% or 90% of their grocery spending for a given year? Knowing this allows you to see where future revenue potential is within your existing customer base and to design campaigns to capture this revenue.
Model 6: Propensity to engage
A propensity to engage model predicts how likely it is that a customer will click on your email links. Armed with this information you can decide not to send an email to a certain “low likelihood to click” segment.
Model 7: Propensity to unsubscribe
A propensity to unsubscribe model predicts how likely it is that a customer will unsubscribe from your email list at any given point in time. Armed with this information you can optimize email frequency. For “high likelihood to unsubscribe” segments you should decrease send frequency, whereas for “low likelihood to unsubscribe” segments you can increase email send frequency. We talked about this in a recent post on “How to find the sweet spot of email frequency”. You could also decide to use different channels (like direct mail or Facebook) to reach out to “high likelihood to unsubscribe” customers.
Model 8: Propensity to convert
The propensity to convert model can predict the likelihood for a customer to accept your offer. This model can be used for direct mail campaigns where the cost of marketing is high for example. In this case you only want to send the offers to customers with a high propensity to convert.
Model 9: Propensity to buy
The propensity to buy model tells you which customers are ready to make their purchase: so you can find who to target. Moreover, once you know who is ready and who is not helps you provide the right aggression in your offer. Those that are likely to buy won't need high discounts (You can stop cannibalizing your margin) while customers who are not likely to buy may need a more aggressive offer, thereby bringing you incremental revenue.
Model 10: Propensity to churn
The propensity to churn model tells you which active customers are at risk, so you know which high value, at risk customers to put on your watch list and reach out.
Often propensity models can be combined to make campaign decisions. For example, you may want to do an aggressive customer win back campaign for customers who have both a high likelihood to churn and a high predicted lifetime value.
C. Collaborative filtering
The common marketing term for collaborative filtering models is recommendations. These recommendation models were made famous by Amazon with their “customer who liked this product, also liked …” suggestions. There are different types of recommendations.
Model 11: Up Sell Recommendations
Up sell recommendations are typically made to customers at the time of purchase, such as at the time of online, phone or in-store check out. Super sizing McDonalds meals would be a classic example, but examples can be found in all industries. You could suggest a higher end version or a multi-pack of the same product, perhaps at a better price. Up sell recommendations are typically tied to a specific SKU: every product has suggested products to upsell to.
Model 12: Cross Sell Recommendations
Cross sell recommendations are also made at the time of purchase. Rather than recommending buying a larger or better version of a specific product however, cross sell recommendations are made to suggest other products that are typically bought with this specific item. The recommendation could read: “customers who bought this time, also tend to buy …” and you could offer a modest discount if the customer decides to follow your cross sell bundle. Cross sell recommendations also tend to be tied to a specific SKU: every product has suggested products to cross sell with it.
Model 13: Next Sell Recommendations
Next sell recommendations are typically made after a customer already has purchased a product from you and could for example be included in the confirmation email. The best next sell recommendations are specific for each customer and take into account more customer data than just their most recent transaction. An example of a next sell use case was documented in data-driven marketing: a home improvement store found that people who build decks tend to be in the market for a grill shortly thereafter and a program was devised to capitalize on this knowledge.
The first rule of predictive analytics …
Predictive analytics models are great, but they are ultimately useless unless you can actually tie them to your day-to-day marketing campaigns. This leads me to the first rule of predictive analytics: always make sure that your predictive analytics platform is directly integrated with your marketing execution systems such as your email service provider, web site, call center or POS system. It is better to start with just one model, but use it in day-to-day marketing campaigns than to have 13 models without the data being actionable in the hands of marketers.