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, we will give you an overview of 13 predictive models you could use to increase revenues and delight your customers. Predictive Analytics in Marketing can be a real eye-opener.
There are three types of algorithmic models marketers should know about: Clustering models (segments), Propensity models (predictions) and Collaborative ﬁltering (recommendations). Let’s go through each and give you a deﬁnition, as well as a total of 13 examples:
Cluster analysis or Clustering is the task of grouping a set of objects in such a way that objects in the same group (called a cluster) are more similar (in some sense or another) to each other than to those in other groups (clusters). It 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.
The most used clustering algorithms are behavioral clustering, product based clustering (also called category-based clustering) and brand based clustering.
Algorithm 1: Behavioural 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 by 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.
Algorithm 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 activewear 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.
Algorithm 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.
Propensity models are what most people think of when they hear “predictive analytics”. Propensity models make true predictions about a customer’s future behaviour. With propensity models you can truly anticipate a customers’ future behaviour.
Algorithm 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 their 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.
Algorithm 5: Predicted share of wallet
With a predicted share of wallet model, 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, one should know whether this is 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 that can capture this revenue.
Algorithm 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.
Algorithm 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. Given 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. You could also decide to use diﬀerent channels (like direct mail or Facebook) to reach out to “high likelihood to unsubscribe” customers.
Algorithm 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. In this case you only want to send the oﬀers to customers with a high propensity to convert.
Algorithm 9: Propensity to buy
The propensity to buy model tells you which customers are ready to make their purchase: so you can ﬁnd whom 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.
Algorithm 10: Propensity to churn
The propensity to churn model tells you which active customers are at risk, so you know the high value of 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.
The common marketing term for collaborative ﬁltering models are 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.
Algorithm 11: Up-Sell Recommendations
Upsell recommendations are typically made to customers at the time of purchase, such as at the time of online, phone or in-store check out. Supersizing McDonalds’ meals would be a classic example, but examples can be found in all industries. Another example is that of the company Target which became famous for its Pregnancy prediction algorithm based on the food materials purchased from the shop and then recommending the pregnant women resources that need to be purchased at different stages of the pregnancy. You could suggest a higher end version or a multi-pack of the same product, perhaps at a better price. Upsell recommendations are typically tied to a speciﬁc SKU: every product has suggested products to upsell to.
Algorithm 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.
Algorithm 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.