What's Next for Your Customers?
Marketing is evolving from communication to the masses to communication to the one. By using a recommendation engine, you can determine the next best step for each customer journey and use that to increase customer loyalty and revenue.
- By Troy Hiltbrand
- February 3, 2017
In the world of marketing, there is an increased focus today on better understanding customers and marketing to them individually rather than seeing them as a mass of characterless faces. Being able to identify your customer and isolate the right next action for an individual can significantly increase customer loyalty and ultimately improve the customer journey, which leads to increased profits.
The challenges are how to determine a customer's next best or next likely action and what to do with that knowledge once you have it. The secret lies in analytics. You need to let your data tell you how to assist that customer through the next step in his or her journey.
How Recommendation Engines Are Changing Marketing
In the past, it was very difficult to isolate the next step that an individual customer needed to take because rules had to be explicitly defined for each interaction in the system. Tailoring these rules to individuals would have been extremely impractical. With the increase of computing power and the application of statistical modeling as part of this process, this is no longer an untenable obstacle to overcome.
Now, instead of hard-coding rules around the steps in the customer journey, these rules can be dynamically generated based on the steps that other customers have taken. The secret is to identify groups of related data and identify how an individual customer fits.
The set of technologies used for this process is best known as a recommendation engine and is based on linking customers to other customers with similar patterns of behavior and using this linkage to recommend a next action. Many major companies have employed technology that excels at providing dynamic recommendations in areas such as which product to purchase next, which song to listen to next, which news story to read next, and even which action to take next.
There are two main types of frameworks for recommendation engines: content-based filtering and collaborative filtering. Content-based filtering is simpler to execute, but it does less to truly personalize the customer journey.
Content-based filtering is based on the attributes of the product or action. Based on the similarity of attributes between items, the recommendation engine can ascertain what products or actions match those that the customer has already experienced. It can provide the customer recommendations such as "because you bought X you might also want to purchase Y."
One of the shortcomings of content-based filtering is that you need to identify this set of attributes for each product or action. This can be laborious.
Collaborative filtering is more complex but it can surface much more useful information about what the customer should do to be satisfied with the next step in their journey. Collaborative filtering comes in two varieties: item-to-item and user-to-user.
In item-to-item, the idea is to isolate other items that people who enjoyed this item also enjoyed. Unlike content-based filtering, this does not rely on the defined attributes of the item itself but on what other items were liked by people who liked this item. This can provide "customers who liked this also liked these items" recommendations.
User-to-user takes it one step further and tries to group individuals together into highly refined segments to make recommendations. This type of recommendation identifies how closely related the actions of two individuals are and then makes recommendations based on what similar customers have done and liked in the past. This can provide "customers like you have also liked these items" recommendations.
Both item-to-item and user-to-user enable refined recommendations about the next step a customer should perform with the goal of increasing the customer's overall satisfaction with the journey.
Many companies use rating information to enhance collaborative filtering. This requires that a portion of the customers actively provide feedback about their experience with a product or service. This could include providing a rating on a scale of one to five, giving something a thumbs-up or thumbs-down, or indicating whether they are happy with the product. In the absence of actual ratings, a surrogate rating could be utilized -- based on the number of times the customer purchased the product or if the product was returned due to dissatisfaction.
This crowdsourcing of characterization can reduce the burden of manually defining attributes for each product or service but it requires you to convince customers that it is in their best interest to participate.
Once you have the recommendation for the next best action, the next challenge is to get it into the customer's hands. It can be delivered as an integrated part of a customer's experience with your service, for example as a list of other products that the customer could put in the shopping cart. When the customer is external to your service, you might send the recommendation through email offers or recommendations.
The power of the recommendation engine is that it creates a very personalized next step for the customer. As a customer acts on these recommendations, the overall experience is improved, loyalty increases, and ultimately your profits increase. Marketing to the individual is becoming a key differentiator for successful organizations. The secret to your success is utilizing analytics to power the process.
Troy Hiltbrand is the chief information officer at Amare Global where he is responsible for its enterprise systems, data architecture, and IT operations. You can reach the author via email.