Machine learning can be used to drive your customer experiences across channels. That means that rather than establishing a single experience with the same messaging and content on your website, in your mobile app, or in each email you send, you can rely on algorithms to modify the experience for each individual, leveraging everything you know about each person to do so.
That concept can make any digital marketer sweat a little. Every marketing team spends a substantial amount of time planning and executing a good experience for its customers and prospects in its digital channels. The specter of losing control — of not knowing precisely what that experience will be for each person — is understandably unsettling.
But the concept is not unlike the customer experience that your company delivers in real life. The experience each person has — in a store, on the phone with a call center agent, speaking with a salesperson, interacting with a customer success manager, etc. — will differ from person to person. And as a marketer, you can’t completely control those experiences because they are delivered by individual humans.
Let me explain the parallels between machine learning-driven experiences and human-to-human interactions with an analogy. For the purposes of this blog post, imagine you are the leader of a customer success team for a technology company.
Transparency and Control Over Algorithms
Any good leader of a customer success team knows that training is critical. Because while you can’t control each customer’s experience, you can provide training to the customer success managers (CSMs) to ensure service excellence.
You give your team guidance about the tone they should use with customers, how to answer common requests, etc. Then you empower them to learn all they can about their customers in order to personalize their communication based on how each customer prefers to communicate, his or her level of experience with the product or service, etc.
When it comes to machine learning-driven personalization, the “training” comes in the form of machine learning strategies or “recipes” which you build out. To do that, you pick a base algorithm (or multiple algorithms) that the personalization will be driven by and then customize it to make it maximally relevant at the one-to-one level. For example, you can adjust your algorithmic recipe to exclude or include certain criteria (e.g. categories, keywords, etc.) to give you more control over the type of information that is shown, and then add boosters based on the preferences and affinities of each individual to completely personalize his or her experience. (For more detail on this topic, check out “The Main Elements of Machine-Learning Personalization.”)
A good personalization platform will give you this level of transparency and control over your machine-learning personalization — because not being able to control your algorithms is just as ridiculous as throwing your CSMs into their jobs without any training or oversight.
Check Experiences for Individuals
Once you’ve trained your staff, you want to be able to ensure that they can deliver a good customer experience before you put them into the field. You may do a few role-playing exercises with the team, where you play the part of a customer with a problem to see how the trainee responds. With this approach, you can correct any issues you see before the trainee goes "live" on the phone.
You should be able to do the same thing with machine learning — check what experience an algorithmic recipe would show for a specific individual on your site, in your app or in your emails. A good personalization platform will allow you to gut-check those experiences to see if they make sense for that real individual’s established preferences and affinities. If it doesn’t make sense, you can alter your algorithmic recipe until it does.
Test and Tweak Algorithms
Finally, as a leader of a customer success team, you are always striving to create a better experience for your customers. If you find that one CSM has a good approach to dealing with a specific client problem, you will coach the whole team on what works for that CSM. You continue to strive to provide a better experience for your customers and to make them as happy and successful as possible.
You can do the same for your machine learning-driven experiences as well. You can (and should) regularly A/B test your algorithmic recipes, find what worked and what could be improved, and continuously iterate to provide better, more relevant experiences across channels.
It can be scary for many marketers to think about losing control over their digital experiences when those experiences are determined by algorithms. But it just requires a different way of thinking.
When a member of your company delivers an experience to a customer in the physical world, you can’t personally control how that experience goes. But you can provide training to the staff delivering those experiences, you can check in to make sure those experiences are going well, and you can continue working with the team to ensure that they keep striving for success.
The same concept applies to your machine-learning strategies with a good personalization platform. To learn more about how Evergage can give you the control you need to deliver human-guided machine learning, request a demo today.