4 Top Machine Learning Use-Cases for Consumer Marketing

You’ve heard lots of hype about machine learning, but how can it REALLY help you get and edge in consumer marketing? We’ve been doing data science and machine learning for many major brands, and have come up with this list of ideas which really DO work.

In modern digital consumer marketing, gathering large amounts of data about existing and potential consumers is both easy and common. The challenge – and the opportunity for those who can solve it – is in extracting the most actionable insights from that data. 

Machine learning offers powerful new tools for answering questions about your data, but to use it successfully in your business, you need to be sure you’re asking the right questions in the first place. Retina has used machine learning to support marketing initiatives for many major brands, and we’ve discovered 4 use cases in the process that really make a difference:

  • Data-driven Customer Segmentation
  • Customer Journey Optimization
  • Customer Lifetime Value Modeling
  • Digital Marketing Optimization

Data-driven Customer Segments

Every business needs to know and understand its customers. Before machine learning, companies commonly hired firms to survey their customers and write up a persona report to accomplish that. But, this time-consuming, manual process may reinforce existing ideas about your customers instead of grouping them based on real data. More importantly, this approach doesn’t tell you which of your customers belong to each persona, so it’s difficult to take meaningful marketing actions based on a persona report alone.

With machine learning, an algorithm looks at the hundreds of data points you may have about each customer. It examines what they do and what they purchase as individuals to find patterns. Next, the algorithm identifies potentially important “features” about each customer based on all of those individual data points (for example, “clicked on an email on a Wednesday”). Finally, you  can apply a clustering algorithm to automatically group customers into segments using those features.

The result is a truly data-driven understanding of the different types of customers you have. You can then label your actual customers accordingly, shape content and strategy for the most important clusters, and generate digital lookalike audiences based on your best clusters on advertising platforms.

Customer Journey Optimization

You’ve worked hard to build your customer data platform and visualize individual customer journeys, but now what? Just like the individual customer data points above, customer journeys come in a great many forms. However, the main idea is always the same—to get your customers to a specific conversion point with the minimum amount of friction or cost.

A top-down approach to customer journey optimization involves discussing various customer objectives (discovery, comparison, purchase, etc.) within your business and assessing how your various marketing touchpoints influence customers to complete those objectives. This can lead to valuable internal discussions, but without a truly data-driven approach, these conversations can be more of an echo-chamber than a revenue-generating marketing optimization exercise.

A data-driven, bottom-up approach to optimizing your customer journeys must be based on machine learning. An algorithm can examine all of the paths that your customers take and score each path. Scores are based on customer acquisition costs (CAC) relative to the value of each customer’s purchases over the course of their relationship with your brand, called Customer Lifetime Value (CLV). 

One such way to optimize customer journeys is to use reinforcement learning. This takes a sequence of touchpoints and determines which next touchpoint will maximize the chances of a certain outcome, like a repeat purchase. It can identify current points of friction along the customer journey and provide data-driven recommendations for reducing that friction at the lowest possible cost. Insights like these have the power to completely transform your customer acquisition and retention strategies for the better.

Customer Lifetime Value

We’ve already mentioned the CLV metric as a key ingredient in a data-driven  customer journey optimization strategy, but it’s such an important figure it also gets its own section. The idea is that the value of each of your customers is not simply a matter of the purchase they made today, or even of all the historical purchases they have made. A customer’s true value to your brand should be measured over their entire life cycle as a customer—past, present, and future.

Traditional approaches to CLV are based on assumptions about how your customers behave “on average.” This may work in reporting aggregate trends, but it fails to account for the individuality of your customers. Two different customers might interact with the same exact product, but as a result, one may decide to stop purchasing while the other is cemented as a loyal customer. The minute you try to combine your one “average” CLV approach to different types of customer segments, you run into the limitations of the traditional approach.

Again, a bottom-up approach using machine learning models comes to the rescue. First, an algorithm trains on individual-level purchasing behavior, and for a single customer predicts when that customer will stop purchasing, how many more purchases they will make before that, and how many dollars they will spend. This offers an objective and data-driven technique for valuing different customer segments and customer journeys–and adjusting Marketing spend accordingly.

Digital Marketing Optimization

All the use cases above use the powerful tools of machine learning to help you identify your target audience and make their engagements with your brand as smooth as possible. But machine learning can also help with the final step: optimizing your paid marketing strategies and campaigns to attract new customers at the lowest possible cost. Paid marketing is often the highest marketing expense in a modern consumer business *, so any optimization to it can significantly impact your bottom line.

The first approach is to make adjustments based on the applications we’ve already examined. You can identify which acquisition channels drive the highest conversion rates in any of your customer segments. Armed with these insights, you can adjust your spend to better target a new niche, or avoid messaging that does not perform well for another. You can also assess how valuable each paid touchpoint is in a customer journey to calculate how much is worth spending for that interaction. Adjusting spend to be in line with customer value generated by that spend leads to large improvements in maximizing total value.

The second way to use machine learning to optimize your digital marketing efforts is to generate new targeting audiences using Google and Facebook data about users. These new audiences can help to give you an edge over competitors on the same platforms, reducing cost while also increasing ROAS. We’ve written a blog post detailing how this can be done for Facebook, so check it out if you want to learn more.

Wrapping Up

It’s easy to get lost in data science projects that are interesting, but do not deliver real business value. In contrast, these 4 are proven techniques we’ve seen make a difference for our consumer clients time and time again. We hope this list makes a difference for you as well.

If you have any questions about how our experience with this might help you, email [email protected] and we’d be happy to chat.