Artificial Intelligence
Jan 17, 2022

Predicting Future Spend And Customer Lifetime Value

How Apteo uses A.I. to forecast what your customers will spend in the future.

Shanif Dhanani

Why customer lifetime value (CLV) is important

Customer lifetime value is an important metric that measures how much money you can expect to make from your customers from their first to their last purchase in your store. As a brand owner, your customers with the highest lifetime value will account for a disproportionate share of your revenue. They'll be the small group of loyal and frequent spenders who keep your store growing. If you can figure out who these customers are early, you can nurture them with campaigns, loyalty programs, special perks, and great offers. But that's a big if.

Knowing what your customers will do next is a bit of a holy grail in business. If you knew what they were going to do, your life would be a lot easier. Fortunately, we continue to improve and update our A.I. systems, including one that can predict how much your customers will spend in the future.

How we're different

Note: the couple of paragraphs in this section are geared towards data scientists. If you don't get it, don't worry. The key takeaway is that we use A.I., instead of other techniques, to forecast what your customers will spend in the future.

Apteo has always been a data and machine learning-first company. Our goal is to use the best analysis and data processing techniques to measure and predict important attributes about your customers. Many companies that forecast customer lifetime value do so by using statistical methods, breaking down expected metrics for the number of future purchases and average order value at a generic, aggregated level.

Our approach is a bit different. We work to forecast the total future spend for each individual customer. Ultimately, this is the key metric that everyone cares about. Since machine learning algorithms are great at identifying subtle patterns from large amounts of data, our belief is that these algorithms can do a better job at forecasting future spend in one go than a more statistical, higher-level approach.

Rather than having to model out sub-components of customer lifetime value, we provide learning algorithms with all of the data that we have (and it's important to note that a subset of this data would be used to create the statistical models used by other companies), and the algorithms themselves can identify what's important and what's not.

How we do it

If you've been following our updates, you're already familiar with the basics of the machine learning that goes into many of Apteo's core systems. Last week, we wrote about how we predict the most likely products that your customers will buy next. This system, which we refer to as our product recommendations service, is created by structuring large amounts of data on your store, products, and customers to identify the patterns that are indicative of someone buying a new product in the future.

Our customer lifetime value forecaster actually uses the exact same data that goes into the product recommendations system, you can find that list from a link within the article. In fact, both systems leverage the exact same data transformation process, which provides a structured set of data points (known as features in the world of machine learning) that our algorithms can use to learn how your customers will behave in the future. In fact, the only difference between the two systems is the thing that we're trying to predict.

For our product recommendations service, we use a system that predicts whether or not a particular customer will buy a specific product that they haven't purchased yet. This is known as a classification model - it classifies between multiple different options. In our case, we've actually updated our model to pick between whether a customer will buy a product (represented as a value of "true") or won't buy a product (represented as "false").

However, in our customer lifetime value model, we predict how much a customer will spend in the future - a numerical value. This type of algorithm is known as a regression model. Using everything we know about a customer and your store, we forecast how much a customer will spend going forward. Then we simply add that to what they've spent in the past to come up with their customer lifetime value.

In both cases, we use a gradient boosting model, as it outperformed other models, including linear regression and deep neural networks. In the case of the classification model, we use a gradient boosted classifier, and in the case of the CLV model, we use a gradient boosted regressor.

How we use CLV

While knowing the customer lifetime value of your customers is important, we go one step further. We just released a new segment of customers based on the top 10% of your customers by expected future spend. This segment, which we call Top Future Spenders, provides you with a targeted group of customers that are extremely high-potential for the future of your store. By paying special attention to these customers (with extra offers, loyalty programs, thank you messages, special content, and other perks), you can maximize their potential and increase the likelihood that they'll spend more with your store going forward.

You can use Apteo's existing integrations to sync this segment to your email tools and target them with special flows, or to your SMS tools to offer them text-only offers, or even target them with Facebook ads using our latest Facebook integration. The key is to make sure you take care of these customers, ensuring you build a good, positive emotional relationship with them so they become loyal customers of your store.

What makes a high-value customer

While your highest value customers will likely include many of your biggest spenders and frequent customers, not all of your highest value customers will be existing heavy customers. Since our A.I. can analyze patterns that make a good customer, even after their first purchase, your new segment of Top Future Spenders will also include new(ish) customers that look like they could become your biggest spenders. Every store and brand has different patterns that lead to high-value customers.

There are some common patterns, though. For most stores, we've seen that geography, which products someone previously purchased, how long ago someone made their last purchase, gender, and how long ago a customer made their first purchase are all important factors in determining whether a customer will be high-value or not. Fortunately, every customer of Apteo gets their own, tailored A.I. model, which learns and predicts customer behavior according to their brand's own data.

Future work

As we continue to research new techniques to improve how we can bring A.I. to every ecommerce store, we plan on rolling out several new features. One of the key areas we're researching is explainability, or providing additional insights into why our models predict what they do. For customer lifetime value, we continue to research new methods that we can use to identify which key attributes are indicative of a high-value customer. Going forward, we hope to provide additional reporting in this area.

About Apteo

Apteo is a predictive marketing and analytics platform for ecommerce. We help brands drive sales by helping them understand their customer journeys, personalize their campaigns, and segment their customers with A.I. so they can make their marketing more effective. If you're interested in trying Apteo for your store, you can sign up for free at


Photo by Austin Distel on Unsplash

About the author
Shanif Dhanani
Co-Founder and CEO, Apteo

Shanif Dhanani is the co-founder & CEO of Apteo. Prior to Apteo, Shanif was a data scientist and software engineer at Twitter, and prior to that he was the lead engineer and head of analytics at TapCommerce, a NYC-based ad tech startup acquired by Twitter. He has a passion for all things data and analytics, loves adventure traveling, and generally loves living in New York City.