Personalized User Experience is the Future of E-Commerce
Customer experience is the essence that can either kill or save e-commerce shops. If the optimization of your site is done correctly, your e-commerce shop will thrive. On the contrary, if you fail to get it right, your e-commerce shop will have a limited future.
You need to understand your customers in order to provide an optimal site experience. You also need to understand what’s more important to them. Maybe they want premium or exclusive products? Or maybe they’re in constant search of good sales? Or maybe they prefer sustainable materials? It all depends on the particular customer.
Recommended: A 2020 Retrospective on Becoming a CMO in 2020
If you’re successful in conveying the right message to the right person at the right time, you can significantly enhance the shopping experience of your customers – boosting your revenue and increase the conversion rates.
Now there’s a new tool that makes showing the right message to the right customer even easier.
Exponea, the customer data and experience platform, has developed Predictive Multi-Armed Bandits, a highly effective and efficient solution that can find the right messaging for a particular customer. This can fundamentally change the dynamics of how e-commerce stores chose their messages, leading to increased site optimization.
Multi-Armed Bandits? What are Those?
Exponea has created Predictive Multi-Armed Bandits as a solution to a specific problem: the multi-armed bandit problem. But what is that? Let’s consider a hypothetical scenario: Imagine you’re in Las Vegas at a casino. You have some money to spend on gambling, so you go for the slot machines (another name for a slot machine is a bandit since they take your money). There’s just one problem: you have lots of slot machines to choose from, but just a limited amount of money to spend.
Some slot machines will pay out a little more, and some will pay out a little less.
You have got to figure out which slot machine pays out more and which pays out less as soon as possible because the amount of money you have is limited. You’ve got to find a strategy that balances looking for the right slot machine and exploiting the highest paying slot machine you have discovered so far. This problem (deciding how to spend your money across a number of machines) is essentially a multi-armed bandit problem.
The same problem can be mapped onto e-commerce. Your limited amount of money becomes your customers. The slot machines in this context become your offers or messages (because some messages will attract more customers while others won’t). The real problem is to find which customers are likely to convert after reading a particular message.
Several solutions have been proposed to the multi-armed bandit problem across different disciplines. However, Exponea explicitly focused on the e-commerce domain. It has also focused on introducing context into the equation.
Introducing Some Context
Before I jump into introducing some context, let’s consider our original slot machine example. Notice that when I initially introduced this problem, the only things I paid attention to was the amount of pay-out from each machine and our initial money.
Now imagine that the machines paid out differently under different conditions. Maybe if the weather is nice, some machines pay out more. Or maybe other machines payout differently depending on what the person using them is wearing. In other words, imagine the machines payout differently depending on the context of the person using them. You’d need to change your strategy accordingly.
What this means is that instead of finding only the best overall strategy for all situations, you should consider the context as well.
It’s similar for e-commerce, except there it’s not hypothetical. Rather than trying to figure out the best message for a majority of your customers, you ought to discover the best message for a particular customer, while taking their context (on-site behavior, historical data, etc) into account.
Exponea’s expertise and experience in e-commerce help you determine what sort of context is important to monitor. This knowledge will help your team develop an optimal solution for the multi-armed bandit problem: Predictive Multi-Armed Bandit personalization.
What are Predictive Bandits?
Want to learn more about Predictive Bandits? Watch the link below to see an AI expert and Product Manager explain the new tech in depth.
“You can think of Contextual Bandits as the magical sorting hat from Harry Potter. All you have to do is to make the customer wear that hat, and it will magically tell you which message is right.”- says Robert Lacok (Product Manager at Exponea).
The context contains all the historical data related to a customer, including their purchase history, email interactions, and clicks on the website, in addition to the data related to their current session and data like what the customer usually searches for. It collects all of this information and chooses which variant is optimal based on the given context.
Then the variant is displayed and something happens: either your target audience converts or they don’t. Then the sorting hat – the predictive bandit – uses this action as an input, and uses it to learn all the patterns and correlations between contexts and variants, and determines which variant is likely to influence conversions and which wouldn’t for each customer. With time, it learns to make optimal decisions.
Is A/B Testing Dead?
As of today, using AB testing to determine which variant of some messaging to display is the gold standard, and it still works well…
Going back to my previous example, where 60% of customers were hunting for sales, 35% desired sustainable materials, and 5% were interested in luxury items, an AB test would show that the best performing variant would relate to discounts.
So after running the AB test, I’d deploy that Sales messaging to the entire customer base. But here’s where things get more interesting. When you used the “sales” variant on the entire customer base, you actually misunderstood almost 30% of your customers and showed them content they’re not necessarily interested in.
In other words, you neglected the opportunity to direct your customers towards sustainable products that are likely to have a considerably higher profit margin.
Predictive Bandits can solve these sorts of problems, automatically and efficiently. Predictive Bandits review and reframe the AB test question – from ‘which variant works best for converting the majority of our customer base to ‘which customer segment should see this variant?’.
This is radically different because rather than thinking whether this variant is appropriate for the whole customer base, we are now focused on finding which variant will have an impact on some segment of customers. All we need to do is to find the right segment of customers.
What does this mean for AB testing? Will it replace it entirely? The straightforward answer to this question is ‘NO’. In some cases it won’t work as an alternative to AB testing. If you’re testing larger changes to your site, like a 3- vs 5-step checkout process, AB testing makes more sense.
However, in most standard AB testing scenarios, Predictive Bandits are a real revolutionary change in the industry.
Predictive Bandit Personalization – Exponea’s New Feature
Exponea’s clients can now use Predictive Bandit Personalization on their sites. So what does it look like in practice? Let’s find out.
Assume a scenario where, after conducting a funnel analysis, you discovered that you’re losing too many customers between the “add to cart” and “checkout” phases. For whatever reason, most of your customers are hesitant to finalize their purchases.
So you plan on testing some new messages, hoping to improve your customer conversion rate. You have four variants: one that represents your certifications by the Better Business Bureau. One that says that you work with high-quality and sustainable materials. Another variant showcases your high customer reviews, and the last variant promises a fast and secure payment gateway.
Typically, you would test your variants by evenly splitting your traffic across your variants. You’d have to be patient with the experimentation and let it run a sufficient amount of time. Once that’s done, you’d determine a winning variant and display that to the rest of your audience from there on out.
But with Predictive Bandit Personalization and contextual data, Exponea will automatically find out which customer should be shown which variant.
This means that not only do you get to keep those variants which could result in higher customer conversion, you also don’t have to wait for sufficient customer data to determine a winning variant. Each variant might be a winner for a particular segment. Overall, you can ensure a more personalized and better customer experience for all of your customers.