Solving the Personalisation Gap with Automation
Many advertisers think they’re delivering better customer experiences than ever before. In the 2017 Adobe Digital Insights report delivered at this year’s EMEA Summit, a full 58 percent of advertisers said they feel their ability to deliver great customer experiences is improving.
But consumers don’t necessarily agree. Only 38 percent of customers say brands are actually getting better at giving them the experiences they want. One major reason for this is that nearly one third of retailers—by their own admission—have limited or no capability to support personalisation efforts at scale.
Personalisation carries a host of challenges, it’s true. Many brands simply don’t know where to begin increasing their relevancy. The process seems difficult; no two companies will follow the same roadmap, and that makes it challenging to determine the right key performance indicators (KPIs) and measure success. Even when a brand recognises the importance of personalisation, many of their channels, assets and profile data still remain unconnected, poorly aligned for coordinated action.
At root, though, the concept behind personalisation is simple: Use CRM and behavioural data to know your customers, deliver them the offers, recommendations, and user interfaces (UI) they want, and use their feedback to improve and automate even better experiences.
Here are four simple tactics you can begin using right now, as your organisation develops its own unique strategy along the spectrum from manual to automated personalisation.
Auto-allocate test traffic
A/B testing has come a long way in the past year or so. The latest software can now automatically shift traffic to the best-performing experience in real time, as it learns—so you never have to manually adjust your tests. This is known as the multi-arm-bandit testing approach, because it’s constantly running tests on 20 percent of traffic while delivering optimised experiences to the other 80 percent.
When using this approach, you don’t have to calculate a run time or sample size for your test up front, as you would in a standard testing model. In the old world, you’d allocate a third of your traffic to each of three different experiences—and while each experience would get the same number of visitors, the revenue for each would differ.
In an auto-allocated A/B test, on the other hand, the software automatically redirects traffic toward better-performing experiences—raising your revenue immediately, in real time, as the testing algorithm learns. This speed is particularly useful when you’re running a campaign pegged to a time-sensitive event (like a holiday or a new product launch) and need to discover a winning experience right away.
And yet, while “three-armed bandit” A/B testing is a powerful one-size-fits-all approach, it’s only one of the array of tactics you should be looking at.
Provide collaborative recommendations
Whether you’re in retail or any other industry, you’ve probably got a large collection of digital assets that need to be personalised at scale. You’re looking for a more precise way to deliver relevant recommendations about those assets to your visitors—serving them the products, videos, content, and other assets they’ll actually want to interact with.
One powerful tactic for achieving this is known as_ item-based collaborative filtering_. It’s item-based because it groups items to make better recommendations; and it’s collaborative because it groups them according to an array of factors, including content similarity, popularity, timeliness, frequency, and similarities with attributes in customers’ profiles. You can also add manual rules, based on your own insights—and the algorithm will automatically factor those rules in, too.
As this algorithm learns from your customers, it’ll recommend combinations of assets you never would have thought of, because they might not make sense from a conventional standpoint—but the results will speak for themselves.
Automate delivery of the perfect offer
In a sense, this is the reverse of the previous section—for times when, instead of a huge library of assets that need to be recommended in clusters, you’ve got a few highly significant pieces of content that need to be served to the right users at the right times. When you’re working with a small number of offers like this, full-factorial machine learning serves as a powerful tool for matching offers to visitors.
This style of machine learning uses multiple algorithms (e.g., residual variance, random forest, and lifetime value) to provide ranked offers to visitors based on their profile data. As population behaviour changes, the model rebuilds itself automatically to weight each of the algorithms differently—in fact, it can even provide the option to weight certain variables more highly in response to specific behavioural triggers.
All three of the tactics described above can be highly useful—but the real impact only appears when you use them all in synchrony.
Auto-target on an experience basis
As all these examples make clear, effective personalisation isn’t so much about delivering the right assets as about delivering the right experiences. Your automation solution needs to be able to A/B test not just pages or products, but holistic experiences involving variations in content, navigation, layout, timing, and a host of other interrelated attributes.
We at Adobe recently went through exactly this process with our corporate homepage. Not only did we A/B test variations in content blocks—our software automatically mixed up the entire layout of the page in a wide variety of ways, and A/B tested each variant using the multi-arm bandit approach. The result wasn’t just one clear winning layout, but a range of winners, each optimised for a particular audience segment—an improvement that’s already driven a major lift in revenue.
In short, the key is to use automated personalisation to speak to customers in the ways they want to be spoken to. But tactics like these are just the beginning of a full-powered customer experience strategy. The coming articles in this series will explore a diverse array of other ways you can use data to surprise and delight your customers with great experiences at every turn. If you’re interested in learning more about these personalisation capabilities, then please take a look at Adobe Target—the personalisation engine of the Adobe Experience Cloud—which is where we’re bringing these exciting capabilities to market.