Swisscom Shows that Fortune Favors the Bold — Especially When Using AI for Personalization

Image source: Adobe Stock / kichigin19.

The Roman poet Virgil once said, “Audaces fortuna iuvat,” or, more simply, “Fortune favors the bold.” That may be good to keep in mind as you reach beyond A/B testing and experience targeting to grow your optimization and personalization program with AI-driven personalization.

That’s the approach the optimization team at Switzerland’s leading telecommunications company Swisscom has taken, and that they shared at Adobe Summit in Las Vegas this past March. There they discussed several ways the company was relying on AI to drive its optimization and personalization work forward. I recently caught up with Nicolas Mériel, senior digital strategist at Swisscom, who has been instrumental in driving this shift to AI to learn more about the AI-driven activities at Swisscom.

In this post, I want to discuss what I learned from Nicolas, including ways that Swisscom is using AI-driven personalization in Adobe Target, and insights he and the optimization team uncovered as they’ve taken the company down this path toward more reliance on AI to scale and grow the program’s success. Indeed, one of the key lessons Nicolas shares is that to win big, you have to be bold and try out big, bold ideas.

By the way, it’s these big, bold ideas that have earned Nicolas and the Swisscom team awards in our annual optimization and personalization contests for EMEA Adobe Summit two years running. This year they won for their AI-powered personalization activities in our ExBE Awards contest, and last year they won for their innovative mobile “shake through” experiment in our Mind-blowing Tales of Optimization contest.

First steps into AI

When Swisscom hired Nicolas five years ago, they asked him to help grow the company’s optimization and personalization program. While they had Adobe Target, they didn’t have anyone responsible for optimization initiatives and using the solution. Once Nicolas joined the optimization team, they started using A/B testing and gradually began using all the capabilities of the Adobe Target solution.

“There was a time not so long ago that there were databases, no data scientists, and no big data or smart data. Then interest in analytics and data started to grow. All of a sudden, data became the new fuel. The new gold. It’s the new thing that we exploit and on which we base many of our business decisions,” he said.

That’s why when Adobe Target Premium launched a few years after Nicolas joined Swisscom, he remembers thinking, “We have all this valuable data, and we have these Adobe Sensei-driven AI tools in Adobe Target. Let’s see what we can do with this.”

The team first tried out Adobe Target Recommendations on the shopping cart page. After seeing success metrics rise on that page, they began using it to recommend products and customer support content and to automate crosslinking between content. The company is now investigating using Recommendations in onsite search and other areas.

Nicolas also wanted to try out Automated Personalization on the homepage, but was waiting for the right time and opportunity to use the powerful automation and AI to deliver personalized offers. At the time, personalization using AI was uncharted territory for the company, and stakeholders weren’t quite ready to fully trust the machine.

Auto-Allocate and Auto-Target fill the gap

Fast-forward a couple of years, and Adobe Target introduces Auto-Allocate and Auto-Target. “There was a gap between A/B testing and Automated Personalization that these two features filled,” Nicolas said. With both AI-driven features seeming like an extension of the A/B testing that the organization was already familiar with, it felt like a natural move to start using them. So just as he had with Recommendations, Nicolas decided to jump on in and give it a try.

Based on the idea that with Auto-Target each visitor gets the experience that wins for them, the optimization team launched an activity with Auto-Target that tested four different hero banners against the default banner. Auto-Target taps into machine learning that uses the Random Forest algorithm. Nicolas shared this activity in his recent session at Adobe Summit, and said that it was already generating conversion lifts in clickthroughs from the banners of up to 40 percent. He expects Auto-Target to be a key tool in the team’s test toolkit moving forward.

Overcoming the “black box” issue and other barriers to adopting AI

Nicolas explains that while AI can deliver incredible results, it’s difficult to pinpoint exactly what aspect of an experience resonated with a given audience and what specific characteristics about your visitors made them more likely to respond to a given experience. It’s the proverbial “black box” issue of AI — you know it works, but you don’t know why, which means you can’t iterate on learnings the way you can with A/B tests.

He says that the mainstage announcement of Adobe Target Personalization Insights reports, currently in beta release and releasing to Adobe Target Premium customers this fall, knocks away much of this barrier to the black box issue. The reports reveal what visitor attributes were most influential in the model that Adobe Target built, and how it grouped customers together into the audience segments it used. Nicolas is excited to use these reports because it lets the optimization team apply learnings from AI personalization. Perhaps as important, the new insights help company stakeholders embrace using AI-powered personalization because they can now see how and why it works.

Nicolas acknowledges that many people fear that with AI they, or some part of their job, will be replaced by the machine. He says that in digital marketing, ideas are key. No tool can help you with ideas — you need people who are creative, and then the tool supports that creativity.

Nicolas observes that while people do like progress, they don’t like change. Shifting to AI is no different. He says, “But when you can prove that you made 50% percent more revenue by delivering the best of 10 personalized experiences to each visitor, the numbers make the decision to move to AI much easier.”

Tips for the content and experiences you use in AI

When using AI, you have to have a lot of experiences—, but not just any experiences. Nicolas advises, “The experiences have to be different enough to provide food for the algorithm to distinguish between noise and signal. They have to be dramatically different in design.,” Nicolas says.

In addition, some experiences have time limits — like back-to-school promotions. Creating all that content for use for just two to three weeks makes that investment even costlier. Meeting the demand for content to fuel personalization with AI can require significant investment.

To address these issues, the Swisscom optimization team is considering dynamic automated creation of experiences by leveraging Adobe Experience Manager Dynamic Media. In addition, they’re exploring how Adobe Experience fragments can further open the content pipeline for personalization in Adobe Target by providing grab-and-go experience building blocks.

Nicolas cautions against creating hundreds of experiences to hyper-personalize to your many visitors. He jokes, “When you’re targeting everyone who lives in downtown Las Vegas who has a cat with three legs and one eye, you’ve reached a point of diminishing returns with your personalization.” He advises finding the sweet spot between this hyper-personalization and delivering everyone the one-size-fits-all experiences so common in the past.

Using AI for omni-channel personalization

Here on the Adobe Target team, we conducted a brief survey to assess how digital marketers are using AI and their ideas for where they can best use it. While only 8% percent were currently using AI to personalize across channels, an additional 35% percent had plans to do so in the next year and a half. A further 41% percent plan to do so, but further out on the timeline.

Nicolas sees great opportunity for omni-channel personalization. For example, when a customer in a brick- and- mortar store interacts in a way that can be digitally captured, when that customer logs in to the Swisscom website or opens the mobile site, Swisscom could push a more deeply personalized experience to them. Conversely, when a visitor to the Swisscom website or mobile site later goes into a brick and mortar store, a sales associate can help them based on knowledge of what they’ve been researching. He emphasizes, “Of course, you have to walk that fine line between being helpful versus being creepy — especially when linking the physical world with the online world.,” Nicolas says.

Final thoughts on AI

Nicolas explains that ideas are always going to be key to succeeding with AI. The technology becomes trivial if you don’t have good ideas, so people will always be needed. And really, that opposition to using AI has much more to do with people’s’ resistance to change. He is excited that Adobe Target is helping address this issue by enabling companies to increase revenue and conversions with AI with features like the AI Insights reports that explain why AI performed so well.

To learn more about how Nicolas and the Swisscom optimization team are using the Adobe Sensei features in Adobe Target, read the Swisscom customer success story or listen to the session at Adobe Summit.