Which banners and what text-boxes work best on the frontpage of the website of an insurance company? Those with two kids on their way to the beach in the summer (it’s about travel insurance)? Those with a young couple (“considering a new insurance scheme – 10 percent online discount”) Or perhaps the one with the businessman (“is your business properly insured”)?

It’s questions like these, Lars Koch Rasmussen is confronted with. Seemingly simple and uncomplicated questions for the Head of Online Sales at Codan, RSAs local Danish brand and the third largest insurance company in Denmark. But far from easy to answer. “In a world where the digital processes are a central part of any business in the service industry, it’s increasingly crucial to be able to attract new customers and grow sales in your online channels. At Codan we found that the frontpage, codan.dk, missed out on engaging customers and leading them through sales funnels on product areas which are strategically important for us” says Lars Koch Rasmussen.


Codan runs AI experiments on codan.dkThe modular frontpage of Codan.dk uses AI and machine learning to understand, what messages in what combinations to show to which customers.

The answers to the questions aren’t trivial either. On the contrary: the answers stem from the intersection of some of the hottest technological trends: Artifical Intelligence (AI), machine learning and solid business savviness.

Machine learning finds the right messages for the right users

In theory, it’s simple: the more relevant the message, the better it performs. Banners and text boxes with travel insurance call to actions works perfect with parents on their way on vacation. Online discounts resonate among those who are looking to swap insurance companies. And the proposition to sign up for a company insurance scheme is likely to appeal to the woman, whose startup is taking off, and now finds a need to consolidate herself.

“The problem is finding out who’s who. Users are rarely logged in, when they arrive at codan.dk. And it is even more rare for us to be able to recognize non-logged in users by other means. At least not in a way, that qualifies his or hers needs in any substantial way” says Lars Koch Rasmussen and continues: ”Luckily we can do without”

Unlock your data

How many visits, pageviews or conversions does your website boast of? Those questions are easily answered by most people charged with the responsibility of running an ecommerce site, thanks to their analytics-systems. Systems, that typically discloses a far greater amount of information on the visitors. Like: What time did the visit take place? Did the user come from a search engine or did he or she come directly to the site? Or: Was the traffic generated from a smartphone or a regular computer?

This applies to Codan as well. But Codans analytics are far more advanced than those of most other eCommerce-sites. Among other things the analytics setup includes an Adobe Analytics Premium installation and an AI and machine learning engine. The AI and machine learning system constantly spots patterns in the analytics systems enormous volume of data. Patterns of timestamps and clicked elements, referrals and bounces. Patterns that, among other things, reveal which segments of users are most likely to react to which specific messaging.

Adapting on the fly

To make things even better, the Adobe installation gives eCapacity and the online team at Codan the opportunity to exploit the artificial intelligence engine to make the site adapt to each user, as he or she enters the site, giving each user the version of Codan.dk most suited for their specific needs.

In this way, the site not only produces qualified guesses at what users will respond most positive to e.g. travel, company insurance messaging, or who’s engaged the most by the possibility of obtaining online-discounts, price-calculation or general sanity checks of the state of his or her present insurance schemes. The page also, automatically, picks the elements to display, that appeals most to the particular user. Thus yielding the highest click-throughs and the best results.

It’s this automated personalization Codan decides to go for. “We were very curious to see, how far the automation and predictive analysis approach could take us” Lars Koch Rasmussen recalls.

60 combinations

As opposed to traditional A/B tests, automated personalization simultaneously tests a multitude of elements – not just two. Even by itself, the frontpage of codan.dk features a dozen or more elements, including banners, teaser boxes and shortcut icons for navigation, resulting in a total of 60 different combinations..

“Before running the test, we had conducted a larger strategic review of the online channels with eCapacity. This had uncovered what elements on the homepage presented the most salient business opportunities for us. Therefore, we had a pretty good idea what part of the page to manipulate in the test” tells Lars Koch Rasmussen.

Surprisingly easy

From there, the work with setting up the test was surprisingly easy. “We put a great deal of effort in aligning our digital and physical channels” continues Lars Koch Rasmussen, “to find the right mix of what to offer customers online, by phone or by personal contact. We worked intensively to create the best creatives to execute our goals. Importantly, we made sure our data-foundation was in place and that our analytics were implemented in a way that allow us to meaningfully measure how we perform against our KPI’s. But the work with AI and machine learning algorithms, that we undertook assisted by the eCapacity consultants, was far easier to deal with than expected. It was surprisingly easy.”

Click-through rates doubled – business effects rapidly materialized

In principle, the personalized automatization never stops. Because the algorithms get smarter and better as time goes by. Because business and online analysts the algorithms tune the algorithms to optimize performance. And because new campaigns, new messages, new product launches and USPs keeps coming along, fueling new impulses to the AI and machine learning engine. But already after the first few weeks, the first results start tricking in. And they’re good.

“The effect of the increased relevancy of the messaging mix is very clear. Some teaser boxes show a 100 percent uplift – click through rates have more than doubled. And the number of users abandoning their purchases before having reached the end of the sales funnels is reduced markedly”, explains Lars Koch Rasmussen. “It gives a really good feeling when your data starts working for you like that.”

Turns insights into solutions, that work

eCapacity helps Codan with both digital strategic consulting, implementation, optimization and analytics. “eCapacity is really good at helping us optimize our business. They’ve helped us analyze our market, our competitors and our customers. They’ve opened our eyes to new insights into both digital and physical channels. They’ve helped us identify opportunities and challenges, and pointed us towards how to grow both our existing customer base and our overall revenue. They’ve implemented our analytics in order for us to track the right KPI’s, and act accordingly. Not least, eCapacity help us activate the data we collect, and turn them into solutions that work. Like they’ve done for us with automated personalization, AI and advanced analytics” Lars Koch Rasmussen ends.

Want to read more? You can find all the eCapacity cases right here.