At its best, personalization should be simple and logical. To make it work, you need access to big data. But more importantly, you need artificial intelligence (AI) to make smart use of the data. 

My previous article looked at some of the ways AI could transform content management. That got me thinking about how AI could be used to personalize better, whatever the size of the data you have. 

SARA, an Assistant with Empathy

AI has made inroads. At last month’s World Economic Forum Annual Meeting in Davos, Switzerland, researchers from Carnegie Mellon University unveiled their Socially Aware Robot Assistant (SARA), a virtual personal assistant that helped attendees find sessions matching their interests and useful people to meet.

SARA starts by asking you how you are feeling today, then frames the conversation to match perceived moods and emotions. This lends a human touch to service interactions. If AI truly moves towards more interpretation and empathy, I could well talk to a robot about matching me to my ideal holiday home.

But robots have not taken over. In the absence of ultimate personalization in the form of thinking and talking robots, you need other ways to personalize and target. When you want to drive heavily into the market, you need the analytics part, targeting customers and addressing segments. For personalization to work and be scalable, you need big data.

Tame the Data

Where would you get all this information? Where could you buy information about visitors? You could connect to customer data platforms (CDPs), which let marketers build unified customer databases that can be accessed by other systems. CDPs are a growing market (expected to reach $1 billion by 2019), but they don’t come cheap at close to $100,000 a year per installation. 

And there are tools like ThinkAnalytics, one of the most deployed real-time content recommendation engines, which harnesses big data to increase media usage. This tool uses social media findings, e.g. Facebook “likes” and Twitter trending topics, to provide personalized recommendations for millions of users.

Don’t fall into the trap of sitting on your data: 94 percent of marketers know how essential personalization is, but 95 percent of big data remains untapped.

One of the toughest nuts to crack is integrating the data, connecting all bits of the customer profile so you can identify individuals correctly across the channels they use and the different databases they appear in. 

A lot of data tends to be siloed and it’s a challenge to piece everything together to personalize in real time. Big data can be too much, too heavy or just plain noisy. But let's use AI to turn that raw data into music and say, compose a perfect algorithm for renting a holiday home. 

Listen, Learn, Move to Changing Tunes

Start by focusing on how visitors are using your site — pages viewed, terms search for, click scoring, referrals. Use cookie trails and study the numbers in your analytics tool. 

For instance, if someone develops a module to personalize content based on my user ID in the system, they could figure out that if I’ve clicked from my mobile device on the weather forecast for the next fortnight in this region, they could show me a page with holiday homes in the area. 

Personalization based on what visitors do on a site is great, because you’re not simply banking on demographics but getting close to behaviors. Double up by applying AI to find patterns in these behaviors, look for repetitions in patterns, and continually adjust to reach me with the right message at the right time through the most appropriate channel. 

Don’t get scared by big data. It’s not “big” in the sense of ploughing demographics from thousands of customers. It’s more about connecting real-time information about individual customers across different touchpoints and tracing and shaping their journeys. 

Learning Opportunities

AI learns about me each time I return, and churns up even more tailored and personalized webpages based on my past, current and changing interests.

Live Research

Use AI to track and monitor your personalized content. A/B testing can be time-consuming as you have to set up different versions of content, e.g. how would Boris react to a page with more action verbs versus one with more emotive adjectives, would Boris click on winter landscape images in dark tones or with more sunshine. 

AI can collect the numbers from these experiments more efficiently and collate the results across many more individuals to derive patterns of behavior.

Step up Your Personalization Game

When starting to personalize, 63 percent of marketers go for a rules-based approach and 22 percent really target individually according to Evergage's 2016 Trends in Personalization. In a rules-based approach, you would set up visitor segments and define rules for when to show them what content, e.g. Boris’s IP address is from Switzerland. Let’s show him holiday homes within a 100-mile radius in Switzerland, France, Germany, Austria and Italy. 

That’s fine and you’re probably already applying AI in a rather blanket but scalable mode.

But why not use AI better to join the few who advance to truly individualized content? In an algorithm-based approach, you would focus less on segments but more on individual behavior, e.g. Boris has clicked nine times on holiday homes in the Alps and three times on lake-side holiday homes. Let’s show him different price categories of mountain holiday homes, and throw in some ski or hiking packages as well. Your personalization becomes more finely targeted and more thoughtful.  

Analyze and Predict

Use AI to continuously analyze the information and predict likely interactions or customer journeys. You could even realign future interactions based on the differences between predicted and actual responses.

The AI tool might have noted that Boris usually goes for higher-end holiday homes, so it serves up such choices. But one day, Boris breaks out of the routine and picks a budget home in the remote outskirts. Are his preferences changing or was it a one-off decision because he wanted more peace and quiet? 

AI tends to uncover the “what,” but not the “why.” Maybe an AI chatbot could step in to talk Boris through his stay. To better understand why customers make certain choices, AI needs to be smart enough to constantly review and adapt. 

Big data is nice to have, but you need to build the algorithms and connections for it to make sense. If you don’t have access to data, start with using simple AI to grow the data that you have and personalize smarter. 

Learn how you can join our contributor community.