Personalisation efforts often start with slicing demographics and trying to understand in which virtual buckets you can place potential customers. Perhaps you look at gender, age, where they’re located, what they earn and occupation – ending up with groups of people for whom you’d like to present customised content.
There is, however, one big requirement for this strategy to be successful. People in the same bucket need to share purchase behaviors and drivers. If not, they are just as randomly distributed as the original group of all visitors.
Just look at yourself. Think about how you would be sliced in a segmentation project.
If you’re reading this at work, you probably share a lot of demographic data with your co-workers. But have a look around. Do you all drive the same car, do you go to the cinema at the same frequency, do you wear the same clothes, do you like the same food and do you travel to the same places. In essence, should a website approach you all in the same way?
If you’re reading this at home. Think about your neighbors. Even though you probably are placed in the same segmentation bucket, do you really have the same purchase intents and drivers?
The key question is (although rhetorical): Is it possible that we tend to overrate demographic data and at the same time underrate one person’s behavior?
Limitations of Demographic Data
There is of course some merit in looking at demographics. A clothing brand may, for instance, decide to present the male or female line of clothes when entering the site. The same company could display different clothes depending on where you’re from. A visitor from Finland may be interested in a winter jacket at the same time as a visitor from Spain is shopping for a summer wardrobe.
It all makes sense. Different people want different things and therefore we try to identify who they are. This is probably why demographic data still is the most collected information by marketers according to CEB’s B2C Marcomm Personalization.
However, demographic data has its limitations. Guy Mucklow, CEO of Triggar, addresses a few important points in his blog post 5 reasons to stop relying on historic data to segment your customers.
When you are placed in a group based on demographics, you are supposed to act the same way all the time. But even one individual has different drivers just depending on what time of day it is.
“They are the same person but have very different needs depending on when, and more importantly, why they are shopping. But as far as the demographic data you have on them goes, they’re exactly the same in both circumstances.” Mucklow says.
Another problem Mucklow addresses is that we only have detailed information about known customers. For new visitors, we might be limited to personalise based on location (IP-address). If that’s the case, elaborate demographic segmentation can only be done for a fraction of the visitors.
“Savvy retailers know that to win more custom they have to refine and manage the customer journey and experience on a micro-level. It is all about the detail … We need progressive approaches that will listen and react rather than guess and assume.” Mucklow sums it up.
Behavioral and Contextual Segmentation
Today, many marketers advocate focusing on behavioral personalisation, preferably in real-time, rather than demographic. Instead of looking at who people are, you look at what they do. And combined with contextual data, such as device or browser used, traffic sources etc., personalisation can become quite effective.
It’s all about refining the on-site experience to suit individual visitors at specific moments.
In this blog post, Peep Laja, Founder of CXL, identifies the following important variables to include when personalising a website. Perhaps apart from location, these are measuring actions and context, not who people are.
- Location – city, country, region
- Device – iPhone, iPad, Android phone/tablet, Windows, Mac, Linux
- Search keywords – did they arrive while searching for shoes or shirts?
- Visitor frequency – First , second, third, fifth time visitor?
- Date and time of day, proximity to payday
- Referring URL – where did they come from?
- Customer history – have bought before, what, how much did it cost?
The next step is to identify patters around how people act when visiting the webpage – navigation clicks, page views, on-site searches, etc. Even though there is an endless array of behaviors, chances are that you’ll find patterns to work with and optimise.
Questions to answer might be: How do we speak to someone who visits the site for 20 seconds, leaves and then comes back? What do we show people who have looked at five pages around pricing? How do we communicate with those who leave items in the shopping basket?
Your analysis will give you the questions for your website.
But the bottom line is: When it comes to interacting with a specific website online, you probably have more in common with someone who acts the same way as you do compared to the person who’s sitting closest to you right now.