What’s better for clustering your audience/users, using Behavioral or Demographic variables?
This is a hot topic since the first websites came along. Since I remember companies use to segment their clients and potencial clients mostly using a Demographic criteria, which means grouping people into groups based on age, sex, income, occupation, religion, race, etc.
Historically those variables are classified into Hard, Soft and moment referred ones.
1- Hard Variables: The Hard Variables are the ones that normally never change like the birthdate or name.
2- Soft Variables: The Soft Variables are those that can change from time to time, like level of education or income.
3- Moment Referred Variables: The moment referred variables soft ones are those, as the name propose, that are happening right now, like I’m getting married, I’m starting a new job, I’m looking to buy a new car, etc.
The last ones are that in reality are the low hanging fruit in order to generate conversions. If you know that someone is looking for a specific thing right now, the conversion probability get higher.
Based on the Meta Analytics model, corporate systems (companies) are living systems that are constantly changing (their products, their offers, their quality, their UX, etc), and so do people. So the best way a company can be 100% relevant would be to get an API connection with their customers. But since that’s not possible (yet) you will have to use data to be as closer to 100% relevance as you can.
Behavior segments are those generated based on what people are doing.
Behavioral data allow you to identify what people are doing in a specific time. It’s moment oriented data, but only about what people are doing, not why. Let’s put an example. A user is navigating content or products related with marriage. That information is not telling you that that person is getting marriage , but just that the person is navigated that content. Why is that person doing that? We have no idea, it could be because is in fact getting marriage or because someone else is getting marriage, or just because is interested in the topic.
You can always use predictive algorithms improve your inferences, but the best way to really improve your predictive capabilities is by adding additional data series to the analysis. Ideally, different type of data.
- Behavioral data answers “What users are doing”. Data from Google Analytics, Adobe Analytics, Double Click, etc, are behavioral data.
- Attitudinal data answers “why users are doing what they are doing”. Survey data are the most common attitudinal type of data.
- Tech data answers “how technology impacted the user experience”
- UX data “how our touch points with our clients impacted in their behavior or decisions”
- Demographic data answers “what’s the demographic profile or your clients or users”
Having different type of data allow you to generate more robust decision making scenarios. It’s like having a jigsaw puzzle, the more pieces you have, the better you see the reality. And as in the jigsaw puzzle you don’t need to have all the pieces to see the image. Actually in most cases having just few additional “pieces” of data generates an additional cost that won’t be repaid with the additional “clarity” that brings to the image (decision making scenario).
DMPs are great platforms to aggregate data. They have real time behavioral information that can be integrated with other data dimensions from the same users (or browsers), from surveys (attitudinal) to tech data (connectivity) to behavioral (purchase). But I would rather suggest to have all that on the company side. Data is the most valuable asset and it shouldn’t be hosted or managed by a third party.