One Size Does Not Fit All – Data Segmentation #eMetrics
Estimated reading time: 2 minutes, 57 seconds
Neil Mason, the SVP Customer Engagement from iJento dives deep into the art and science of segmentation in the second to last session of the day at eMetrics in London 2012.
He looks at different approaches across different types of data so we can learn about simple models and advanced data mining techniques to help you become a ‘segmentation believer’.
The starting quote on the slides sum up what this session is about nicely:
“One size doesn’t fit all….. Meaning…. Segmentation is a must in the online world”
When we think about ‘big’ data we lose sight of what that actually means. We are looking at data for all customers on a website but we don’t use the data in the right way. All the customers are grouped together, not segmented and then data analysed is ‘averages’.
The problem with averages:
- Average time on site
- Avrage number of pages viewed
- Average conversion ratio
- Average satisfaction
We look as averages and pull everyone into the narrow segment but we don’t actually know what individuals are doing. Grouping people together isn’t going to give you the interesting and valuable data that you need. Averages rarely show the true picture, we need to segment!
What do we mean by segmentation?
Within our overall customer database, which ones have something in common. We need to find the groups of people, understand them and make some commercial value from the different groups. The three types of segmentation:
- Demographic – You can segment visitors by demographic but when we are looking at websites, this information is not great as we can’t make great use of it
- Attitudinal – You won’t know if customers are happy or unhappy until they complete a survey online but that isn’t going to be helpful initially
- Behavioral – In the online world, segmenting visitors by behaviour is key. We can optimise a website experience a lot faster with this type of segmentation
Let’s look at an example where we take new visitors vs. returning customers.
First time = 30% of visits
Repeat = 70% of visits
First time customers convert at 2.00%
Repeat customers convert at 5.00%
Overall conversion rate – 4.1%
If you run a campaign which ends up bringing in more first time customers, your overall conversion rate will drop. If you had segmented your data first and understood how your new vs. returning customers convert first, you would look to create a campaign to increase sales from the returning customers as they convert at a higher rate.
The Data Mining Process
This is a simple analytical process which you will continuously need to refine. Once you get to one stage, you will almost certainly find that you need to go back a step and refine some more before you finally get the data into a format that you can use.
Business Understanding > Data Understanding > Data Preparation > Analysis and Modelling > Evaluation > Deployment
TIP – When looking at mining data and grouping visitors into clusters, it is often wise to remove visitors who have only visited your site once. You want to focus on the customers that engage with you more frequently.
- Segmentation is critical in extracting insight from any digital data
- Means are generally meaningless and useless
- Never underestimate the data cleaning and integration aspects of any data mining
- To get meaningful and useful segments takes a lot of iteration, a lot of iteration, a lot of iteration….
If you want to find out more about the speaker, check out the below or follow them on Twitter: