Customer segmentation

The performance of a marketing campaign relies on accurate re-targeting enhanced by upstream customer segmentation.
Here is an example of customer segmentation we are able to realize.

1. INITIAL QUESTION

A company wants to launch a finely targeted advertising campaign, with a high ROI.
Company activity: e-commerce.
The company wants to know which kind of visitors are buying the most.

2. DATA COLLECTION

We help the company setting up web traffic analytics tools. These tools allow us to profile the website visitors. For this example, we only consider age, gender, estimated annual income, and whether the visitor bought something or not.
Here we analyze the data from 400 visitors, but there is no restriction to this number. We can analyse the data of ten of thousands visitors, or more.

3. DATA ANALYSIS

We use R statistical programming language, and not an oversimplified drag-and-drop software, which allow us to tailor the analysis at will. For every analysis, we thoroughly test the model.
Here we use a decision tree, that allow partitioning visitors depending on their characteristics.

  

4. RESULTS

Visitors older than 44 are those who buy the most. More than 80% of these visitors buy something when visiting the website. Then, visitors younger than 44 but with an annual income greater than 89’000 USD buy something 80% of the time. Visitors younger than 44 and earning less than 89’000 USD almost never buy anything. The model accuracy is 89%, meaning that in 9/10 cases, the model will properly identify if a visitor is a good advertising target. This is estimated by a confusion matrix comparing the prediction of the model on the test dataset and the real data of the test dataset.

Customer segmentation DataTailors example

5. CONCLUSION

For our client advertising campaign, we recommend targeting an audience older than 44 or younger than 44 but earning more than 89’000 USD annually.
We could tune the potential revenue from each targeted visitor by estimating the visitors CLV (Customer Lifetime Value). This would be helpful in case of tight budget to maximize the campaign ROI.