Data Science in eCommerce — Part1

Making sense of digital touchpoints — Visual exploration

Did you ever wonder how to optimise customer journey to get more conversions? Data science and advanced analytics will help to achieve that. In this series of articles, I am going to show how to bring an understanding of touchpoints on customer journey and its impact on the conversions.

We can start with one of the most popular online analytical tools — Google analytics. It provides eCommerce data insight called multi channels funnels report( MCF). Report show unique paths to conversion. Each path is made out of individual touchpoints.

Multi channel funnels report for eCommerce

For each path we can see corresponding number of conversions and conversion value. Report is based on the default channel grouping set for the account (read more about default channel grouping).

MCF Report — Data extract

Above reports provides good starting point for the visual analysis. Let’s plot all all unique paths with corresponding conversion value:

Plotting all values on the single diagram shows distribution of conversion value against all unique paths. In this case, we have a high number of unique paths to conversion (1,800+) — axis x.

Only small subset of all paths produces meaningful results. In this case, we set a threshold for viable conversion value at $1,000. Setting threshold will help us with the analysis. Conversion value is a cumulative number made out of individual conversion over certain period of time.

Let’s zoom in into the area of the conversion.

Let’s zoom in — only small subset of paths produces ‘meaningful’ conversion values

There is couple of interesting take outs and questions we can asked:

• We have more than 1,800 unique paths to conversions but only a relatively small subset is producing conversions. Can we identify them?
• Is there a pattern in the paths that produces the highest conversions?
• Impact on conversions between short path and the long path?
• What is the impact paid media, social, email marketing activities? Do they increase conversions, shorten the path or maybe are ‘killing’ the conversion value?

To answer above questions, we will have to transform our data.

Learn how to transform data in Part 2 — Data Preparation