Data Science in eCommerce — Part 5

Finding patters

In the last chapter we discovered the following insights:

Cumulative conversion value (percentage) growing with each channel.

• Customers with more than 4/5 touchpoints are highly likely to do not generate any substantial revenue. Should there be a trigger to stop paid marketing activities for those users?
• Journeys with the short customer path — one, two three touchpoints are the most successful from the conversion perspective.

Let’s uncover some patterns! We will analyse customer journey made out of one, two or three touchpoints. Those customer journeys accounts for almost 80% of the revenue.

1. Customer journeys made out of one touchpoint

Customer journeys made out of one touchpoint vs. conversion value

How to interpret above graph? It can be presented in different form:

• Organic search is responsible for more than 50% of revenue within one touchpoint customer journeys,
• Organic search and direct are responsible for almost 90% of revenue within one touchpoint customer journeys.

Let’s summarise it. One touchpoint journey is responsible for 40% of the revenue. Within this length of the customer journey, organic search and direct (combined) are responsible for 90% of the revenue.

Numeric values:

2. Customer journeys made out of two touchpoint

Customer journeys made out of two touchpoint vs. conversion value

The chart above shows 30+ permutations (9 distinctive channels, two channels to conversions). You can calculate all permutations using this tool.

Let’s take a look at the cumulative conversion value of the journeys:

The first two permutations:

• Organic search > Direct
• Direct > Direct

makes up almost 90% of all conversion values for the two touch point journeys. Can we continue the discover going through one, two, three, four, touch point journeys. Chart below shows customer journeys made out of three touchpoints with corresponding conversion value (sum).

Customer journeys made our of three touchpoints

Customer journeys made out of four touchpoints

Is there a different approach? Let’s take a look at a quicker way to find the most valuable journeys without examining all lengths of the paths to conversion. I have setup a cut off point at the certain conversion value to restrict number of paths shown.

Let us take a look at he cumulative distribution: