In this article, you’ll find out how to attribute your purchases by the last.session_start value of the classification by return-time. This time-based purchase attribution analysis will let you answer questions like “How many purchases come from returning customers?” and “How many purchases are made by customers in browsing mode (meaning they return within 0.1D or 1D)?” and “How many purchases occur during a customer first session?”
This analysis comprises of using an Expression and Event Segmentation in Exponea, which you have built in the Return-Time Session_Starts Classification use-case. These are applied to a Report, using a Running Aggregate to assign a new event attribute to purchase events based on the events preceding it.
What is time-based purchase attribution?
Purchase is an event that is tracked by Exponea to the customer when a transaction occurs. Because customer behavior events such as purchases and session_starts are stored in a Single Customer View, we can use Running Aggregates to “add” an attribute of a preceding event (session_starts always precede a purchase in e-commerce). We can use this function of Running Aggregates to build time-based purchase attribution.
This analysis uses an Expression and Event Segmentation already outlined and explained here.
Why you should use this analysis
Be able to distinguish your purchases. Not all transactions are equal. What is the proportion of purchases occurring within the 1st session_start of a customer? When returning customers start their sessions, how many of those returning sessions convert into a purchase?
Customer Personalisation. Based on this classification of purchases it would be possible to also classify customers into “spontaneous” and ready to buy buyers during their 1st session_start or in a returning session directly; while others are more “browsing” before their purchase which those purchases occurring within 0.1D or 1D of a previous session.
How do you implement the solution in Exponea?
The step-by-step guide on how to build the component parts for this analysis (an Expression for event session_start and an Event Segmentation for event session_start) that are prerequisites to building the below report. Click here for the instructions
1. Date Filter
Set for which period you want to report (last 14 or 30 days recommended)
Format: Year month day
Find running aggregate (under events) and create a new one.
When creating a running aggregate, choose last session start. As an attribute, select the event segmentation you created in the previous steps. It will appear in the list of events under Event Segmentations.
Create a new simple metric and select count > event > purchase
5.Customer and event filters
The final set up of the report
In this section we would like to show you anonymized examples from three different e-commerce verticals – fashion, household, electronics.
Digital-native fashion e-commerce
For this fashion brand we can find the following insights:
the number of “spontaneous” purchases during 1st session_start is healthy for its price point
the proportion of “browsing” purchases occurring within 0.1 or 1 day of the last session is healthy
purchases consistently dropping over the weekends, reaching the lowest value on Saturday, followed by growth on workdays; throughout the working week approx. a third of all purchases are made by those that return after more than a day
throughout the whole month the proportion of purchases done within 0.1 and 1 day is almost equal – 19% and 16.5% respectively, with decreasing trend on Sundays; while the ones did during 1st session_start is account for 25% of all purchases per month
A new player in household e-commerce
For this household brand we can observe these high-level insights:
the number of “spontaneous” purchases during 1st session_start is higher than in the other two companies; suggesting they have retention issues, and also that customers that arrive on the site had already decided to make a purchase (note: for this business, their website is their smallest distribution channel)
within 0.1 and 1 day, there is not a very high proportion of people making a purchase
the trend changes significantly starting on the 13th of September and continuing for two weeks. During that time, most purchases are made on weekends or first workdays; with Mondays having the highest number of 1st session start purchases and an almost doubled number of 30+ days returnees purchases; suggesting some marketing activity resonated with existing customers, nudging them to return to the site
the highest peak in purchases was on the 15th of September, possibly because of a public holiday that day (note it is possible this business also had paid TV adverts during this time)
Established electronics e-commerce brand
For this electronics brand we can observe the following insights:
Workday/weekend pattern: a substantial fall in purchases over the weekend; growth during the workdays with peaks on Mondays. During the working week approx. a third of the purchases are made by customers returning to the website after more than a day; with a peak on Mondays of visitors returning within 2 to 7 days
Only a small proportion of purchases each day are a result of “spontaneous purchases” at the customer’s 1st session_start. It is consistently lower when comparing to the other two companies; (note this business has a substantially higher average price per product)
The high proportion of actively “browsing” customers that purchase within 0.1 and 1 day of their last session. This suggests the business has achieved “browsing behavior” on its e-commerce website successfully (note this business is 10 years in the running)