User Segment Analysis

6 min readAug 14, 2021

In this post (Part II-III), I will do some analysis on user behaviour and in our ecommerce and how to segment the user based on their behaviour.

To start, let’s look at the purpose of our analysis :

What is the growth for new user & loyal user for our ecommerce?
What is the average spending for new user & loyal user for our ecommerce?
How many user in each recency and monetary segment based on user order behavior?

Background

In ecommerce, the more we understand the behavior the more efficient action we can take be increase the likelyhood to target our best customer. With this in mind, I’m trying to implement a simple RFM (recency, frequency and monetary) analysis based on our customer data with the end goal to understand more of our customer and able to identify which is our best customer based on simple segmentation from order behaviour.

Why is it important to segment our customer? Based on Salesforce, customer segmentation is a process to divide a customer base into groups of individuals that are similar in specific ways relevant to marketing, such as age, gender, interests and spending habits. One way to achieve this process is by segmenting based on their behaviour when using our ecommerce application (in this process, I will focus more on the recency and monetary).

Not only that, the user growth (either new user or returning user) is also as important for categorizing user. By understanding how is the growth on our new user and returning user, we can plan new marketing plan that can furthur impact those area or to improve things in the ecommerce that can boost user satisfaction.

Remember! The transaction isn’t over after they click “Purchase”. When we get new customer but you’ll leave a lot more on the table if you stop there. Repeat customers tends to spend more on average than first-timers, and because all your marketing costs were sunk in acquiring them in the first place, that extra spend is almost pure profit. That’s why by analyzing our current customer, we can improve our service to make more repeat customer.

Before we do our analysis, I want to remind the viewers that I will try to approach the problem based on user recency & monetary (and basket size) since most user have only been using our ecommerce for first time (or only one time).

New User vs Returning User

From plot above, we can see that :

The average for new user growth each month for our ecommerce is about 68.3% (early — mid 2016 & late 2018)
The average for returning user growth each month for our ecommerce is about 41.53% (early — mid 2016 & late 2018)

Based on the line plot, we can see that both new user and returning user for our ecommerce is steadily increasing (which is awesome!!). With the increasingly new user and returning user, we need to analyze their impact on business perspective. From below plot we can see that:

From user percentage growth between new user and returning user, we can see that 2017 might be the best year for new user in our ecommerce. Part of the reason might be because there are peak season in November 2017 (remember the legendary Black Friday)
From the average spending chart itself, we can see that along the time, returning user spend more than new user. This is one of the reason why we must make sure that returning user will keep using our service

User Distribution based on Lifetime Order

We have 1:28 ratio of returning user vs new user. It means that for every 29 new user, one of the would be our loyal customer, the rest will be churned. Not only that, most user (around 92k) only purchase 1 times from the ecommerce. Meanwhile, there are about (~ 3k) user that purchase more than 1 times.

Let’s analyze our loyal user using below plot. Majority of customers who repurchase order need between 22 and 168 days to do so. The upper quartile (Q3) suggests that there is a rough interval of 183 days which sits about 6 months.

By using the upper bound of that interval, we can have this as our cut-off point for recency, ultimately any purchase within 6 months (~183 days) will be considered as a recent purchase. Using this recency cutoff duration, we can segment our customer based on the recency of last purchase.

User Segment Distribution

Based on plot above, we can see that there are roughly the same amount of customers considered active and inactive, with the majority of customers being classified as active.

With this information, even though there are less than 10% user that come back (multiple purchase), we could see that there still about 30–40% user that still active based on our recency status segment (6 month) which probably new users.

Not only that, using the same recency segmentation for our ecommerce user, we can identify the order behaviour from our customer. To do this, I will simply cut all of the customer into 4 parts :

High Value & High Volume : Spend good amount of money + Order in high quantity. (HVHV)
High Value & Low Volume : Spend good amount of money + Order only in small quantity (HVLV)
Low Value & High Volume : Spend small amount of money + Order in high quantity (LVHV)
Low Value & Low Volume : Spend small amount of money + Order only in small quantity (LVLV)

Based on the purchase behaviour distribution, we can see that most of our customers are low value and low volume, then followed by customers with a high value but low volume. It might be interesting to see if we can furthur more classify our customer based on their other shopping behavior, but for now we can indicate that most of the time the customer for our ecommerce is indeed buy in small cart size, and most likely to prefer cheap product than expensive product.

Insight

As final takeaway, here are the things that we need to look out for :

The average for new user growth each month for our ecommerce is about 68.3%, while the average for returning user growth each month for our ecommerce is about 41.53% (early — mid 2016 & late 2018)
As a side note too, the new user and returning user of our ecommerce is steadily increasing, while the growth is steadily decreasing to normal ecommerce rate
From early — mid 2016 & late 2018, the ecommerce have about 1:28 ratio of returning user vs new user. It means that for every 28 new user, one of the would be our loyal customer
Most user (around 92k) only purchase 1 times from the ecommerce. Meanwhile, there are about (~ 3k) user that purchase more than 1 times.
Majority of customers who repurchase order leave between 22 and 168 days to do so. The upper quartile (Q3) suggests that there is a rough interval of 183 days which sits about 6 months, which can be used as cutoff interval for recent purchases.
Based on our simple order behaviour customer segmentation, we can see that most of our customers are low value and low volume, then followed by customers with a high value but low volume

Recommendation

Based on insight that we get from the analysis above, we can recommend few things:

Use customer retention strategies to prevent customer lost. We have about ~17k inactive user that were important customer. Offer personalized discount voucher, personal email marketing to engage with them.
Give reward for our loyal customer. We have about ~40k active user that might be potential loyalist customer. Offer marketing loyalty program, recommend other products, make a higher quit gate.
Create more personalized recommendation for each customer segment. With more accurate recommendation, we can create a good user experience and increase in sales