Order Trend Analysis
👈 Part I-II| TOC | Part II-II 👉
In this post (Part II-I), I will do some analysis on order trend in our ecommerce.
To start, let’s look at the purpose of our analysis :
- When is the peak season of our ecommerce ?
- What is the growth rate for the ecommerce based on purchase rate?
- What time users are most likely make an order or using the ecommerce app?
Background
One of the best information that we can get to improve our ecommerce are by digging our own ecommerce past data. Using analytics, we can define which offer given to customer that were effective and not effective. On this analysis, we would want to know when the peak season of our ecommerce in order to understand what offer is the best for user (promotions, ads channel, timing, etc). Knowing when is the ecommerce user most likely to purchase (peak season) might be able to help ecommerce peak season preparation.
Not only that we also want to find sales trend as it is one of the most important aspect in ecommerce. As one of the main business process e-commerce can offer, knowing the growth of the ecommerce might help furthur business direction. The insight that we get from understanding the trend in the ecommerce can be used to predict furthur sales and what to prepare for furthur development for the ecommerce. Understanding when the user is most likely to purchase or using the application might help too, since it helps us to understand certain purchase pattern and can make better marketing related decision.
Before we do our analysis, I would like to assume that as more people in Indonesia start to use more of technology, with the average daily usage reaching 4 hours a day (twice than people in USA), I would like to say that most likely the ecommerce that we are going to analyze will have growing trend. Meanwhile, for the peak season I would say that most likely happen to special event such as Harbolnas or any similar events.
Distribution of order trend along the year
To answer that, let’s see below plot…
Data given only provide us information about order in ecommerce from Sept 2016 — Sept 2018 (2 years). That might be one of the reason order frequency in 2016 is so low.
Based on the line plot, there are some notes that might worth the attention :
- There are order / sales spike in 24 Nov 2017. Based on information in internet, that day is the famous Black Friday.
- At the end of the 2016, the sales is weak. Is it possible that most seller in the ecommerce have end year holiday ?
Because of the lack of data for early 2016 and end of 2018, we are unable to conclude more of significant findings here. But we know for sure that 24 Nov 2017 — end of Nov 2017 is the peak season for our ecommerce. Find what offer that the ecommerce gave to user to increase the sales!
Peak season of our ecommerce
Based the chart above we can conclude that ecommerce that we analyze has a growing trend along the time. We can see some seasonality with peaks at specific months, but in general we can see clear that customers are more likely to buy from our ecommerce than before.
Not only that, we can see that Nov 2017 is the peak season for our ecommerce. Most likely it happens because of the previous finding of Black Friday event that the ecommerce held. As a side note, we can see sharp decrease of order frequency from August 2018 — Sept 2018. It looks like there is some cutoff from the data (noise).
In order to see more clearly for the ecommerce growth, I use the data between Jan 2017 — Aug 2017 and compare it to Jan 2017 — Aug 2018. Based on the data, we could see that the ecommerce really improve within 1 year span.
There are increase about 142% order frequency from Jan — Aug 2017 vs Jan — Aug 2018. From the order percentage comparison itself, it we could see that as time goes on, the difference started to stabilize around 35% — 40% vs 60% — 65% for 2017 vs 2018.
Peak season per day in month
From day in month seasonality, we can also see that it distributed equally. But we can see that most user prefer not to order at the end of the month because the bar plot slightly lower in end of the week.
Weekday vs Weekends
Based from the day in a week distribution, we can see that our customer tend to make order in the weekdays than in the weekend.
Usage Tendencies by hour in a day
What about time seasonality? Is there any specific hour over the day that indicate more usage on our ecommerce ?
From the distribution of order over hour, we can see that order frequency steadily rises as the day progress and reaches the peak after noon and continues until about 22.00, then lower with dawn having the lowest order frequency than other daytime period.
Not only that, from the order time heatmap, we can see a quite interesting facts about user order time preference weekdays & weekend. On weekdays, user most likely to order from morning — night, while on weekends user tends to order from nightime.
Insight
As final takeaway, here are the things that we need to pay attention :
- Data provide us with order information from Sept 2016 — Sept 2018. Few reason on why 2016 data is so low are either ecommerce only started or many data loss due to not having a good system.
- There are order / sales spike in 24 Nov 2017. Based on information in internet, that day is the famous Black Friday.
- At the end of the 2016, the sales is weak. It seems possible that most seller in the ecommerce have end year holiday in 2016, but when we see the end of 2017, the same pattern not occuring, might indicate that the data in the end of 2016 have noise
- Our ecommerce has a growing trend. While there are some We can peaks at specific months, but in general we can see clear that customers are more likely to buy from our ecommerce than before.
- Based on comparison between Jan 2017 — Aug 2017 and Jan 2017 — Aug 2018, we can see that our ecommerce have 142% growth of order frequency
- By seeing user tendencies in purchasing, we know that user most likely to buy in weekdays than in the weekend, and not likely to purchase at the end of the week
- From hour over the day tendencies, we can see that order frequency steadily rises as the day progress and reaches the peak after noon and continues until about 22.00, and while during weekdays , the order frequency increases steadly after 9 AM, meanwhile the order frequency picks up only after noon during weekends.
Recommendation
Based on insight that we get from the order analysis, we can recommend few things:
- A/B Testing peak season offer to find out which offer actually customer into to help us learn more about what resonates with your audience and drive online sales. (Can be about offer, customer service, discount, cashback, and many more through that certain timestamp of peak season).
- Make much more offer similar to Black Friday Event, such as HarBolNas (Hari Belanja Nasional) to improve sales and lower new customer cost. We can see that by such event we can get more sales up to 3x from normal days
- Based on when user is using the ecommerce application, it is safe to say that most people like to use in weekdays (Mon — Fri) afternoon — evening (11.00–16.00). Use that time to make special offer in a specific duration only such as flash sale to increase sales & popularity.
