As an average store owner, I was always curious how much I will earn tomorrow, how can I make those sales better and when I should give or should not give discounts to my customers. Why my emails have low opening index even if I write useful emails with shopper related products or information?
That’s the time when I started asking. How can I predict the behaviour of my client? When should I send him/her an e-mail or make them a call? When should I remind users that “Here I Am“ and when I should push them to interaction? That’s when I found Latency Matrix. It is a magic ball to look in and predict when any specific customer should make the next interaction and what I should do in that situation.
Q – We should offer discounts to users as often as possible to get more sales
A – Wrong
The discount shouldn’t be used to bring shoppers back at all times. You should be the person who offers what they needs in every specific moment and you should keep a finger on pulse to make sure your customers will get back to you next time. Discount that is offered is not the right method to achieve that.
Let’s have a look how the predictable charm works with Latency Matrix
It looks scary at first, but we’ll check it step by step and you’ll know how to read and build same one on your own.
First thing we need to do is to group our buyers by 2 dimensions – the number of days it took for them between orders (days between interactions) and how many orders the exact customer has placed. When filtering them we will use those 2 dimensions to see normal customers behaviour at the store and to predict their future actions.
Let’s try to build our Latency matrix now.
When we’ve exported all customers along with their orders we need to find those who have placed only one order. We count how many days it took for customers to place their first and the only one order after registration. Skip shoppers with 2 or more orders, we’ll get back to them later. This value, days between registration and the first order, should be placed in the first cell next to 1-st order made by customers with only one order (order count).
Days it took for client to place 1 order after registration
In our example it took 2 days (on average) starting from registration date to place the first order.
In the same way we filter customers who have made their first order depending on the number of orders that they have made at your store. This way we get number or days it took to make first order if the customer already have 2 orders, same for 3, 4, 5 and 11+ orders per customer. We get the following picture:
Days it took for buyers to place the first order depending on the total number of orders that they have placed at your store
It is the point where we can can start our prediction. If the total number of customers that were used to calculate the number (in our case it is 694 customers) is high enough, we can say that it is the period when all registered users who are going to make the only one purchase should make their first interaction within 2 days (max), otherwise with a high degree of probability, they will not do it at all. Keeping that in mind we can setup some automated reminder that will tell purchasers something interesting to attract them to your store.
The maximum time it took for customers to place their first order is 12 and 17, though the number of clients with such behaviour is 49 and comparing to all customers you can ignore them. The minimum time it took for shoppers to place their first order is 0, it means that they have placed their order same day when the registration happened.
What is the correct time for the first reminder?
You expect clients to make their order in 2 days, if you have average time for customers with one order and it may take up to 4 days taking into account all shoppers. It is a good idea to set us your reminder to be sent on the 5-th day to let your them do their first step themselves.
Your numbers will surely change in time, as you will help your clients perform themselves as you expect.
Why you shouldn’t send the letter to All potential customers
If you use any third-party mail systems, this method may save you time and money. Your shoppers won’t treat you like a spam as you will not send mails to all of them too fast or too late when they lost interest in your store already. You will not lose potential profit if you offer a discount for such users in 2 days as most likely they are going to make this interaction themselves (on avg. up to 4 days).
So we get the list of registered users without any orders and send them a reminder in 5 days.
That is the scenario of the first reminder, but what about existing customers? One-time buyers are the most expensive for store owners, so this group is very important and you should predict when they will make their next interaction and when they should do it.
To understand that, we need to gather data about all second orders made by your customers. So we get the list of shoppers with 2 orders and calculate in how many days they made the second order from the first order and filtering them by the number of orders that they have at your store. So 120 is the number of days it took for customer to place the second order starting from the first one and the shopper has made only 2 orders, and so on.
2-nd – Number of days it took for customers to make their second order
Also you need to calculate average number of days for all customers between the first and the second order to see average statistics. Also keep in mind that those numbers are less accurate as the number of clients who have placed 2+ orders is lower than those with only one order. It is normal situation for an average ecommerce store, and this report should help you push one-time buyers and make further interactions.
In our example it took 120 days to place the second order for shoppers with 2 orders and on average it took 110 days to place the second order for all customers. So we can predict that the shoppers with only one order should place their second purchase within 110 – 120 days. Make a reminder for them in 121 days and push them to the second purchase.
It is time to offer discounts at this point as a discount is cheaper than new user. You can turn your one-time buyers into loyal customers now.
Same way you are doing with the third order, fourth order, etc. You can take as many orders counts as you need for prediction. Basically loyal customers who make more than 10 orders should be treated differently, so you can stop on those numbers.
As the result of calculations you get the following grade called Latency Matrix.
Basically what you need here is the every last order of each group and all average numbers. So the prediction of the upcoming sale for clients with 2 orders is the number of days it took to make 3-rd order in the group of customers with 3 orders (that is 96 in our case) and its average number that is 110. The predictiction of the next interaction for shoppers with 3 orders is 87 – 59 days, etc.
What is more, you can offer great deals to bring users back to your store by scheduling automated replies in 3, 7, 30 days of inactivity of all customer groups having our predictions as mail trigger “start point”.
In time this live organism will change and numbers will be more accurate and the % of sales will increase.
Also for your convenience it is better to get totals of customers in each group as you will rely on totals and they may refer to small number of buyers and may not be accurate.
Also in case you filter clients basing on time from last transaction, you can get the list of customers to send mail to and potentially lost shoppers with over 12 month of inactivity.
How often should I check this matrix?
It is a good idea to check it once a month for newly-created stores, once a 3 month for store with good reputation and quite stable sales and at least once a year for stores with huge history and good rankings.
What is not covered in this matrix?
You do not take into account time when the interaction was made, so all orders made 2 years ago and newly created orders are taken into account. Behaviour will change in time by speeding-up your new sales, though old numbers will bring your statistics down.
Predicting customers behaviour can make you a hand with your marketing strategies and mail marketing that you will start trusting to. Do not leave your clients on their own when they are about to leave your store forever.
This report is time-consuming, but can be done both in Excel or with Store Manager for Magento application, Enterprise Edition (Supports both Magento Community and EE).
Find more Magento reports by eMagicOne at – https://www.mag-manager.com/magento-report/
Hope you’ll going to grow fast and our article will be a small door opening your large future.
This post is provided by guest contributor Oksana Semenyuk, CMO at eMagicOne – company offering smart and convenient ecommerce solutions that make maintaining online business very easy and effortless.