"The client is gone - is it forever?" How to count customer churn in SaaS and what's wrong with basic metrics

    On the one hand, outflow management increases revenue. On the other hand, a client who is thinking about leaving the service cannot be loyal. Therefore, it is important to know the leaving customers, communicate with them and return to cooperation.

    But you can only effectively manage what you measure.

    I’ll tell you how to determine the level of customer churn: what metrics and in what section to look at. I will show you specific examples of how we do this at UniSender.

    My name is Andrey Churanov, I am the leader of the UniSender customer experience management team .

    UniSender is a simple email and SMS mailing service. We help marketers create beautiful letters, combine them in series, conduct tests and monitor the effectiveness of marketing.

    Every day we examine customer experience at all points of contact between the client and the company. The metrics we use will be close and understandable to many SaaS.

    Features of measuring customer churn in SaaS products

    When we talk about outflow, first of all, we need to answer two questions:

    1. How stable is our customer base? Ideally, the number of customers at the beginning of the month and at the end is constant.
    2. What proportion of customers left the company? Ideally zero.

    Two important metrics will help answer these questions:

    Customer Retention Rate (CRR). You can

    calculate by the formula: Churn rate (CR) . We believe this:

    Using these metrics, you can calculate the proportion of customers who no longer use the services of the company. Formulas take into account new customers - this way we can neutralize the impact of the rapid growth of the base in any period. In total, CCR and CR give 100%.

    How to consider these metrics as an example:


    • Customers at start of period = 100.
    • Customers at end of period = 105.
    • New customers = 10.

    Consider Customer Retention Rate:

    • CCR = (105-10) / 100 = 95%. In our example, we see that the number of customers is decreasing, since the indicator is less than 100%.


    • Customers at start of period = 100.
    • Number of departed customers = 5.

    We consider Churn Rate:

    • CR = 5/100 = 5%. This is an indicator of outflow. Whether it depends a lot or a little on a specific company and industry.

    What is wrong with Customer Retention Rate and Churn Rate

    The problem with calculating CCR and CR is fixing outflow. To stop using the SaaS product, in most cases, you just don’t need to replenish the balance. And if the client did it temporarily and after a few months will return? In this case, the calculation of CCR and CR may be inaccurate.

    This is a feature!

    We must choose a hypothesis of behavior that will show whether the client left forever or is it a temporary inactivity.

    I tell you what metrics we use to more accurately calculate the outflow of customers.

    Process Performance Metrics

    These metrics will help you understand how much our outflow management work pays off.

    Active customer share

    I note right away that each company has its own concept of “active customer”. For example, we at UniSender consider active those who, during the billing period, made any newsletter and replenished their account in the service. You need to understand what actions a client must take in order to become active. After that, you can calculate the proportion of active customers from all customers of the company.

    Why count. It’s easier to answer the question: “What proportion of customers use the service”? If this share begins to grow or fall, we must respond: find out which segment of customers is changing and decide what to do with it.

    We recommend that you consider the percentage of active customers for such periods:

    • Active customers for one month (Monthly Active Users, MAU).
    • Active customers for two consecutive months (2Monthly Active Users, 2MAU).
    • Clients active during the quarter (Quarter Active Users, QAU).

    Active customers bring money. Inactive - this is an outflow.

    How to calculate MAU, 2MAU and QAU on the example of 4 clients with different activities ("1" means that the client was active this month):

    • MAU = 3/4 = 75%.
    • 2MAU = 2/4 = 50%.
    • QAU = 1/4 = 25%.

    What to look for. It is logical that you need to focus on increasing customer activity. First of all, the causes of inactivity are studied, in each period they can be different. For example, in UniSender, customers stop using the product because:

    • consider that email marketing is not effective;
    • there is no newsletter for a new mailing list;
    • there is no time / specialist who will be engaged in mailing.

    You can read about the main reasons why customers refuse email marketing in our article .

    You also need to study the drivers that lead to activity. For example, in UniSender, an activity driver is a client’s actions in a service (login to the office, creating a letter template, creating a newsletter, sending a newsletter). These actions are often followed by replenishment of the account balance. That is, we must encourage the client to “touch” the service as often as possible, and not immediately sell it.

    Returned Customer Income

    We determined that the user leaves the service and contacted him in some channel. The client agreed to continue cooperation and paid.

    Income from a returned customer is the amount that the customer paid after our communication in any channel.

    Why count. This indicator is used to calculate the average check. The average check should be considered for each client separately and, in general, after all outflow communications.

    What to look for. A separate cut for the analytics of this indicator can be a communication channel (phone, email, personal meeting) and a client segment. You can segment by several indicators:

    • What income was earlier?
    • When was the last deposit?
    • What services did you use before?

    For example, you highlighted customers who left the service and made calls to them. Naturally, there will be customers whom you did not reach. Compare the amount of revenue after dialing customers and not dialing. If the choice of the outflow segment was made correctly, then the amount of income after dialing should significantly exceed the amount after dialing to customers.

    The number of payments after 30, 60, 90 days after communication

    The number of payments characterizes the stability of the process and the correct choice of outflow management strategy.

    Suppose you returned the customer to the service. After a long break, he replenished his balance. It's good! What next? Do you control the future? We believe that this is necessary.

    Why count. Normally, the share of recurring payments from returned customers with whom you communicated should significantly exceed the percentage of recurring payments from customers with whom there was no communication. This trend confirms that we have chosen the right strategy.

    To find out this indicator, we simply check if the client has paid again after the first deposit.

    An example from the life of UniSender.

    1. We make a sample for the quarter. We look at the number of payments after dialing and without it. The share of payments after dialing is higher. Our strategy is correct - we influence the client’s decision to return to the service.

    Payments after dialing and without it

    2. Next, see what percentage of customers made more than one payment.

    After dialing, customers often make more than one payment

    3. Fix financial indicators. Compare the average customer check after dialing and without it. The difference in terms of the number of payments is obvious. We also see that customers who have already managed to make 4 payments after returning have an average check higher. The share of such clients in our case is small (about 3%), so there can be any trends.

    The average customer check after dialing is higher

    . What to look for.If the client returned, then he believed that the problem due to which he left was resolved.

    In order for the indicator to grow, it is necessary to analyze the reasons for the outflow of customers and develop retention tools by the reasons for the outflow.

    For example, a client left the service for email newsletters due to poor email delivery. We contact the customer and give recommendations on how to improve delivery . After his return, we must check whether he used our advice. If we failed to raise the indicator, we are once again working on ways to increase deliverability. If this is not done, the client runs the risk of getting a negative experience again and abandoning the service.

    Process Quality Metrics

    Quality indicators show a balanced process. For example, we received $ 10,000 from returned customers - is this good or bad? And if one client brought this money? And if 10,000 customers? The indicators below will allow you to see if there are distortions in the process.

    Average check

    To calculate the average bill of returned customers, you need to divide the income from customers with whom you had communication, divided by the number of customers who made payment after communication.

    The calculation of the average value has several disadvantages. For example, the average is distorted for arrays with a large spread in values. For numbers 100, 200 and –300, the arithmetic mean will be 0, and this can not always be interpreted. Therefore, we additionally recommend measuring the standard deviation and the median.

    The standard deviation shows how many units each indicator deviates on average from the average value of the sample. For calculation, we use the formula in MS Excel:

    The median splits the sample into two equal parts. Half of the observations lie below the median, the other half is higher. We use the formula in MS Excel:

    What to look for. A good way to increase the average bill (tariff changes do not count) is to connect withholding tools. A company offer for customer returns should be helpful and understandable. And most importantly - contain a solution to the problem that the client encountered.

    For example, one of our offers contains personal expert advice. With this tool, we listen and hear the client, and he is us. We give additional value and after that the client is ready to pay his previous payment, and not the minimum.


    Conversion - the proportion of customers who returned to the service, from all customers with whom there was communication.

    What to look for. In short: do not take into processing those customers who are "not outflow." To do this, you need to constantly analyze customer behavior for outflow.

    For example, we at UniSender look at how long there have been replenishment of accounts and mailings in the service. For us, a combination of these two criteria allows us to determine the outgoing customer.

    I recommend comparing the conversion rate for customers with whom you spoke about returning with conversions when there was no communication. If the conversion in the second case is higher, then the actions are aimed at the wrong customers. Customers pay without our efforts. So these are not outgoing customers.

    Such actions save company resources and help focus on customers who have left the service.

    Customer feedback

    The reviews help to understand if the customer is satisfied with the service after returning. Naturally, we must focus on improving this assessment.

    What we consider:

    • CSAT (Customer Satisfaction Index) . It shows how customers are satisfied with the service at the point of contact with the company (phone, email or point of sale). For example, in UniSender, customers evaluate the quality of technical support immediately after communication. Now CSAT UniSender 95 out of 100%.
    • CES (Customer Efforts Score) . Shows how convenient it is for the client to use the service. Our question is: “How easy was it to register with the service?”
    • NPS (Net Promoter Score) . Customer Loyalty Index. Shows what proportion of brand promoters among customers. For example, we ask: “Based on your experience with the company, are you ready to recommend it to your friends and acquaintances?” NPS can be from -100 to 100. UniSender +41.

    We tell you how to collect customer feedback, and how it works in UniSender.

    What to look for . Low scores should lead to a more detailed study of dissatisfaction. Analytics and changes should relate to specific employees, employee groups, retention tools, or company products.

    The process of collecting feedback should be constant, and its analytics is cyclical. The main tools for improving the quality of the process and customer satisfaction are:

    • The constant collection of feedback.
    • Cyclicity in measurements.
    • Collection of proposals for improving and monitoring the effectiveness of implemented changes.

    In which sections to measure outflow

    Stationary intervals

    We at UniSender measure outflows at such intervals: day - week - quarter - year.

    I recommend measuring indicators as often as possible. New actions occur in the service every day. Therefore, we must control them.

    Customer segments

    It is normal that different segments will have different indicators. We set different goals for them in conversion and average bill.

    Segmentation can also optimize outflow management costs. For example, for more profitable customers, we use phone calls, and for less profitable clients, we set up automatic conversations.

    Communication channel

    We determine which channel holds customers better: a phone call, a message in Viber, or an email. For each channel, we get different outflow metrics.

    It's all?


    Collecting statistics is not enough to effectively evaluate the outflow. We at UniSender are building mathematical models to predict various outflow indicators. This will be our next article.

    In the meantime, you can read our other publications about outflow management:

    In the near future we will write how to create a mathematical model for predicting the outflow. Not to miss, subscribe to our blog on Habré .

    Also popular now: