In the past year, have you wondered why your churn analysis is inaccurate when, once upon a time (pre-2023/24), your churn analysis was pretty reliable?
If so, you aren’t alone.
Misguided and inaccurate churn analysis has contributed to businesses’ overall challenge in forecasting accurate revenue numbers.
Often, a churn analysis consists of a basic analysis based on historical churn data, which would then be extrapolated to get forecasted churn results.
In 2021 and 2022, this may have been sufficient, but with so many changes in the economy and SaaS industry, this approach is showing its flaws.
When you do a churn analysis that only looks at historical churn data, key inputs that can have a material impact on your results are missing.
Before we get into the often missed inputs, let’s level set and look at the basic churn analysis I’ve seen most revenue leaders conduct. It generally consists of the following:
- A list of all customers who have churned in the past year or whatever period you are looking to analyze.
- Churned customers are segmented by certain variables, such as industry, size, cohort, etc.
- Data analysis to identify historical trends.
- Data is extrapolated out to predict future churn.
While these steps are reasonable as part of a churn analysis, they need to be more comprehensive, but they’re unfortunately representative of many churn analysis practices I’ve witnessed over the years.
So now let’s look at what a thorough churn analysis looks like that will result in higher churn forecast accuracy (not to mention better churn mitigation strategies).
Let’s assume we are analyzing our churn for the past year. We’ve pulled data on all churned customers, and we’ve calculated our overall churn rate. Now, we need to understand some of the deeper layers beyond this overall churn result.
Step One – Segmentation and Cohort Analysis (Basic Churn Analysis)
When it comes to understanding churn and building predictive models, a critical step is to divide your customer base into segments and cohorts. This allows you to identify patterns of churn and retention within these groups by analyzing their specific characteristics.
As you investigate what is causing your churn, you’ve likely started to group your customers based on specific attributes. These segmentation factors typically include:
- Industry
- Company size
- Geographic location
- Utilized products and use cases
- Customer tenure
- Cohorts based on their tenure of being a customer
To start, look closely at past trends in your customer groups. This examination helps uncover signs of churn and also shows how each group has grown over time.
One thing to be careful about is anomalies in your data. For instance, a high-value customer leaving a certain segment can skew your analysis. Adjust the data to account for these anomalies, and don’t rely solely on averages. Consider median data, too. We’ll discuss this a bit more later in this post.
Step Two – Identify the Layers of Churn Drivers
To make sound decisions with our churn analysis, we need to truly understand the root cause of the churn. To do this, we’ll need to uncover a few layers of details.
For example, a common categorization I see for churn is ‘low product adoption.’
While this detail helps point us in the right direction, it is not good enough on its own to categorize it as a driver of churn. We need to understand why there was low product adoption. To bring this to life, I will give a real-life example from a client I worked with.
Over 40% of their churn was categorized as low product adoption. When we did some digging, we uncovered the root causes of why product adoption was low. They were:
- No Sales to CS Handoff leading to slow CSM engagement and the client ultimately becoming disengaged
- Poor fit client and could not address needed use cases
- Loss of the main point of contact and the CSM was single-threaded. The loss of the champion resulted in a drop off in usage as they were responsible for being the admin of the software.
As you can see, there are numerous layers to the churn reason, but identifying these root causes was extremely impactful. In this case, we immediately collaborated with the sales and CS leader to create a Sales to CS handoff, agreed with the leadership team on qualifying out bad fit prospects, and worked with the CS team on multi-threading.
The result was a churn rate improvement from 89% to 95% just by addressing this one churn reason.
It’s worth putting effort into identifying the layers of churn reasons as it enables organizations to make decisions on activities that will be high-value and impactful.
Step Three – Pull Insights from the Voice of the Customer
One of the best ways to understand the layers of churn drivers is to talk to your customers.
Sure, not all customers will respond and talk to you, but I bet you’d be surprised at how many will, especially if you give them the chance to vent.
If you have a large customer base and it’s not possible to talk to all churned customers, then there are ways that you can scale this effort.
- Conduct surveys
- Listen to call recordings where they may have discussed churn reasons
- Read online reviews
- Review emails
There are a number of ways to get this information, and it’s worth the effort.
There is nothing quite like getting the information straight from the horse’s mouth, which can be done through a multi-pronged approach of talking directly to customers and reviewing the data you have with what customers have already said. AI and Natural Language Process (NLP) have made the latter much more efficient and effective.
Step Four – Identify ‘Look Alike Customers’ Who Stayed
In other words, analyze your customer cohorts and also look at your customers who stayed. Often, churn analysis only involves looking at the customers who left, but we also need to look at the customers who stayed and why.
What can we learn from the customers who continue to renew, and how can we apply that to our at-risk customers?
Break your customers into appropriate segments and cohorts. Then, compare who stayed and why to who didn’t. You can identify factors like:
- How customers’ behaviors change over time, and what the common behaviors are of those who stayed.
- How are customers who have stayed using your product and engaging with your company compared to those who leave? For instance, do customers who join and participate in your community tend to use your product more and in more meaningful ways? If they do, this insight can guide your retention efforts. You can pinpoint product features or services that keep customers and prioritize promoting product adoption accordingly.
- Which retention strategies have had an impact and which didn’t. Cohort analysis of retained clients also allows you to track the effect of retention efforts over time. By comparing churn rates before and after implementing churn and retention strategies, you can assess their effectiveness and iterate as needed.
Step Five – Take Externalities into Consideration
Historical data is only going to give you part of the picture. The other part of the picture comes from externalities and macroeconomic factors. This is a component that often gets missed. Here are some factors that need to be considered:
- How the economy and industry that you operate in is performing. As we’ve seen this year, tech companies have been hard hit, especially those whose customer base is largely made up of other tech companies. Understanding what’s happening in the market can help you identify risk due to external factors.
- The economy and industry performance of your clients’ customers can impact your churn as well. For example, during the pandemic, companies that served the travel and hospitality industry were hard hit. Meanwhile, video conferencing companies saw a boom. While this is a more obvious example, there are often more subtle but powerful factors at play that are important to be aware of.
- Regulatory, compliance, and law changes can also help or hinder your churn. Identifying how these changes affect your clients can help identify both risks and opportunities.
- The competitive landscape is another factor to take into consideration. If competitors are releasing features in high demand or becoming more aggressive with pricing, this can signal a potential churn issue. Conducting a win/loss analysis is a great way to identify if this has already happened in both new business sales and with existing customers.
Although this isn’t an exhaustive list of external factors that could affect your churn, these are some common ones. Consider your industry, your operating environment, technology changes, and macroeconomic conditions, and you’ll cover a lot of data key points that can signal potential churn issues (or opportunities).
Step Six – Assess the Percentage of Churn within Segments
Your overall churn target is one thing, but understanding what percentage of churn is coming from your ICP and your bad-fit customers is another.
Calculate this by logo and revenue.
If a significant portion of your churn comes from your ICP, that is alarming. If it’s coming from your bad-fit customers, then it’s less of a fire.
If you don’t understand the context of your churn, then any decision-making and planning could be flawed.
Step Seven – Take Outliers into Account
Take a look at your data range and identify if any anomalies in your data are having a material impact on your results. For example, if you have a client that churned and their contract value was $100,000, but most of your contracts fall into the $20 to $50K range, then it’s worth noting and doing analysis around that.
For example, that churn value would significantly skew any cohort data that they would be included in and could lead to some incorrect conclusions. To avoid this, look at the data analysis with and without any anomalies before jumping to conclusions and making decisions.
There are also techniques to look at how representative our churn data is. For example, if we’re using a data sample to extrapolate customer behavior, you need to be sure that your sample is adequately representative of that customer base. It’s common to have significant outliers in data, which can greatly affect the results. For instance, if the average deal size seems high and is actually driven by a few huge deals, we might make decisions based on misleading information if we don’t consider these outliers. Simply averaging everything could be misleading, so accounting for those anomalies is important.
To test churn hypotheses or any hypotheses, you can split your data into two separate sets. Use one set to develop your hypothesis and the other to test or confirm it. This method ensures that your hypotheses accurately reflect the data you’re modeling and that you aren’t unfairly influenced by the data used to create them.
This may feel like a lot to analyze, but by following these steps, you’ll gain a fuller understanding of what caused past churn and what risks could affect future renewals. Instead of just relying on past data and projecting it forward, consider other crucial factors both inside and outside your organization.
This method also reduces the chance of using misleading data, which can create confusion between causation and correlation, potentially leading to ineffective decisions.
If your churn analysis is surface level, then you risk drawing conclusions about your past and future churn based on a coincidental correlation instead of getting to the root cause of your churn.
