You know those “How am I driving?” stickers you see on the backs of 18-wheelers? The ones that give you a phone number to call if—for some ungodly reason—you want to criticize a truck driver you’ve never met? Outrageous. The people who call those phone numbers are the same people who get banned from their kids’ soccer games for screaming at teenage referees.
That’s not to say that feedback doesn’t have a time and a place. In fact, if you’re on the customer success team at a SaaS company, regularly taking your customers’ temperatures is an essential part of the job. By figuring out how satisfied and engaged a particular customer is, you empower your team to make educated decisions going forward. Does this customer make for a good upsell opportunity? Are they having trouble with certain features within your product? Are they at risk of churning if you don’t take action?
The key to answering these questions—and many more that are equally insightful—is customer health scoring. I sat down with two managers from our customer success team—Phil Kowalski and Bobby Kittredge—to learn more.
Here’s what you’ll know by the end of this post:
Let’s do it.
In a nutshell, customer health scoring is the process of scoring customers based on the likelihood of an outcome you consider important. In other words, customer health scoring enables you to predict how your relationship with a given customer may change in the future. In turn, you can use that information to improve the way you interact with them.
You want to offer a bit more than this.
If that sounds a little vague, that’s because different companies care about different outcomes. Whereas Company A may want to predict the likelihood of a customer churning—that is, ending the relationship—Company B may want to predict the likelihood of a customer upselling—that is, making an upgraded purchase. Of course, there’s no law that limits your company to only predicting one outcome. The point is that there’s no universal, one-size-fits-all approach to customer health scoring.
Succinctly summing up the importance of customer health scoring isn’t easy, but I’ll give it a shot: When their needs are met and their expectations are exceeded, your customers are the best marketers in the world. Think about a time when you experienced extraordinary customer service—from a SaaS company or otherwise. It sort of made you want to scream about your love for that company from the top of a mountain, right?
Well, your customers are no different. The better the service you provide—keeping in mind that what constitutes “good service” varies from customer to customer—the more likely you are to catalyze positive word-of-mouth. Customer referrals are incredibly powerful, especially when you’re selling an expensive product that requires a long-term commitment. By providing stellar customer service, not only do you increase retention of your current customers; you also increase acquisition of new customers.
Customer health scoring is the key to providing that stellar service. Once you’ve set up your system and collected the necessary data, you’ll have the insights you need to address each customer’s unique pain points. In return, you’ll enjoy less churn, greater retention, accelerated acquisition, and more success overall.
Sound good? Let’s move on, then—here are the five steps you need to take to build a customer health scoring system at your company.
At a high level, customer health scoring is all about making educated predictions. This leaves us with a rather important question: How do you make those predictions? That’s what we’ll address in this section of the post. At the risk of oversimplifying, the following are the five basic steps involved in customer health scoring. As an added bonus, we’ll include a best practice to accompany each step in the process.
Ultimately, a customer’s health score is based on the likelihood of a particular outcome—churning, renewing, upselling, etc. Ideally, you should be able to look at someone’s health score and immediately think “They’re liable to churn soon” or “This could make for a great upsell opportunity.”
Determining which outcome you want to predict is your first step towards establishing a customer health scoring system. The outcome you select, of course, depends on the unique circumstances at your company. If you’re having trouble keeping customers on board for more than a few months, you may want to predict the likelihood of churn. If you’ve got a fantastic retention rate and you’d like to drive some additional revenue from your existing customer base, consider predicting the likelihood of upsell.
As I said before, you don’t need to limit yourself to a single outcome; I’m only sticking with the single-outcome model for the sake of simplicity. You may find that starting out with one outcome and eventually transitioning to a more complex system works best for you.
Best practice: Get (and keep) your customer data in order. As you read the rest of this post, it’ll become increasingly clear that customer health scoring is an exercise in data science. Without accurate, up-to-date customer data, the health scores you assign won’t be reflective of reality. Therefore, before you move on to the following steps, you need to make sure your data is clean.
Fortunately, you don’t need to take that on by yourself. Gainsight, a company we’ve worked with ourselves here at WordStream, offers a number of products and services to help you get all your relevant customer data in one place. We highly recommend giving them a look!
Once you’ve determined which outcome you’ll use as the base of your customers’ health scores, you need a way to predict the likelihood of that outcome. This is where predictive signals come in. Essentially, a predictive signal is any behavior that’s related to the outcome you’ve selected.
For simplicity, let’s say you’ve decided to predict the likelihood of churn for each customer. In order to determine the best predictive signals, you need to ask yourself: Which behaviors may indicate that a customer is liable to churn? For now, let’s stick with just three answers:
I think the importance of picking the right predictive signals is pretty clear: The closer the correlations between your signals and your outcome, the more accurate your customers’ health scores will be. I like to think of each signal as an ingredient and the overall health scoring system as a cake. Without the right ingredients, a cake tastes bad. Without the right signals, a health scoring system tells you nothing.
Best practice: Challenge your assumptions. Naturally, selecting your predictive signals requires making some assumptions. For example, you assume that there’s a positive correlation between how often a customer uses your product and their overall health score. Although you can’t get around making assumptions, you can reduce the likelihood of them being costly. How? By staying on top of your data. If you notice that a significant amount of churn among customers who use your product often, you’ll need to adjust your signal selections.
Something that may not be immediately obvious to you upon selecting your predictive signals is the fact that they’re not equally indicative of the outcome you’re tracking. For example, whereas breadth of features a particular customer uses whenever they log into your product may only be somewhat predictive of churn, the degree of impact your product has on their business results is probably highly predictive of churn. If you treat signals of different importance as if they’re equally important, your customers’ health scores won’t be accurate.
That’s why you need to assign a weight to each signal. If your signals are ingredients, their weights are the measurements. In order for the cake to taste good—in order for your health scoring system to be meaningful—you need to be thoughtful about the measurements. So how do you do that?
Best practice: Talk to your customer success representatives. As the people who regularly interact with your customers, they’ll have a strong sense of which signals are more or less important. If some of them have clients who use few features and rave about the product while others report the exact opposite, you can safely conclude (for now, that is) that the breadth of a given customer’s product usage is only somewhat predictive of churn.
OK—you’ve determined which outcome you want to predict and how you’re going to make that prediction. Believe it or not, once you’ve completed the first three steps, you’re technically ready to assign health scores to your customers. However, these health scores are completely meaningless unless they’re benchmarked against some form of standard. For example, a health score of 70 means one thing when the highest possible score is 100 and an entirely different thing when the highest possible score is 1,000.
That’s why your next step is to develop a health score scale—to make your customers’ scores meaningful and useful. Because it’s already familiar to most people, using 0-100 as your scale is a pretty safe choice. This scale also makes it easy to distinguish between different buckets of customers. For example: Anyone scored in the 0-39 bucket is labeled unhealthy; anyone scored in the 40-79 bucket is labeled neutral; and anyone scored in the 80-100 bucket is labeled healthy. A common way to make this scale even more intuitive is to assign each bucket a color: The 0-39 bucket is red; the 40-79 bucket is yellow; and the 80-100 bucket is green.
As rudimentary as this approach is, it makes a tremendous impact on your health scoring system. All of a sudden, each customer’s score communicates a meaningful message about their relationship with your company.
Best practice: Segment your customers specifically. For the sake of simplicity, I used a traffic light model (red, yellow, green) as an example of a scale. Although that may work for you, the general rule is that more segmentation yields better results. By “segmentation” I mean the specificity with which you categorize your customers according to their health scores. Because you’ll be developing a strategy for each bucket of scores, it follows that more buckets enable you to develop more pointed strategies. In turn, more pointed strategies increase the likelihood of your customers feeling as if their needs are being met.
The whole point of customer health scoring is to enable your team to improve the company’s relationship with each customer in a logical, mutually beneficial way. Accordingly, you need to plan how you’re going to respond once each customer has been scored; otherwise, the data you’ve worked so hard to collect and organize is effectively useless. For example, you could incentivize unhealthy customers to stay on board by offering them discounted prices or free access to additional features that typically cost extra. At the other end of the spectrum, you could gauge your healthy customers’ interest in upgrading to a premium subscription.
Best practice: Refresh your system regularly. Because the initial iteration of your health scoring system will be largely based on assumptions, you’ll almost certainly need to make adjustments. Things won’t go entirely as expected—and the success of your system depends on your willingness to adapt accordingly. Additionally, with new customers come new pain points. As your company grows, so will the range of needs your customers bring to the table. If you stubbornly stick with your original health scoring system, a huge chunk of your customer base will get left behind. Alternatively, if you update your system on a regular basis—annually is a good place to start—you’ll ensure its continued success.
If you find yourself wondering whether customer health scoring is necessary, ask yourself this question: Is succeeding as a company over the long-term necessary? Exactly.
The longevity of your SaaS company is directly correlated to the success of your customers—which, in turn, is dependent on your capacity to deliver solutions to their problems. Developing a customer health scoring system immensely enhances that capacity. In short, it gives you access to the insights you need to serve your customers as well as you possibly can.
A huge thank you to Phil and Bobby for their help with this post.
Please read our Comment Policy before commenting.