Best practices for configuring a perfect customer health score (and why single health score fails)
TL;DR A single customer health score is not ideal for a B2B SaaS business. The best practices for configuring a health score include
- Monitoring the right metrics, i.e., don’t mix support signals with product adoption signals
- Measuring leading indicators with short lookup windows
- Having a simple risk strategy, i.e., poor score must be broad and should consider each churn signal separately
- Acting on poor leading metrics till health score is good
- Refreshing your configurations periodically (we recommend 3 months) on the basis of new learnings
Why calculate customer health score?
First, let’s answer the question, what does a typical customer journey in the subscription economy look like? A SaaS customer goes through acquisition, onboarding, retention, and upsell.
You look to deliver early value during the trial / demo to convert the prospect. Then, you look to deliver early success milestone during onboarding as the maximum churn happens in the first 90 days. Later you try to deliver consistent value to prevent churn. Ultimately, you have to identify unmet needs so that you can upsell / cross-sell.
At each stage, you need some kind of a health score to evaluate the status of the account and decide how you would engage with the account to take it ahead in the journey.
So, what are the signals that you have today?
- Billing: With a billing system, you get to know an account is in poor health via a cancellation request or a card removal or overdue invoice(s). But this is a lagging indicator as you come to know only when the action has already been taken by the customer.
- Support System: You usually rely on a low CSAT score or complaints to know that the account is not in good health. You are relying on the customer becoming frustrated with your product so this is not an ideal signal.
- Touchpoint: If a promoter becomes a detractor or there is a lack of engagement, it’s an indicator that the account needs more handholding. With this, you are still being reactive as there is no way you can predict if someone will become a detractor.
- Product usage: Tracking things like low usage, low login frequency, declining active users and even unmet outcomes can be a leading indicator of the account health. This is where customer success technology comes in.
- Early warning system
- Driving specific action
- Validate specific outcomes
Traditional single health score — why and where it fails
As recent as last year, CustomerSuccessBox also calculated a single health score but after early feedback from customers, we realized that a traditional customer health score does not work in the modern success environment.
So why does it not work?
Health never comes in one shape and form. Imagine your own health — can there ever truly be a single value to define your health? Our health is not a unit, not a metric that can be looked on at in aggregate. Our health comes in many shapes and forms — mental health, physical health, relationship health — just to name a few.
Applying it back to health of an account, a single customer health score is a deception.
With a single health score, you are expected to be a data scientist. You are expected to set up a complex formula and define the weights. And after all this, you get a single score which is supposed to define the tasks for you.
the customer success team that doesn’t understand the why behind the health score, does not know what to look at and which outcomes to drive for
The problem is there are false positives and false negatives all across. For example, you have 10 parameters for good health configuration and you are dealing in averages. If one parameter, say, last login was 10 days ago is negative and the remaining 9 are positive, health score will show 90 which is good health. In reality, if last login was 10 days ago, it should get flagged and alert the CSM.
This leads to customer success teams having low confidence in the health scores and leads to missed opportunities as accounts can keep showing up in good health and then churn.
Also the customer success team that doesn’t understand the why behind the health score, does not know what to look at and which outcomes to drive for. For example, if a team has to increase the health score from 50 to 70, they have no idea what to do as the health score is based on 10 random parameters.
One of our on-prem customers had come up with a better excel formula than technology could provide for health score calculation. And that is what provided the inspiration behind our framework for measuring a health score.
Framework for measuring customer health score
The framework is fairly simple and comprises three steps.
- List Signals: Signals here are any metrics or indicators that can be tracked and are periodically refreshed, e.g., usage frequency, last login, NPS, etc.
- Bucket: Bucket related signals into groups, where each group indicates a dimension of health. For example,
- Configure health: Define good and poor health for each group.
Perfect health for a B2B SaaS account
Using the framework above, you can calculate a perfect health for a B2B SaaS account.
List signals such as usage frequency, last login, due invoice, NPS, last touchpoint, key feature usage, manual input such as CSM risk score, and more.
Start segmenting the signals into groups based on logical grouping. For example, revenue trend, due invoices and days since last renewal can be in one group which is actually a measure of financial health. Similarly, unresolved tickets, frequency of new conversations and CSAT scores are in one logical group and indicate service health.
So, once you list all signals and bucket them into groups, you will come to realize that there are five dimensions of customer health.
The five dimensions of customer health score
- Financial signals: Revenue trend, due invoices amount, days since last renewal, etc.
- Product signals: frequency of usage, depth of usage, active user trend, license utilization, and more.
- Service signals: critical unresolved tickets, frequency of new conversations, CSAT score, etc.
- Relationship signals: touchpoint frequency, NPS, recency of touchpoint, etc.
- Human/Market input: CSM input, sales inputs, and market trends
Lagging indicators only alert you when a fire has to be put out and not when the fire is about to start
Best practices for configuring a customer health score
KISS — Keep it simple, stupid
You want to ensure that it is simple to configure and easy to understand, and prefer leading indicators over lagging indicators. Lagging indicators only alert you when a fire has to be put out and not when the fire is about to start.
- Track right metrics in the right dimension. Don’t mix service signals with product adoption signals.
- Prefer leading indicators with short lookup windows, i.e., signals which refresh over a short period of time.
- Follow a simple risk strategy
- Good score must be specific and should consider all signals combined, e.g. if you have five signals for good health, all five should be meeting the criteria for a good score.
- Poor score must be broad and should consider each churn signal separately. If you have five churn signals, even if one churn signal matches the criteria, you should have a poor score and flag the account.
Act on it
- You should know instantly which dimensions need work and should be aware of your strengths when engaging with a customer.
- Do a quick lookup of leading metrics to understand what exactly is going wrong.
- Take proactive actions to overturn the health score. Rinse and repeat until the score is ‘Good’.
Correct the course
- Refresh your configurations periodically on the basis of new learnings. We recommend revisiting health score calculations every 3 months.
- Adjust rule thresholds to identify and focus on retention first.
Bucket accounts for a more accurate health score
There are three primary strategies to segment accounts for a more accurate health calculation.
- Bucket by desired outcome. Example, for a CRM, all accounts which want to double the conversion rate could be in one bucket.
- Bucket accounts by purchased plan. This is fairly straightforward. Plan here is a proxy for a bundle of use cases. So you could have starter, enterprise, professional segments.
- Bucket accounts by lifecycle stage, i.e., trial, onboarding, acquisition, etc. Stage indicates what’s important and what’s not.
If you still have doubts regarding how to calculate a perfect customer health score, watch this webinar with Anadi Raj Tiwari, Product Lead at CustomerSuccessBox or feel free to check out what Account Health 360 is all about.
If you enjoyed this article, feel free to hit that clap👏 button a few times to help others find it 😄
— — — —