The health score trap | TDSU Ep. 43

(0:00 - 0:27)

You want it to be simple because you want it to be usable. And if the health score is effective, a CSM should be able to just know in an instant to be able to diagnose what the problem is. What's up, Lifers, and welcome to The Daily Standup with Lifetime Value, where we're giving you fresh new customer success ideas and debates every single day.

(0:27 - 1:50)

I've got my man Rob here. Rob, do you want to say hi? What's going on? And that's it, disappointingly. It's just me and Rob.

I am your host. My name is Dillon Young. Rob, before we started recording, you told me you've got a topic.

And so you know what we do here, right? We ask one single question, what is on your mind when it comes to customer success? And so here's your opportunity to tell us what's on your mind. Tell you what's fresh on my mind. I've been asking myself, I've been talking with so many customer success teams who all say the same thing.

They say, we have a health score, but we don't trust it. And I did a poll with a group recently. Most people were saying their confidence in their health score is between 40 to maximum 70%, which is kind of crazy when you think about it.

It's like looking at your speedometer on your car and not knowing if it's accurate or your gas gauge, right? That's kind of alarming. And so hot take that I'll drop is that more companies, not all companies, that more companies should scrap the idea of trying to use a conglomerate health score to represent their customer base. And they should replace it instead with specific markers of account health that each predict different outcomes.

(1:51 - 2:25)

So one predicts retention, one predicts expansion, one predicts, and I say one, but it could be multiple, predicts customer advocacy, all different outcomes. That's what's on my mind. And so advocacy, retention, and what's feeding that as a single metric, or that can be a conglomerate? Because then really what you're doing is you're just creating an additional layer of conglomerate, right? I'm saying that I think we could live in a world where there are no conglomerate health scores.

(2:27 - 7:37)

And so there would still be health scores, they would just be way simpler? They would be markers of an account. For example, the analogy that Jamie, the CEO of Vitaly, used was he was like, imagine you go to the doctor and they say, but you're fine, you're 80% healthy, you're good to go. You're like, what 80% healthy? I feel like I need to know, like, what that 20% is.

That's the thing that's keeping me up at night, right? So maybe we have to look at the health components of an account, instead of the conglomerate score, and figure out that if we optimize for each of those different components, we'll do the thing of producing a healthy customer and getting whatever outcome we're looking for, whether it's retention or expansion or advocacy. But we shouldn't necessarily rely on one bundled score. And I believe the way Vitaly does it is you can create, this is not a, Vitaly is not a sponsor.

Once again, here's another miss on our part. But I believe in Vitaly's case, they allow you to create like sections, and then those can roll up. They can also be segmented by your different customer segments.

They can be rolled up in different ways, aggregated in different ways. But I'm curious, when you did the poll, did you also ask those folks like on average, how many components are in your health score? No, it did come up. Some people recommended having no more than about four, because they think, for example, my friend, Diane Gordon, she's got a ton of experience in this area.

She was saying that she thinks that it's all precious real estate, what you put in your health score. I would add to that that I thought it's also, you want it to be simple. You want it to be simple because you want it to be usable.

And if a health score is effective, a CSM should be able to just know in an instant to be able to diagnose what the problem is. So oftentimes, the more complex a health score gets, it can be overboard. I was just recently talking with friend of the pod, Byron Tarugno, about adoption and how we often get confused about health and activity, how we often we look at things like logins, how somebody might log in 15 times a day.

And folks would say, oh, that's great. They use the system a ton. But I think there's second level or even tertiary level data there that says, oh, that person is so busy that they log in and then they never do anything.

They log in, they get distracted and have to go do something else. So it's really only once or twice a day that they log in and actually navigate through the system and accomplish tasks. Well, I would say that is not a healthy customer versus somebody who's, oh, they logged in 15 times.

And I think we do that a lot with our data where we're like, oh, I have all this data. And all I need to know is, do I have enough data to say that they're healthy? When in fact, we're reaching this age where we've got to think a little bit more critically about whether the data we're looking at is actually indicative of what we're looking for. Back to your point.

And so I guess my question would be, or I'll just ask another question. Does sentiment have a place to live in your health score? Yeah. I mean, in some models, particularly in enterprise models, a lot of times you can't measure the sentiment if you're working like on a scaled model, right? Well, it's not that you can't, but it's not usually an in-depth analysis.

Usually it's nothing more than an NPS survey, which I know is in the doghouse in the success world. There's also, there's these new tools, and I'm not going to name these, not because I want them to be a sponsor, but because I don't remember the names of them, that will actually analyze the verbiage and the language used in your tickets to help you identify sentiment at that scaled segment. They'll say, oh, this person used 14 exclamation points and all caps.

They're probably not very happy with you, right? Though they could be saying, I love you, you fixed my problem, but typically it's more like, I can't get this thing to work and I've got a deal I'm trying to close. That one's so interesting to me because imagine a client says, it's fine, but that's so different than if they say, it's fine, or like, it's fine. Well, that's why it's AI.

You just trust that the overlords are, they're splitting those hairs for you. It's funny because it speaks to other ways that I've gotten this wrong in the past. For example, I was thinking through an experience where, well, actually I was thinking through the conversation you and I had a few days ago, which was like, a lot of people say, well, you want more time in app, right? But that might be true for Instagram, but it's not true for TurboTax where they want to minimize time in app.

Or I remember a time where I was trying to measure time in app and I realized people were spending more time in the app and I was like, I love this, this is awesome. It turns out our page loads were painfully slow. It sucked.

(7:37 - 13:58)

Yeah. And they were angry about their experience as a customer, but I wasn't capturing their sentiment. So there's room for sentiment analysis.

It depends pretty substantially on the model as to how you try to measure it. Yeah. Well, I think sentiment truly, my opinion there is that it only really works when it's juxtaposed with hard data.

If somebody says they're happy, but they've never logged into your platform, well, it's only a matter of time before their CFO says, Hey, do we really need this platform? And it turns out that they were your best friend, best man at your wedding maybe, but they don't need the platform. So they're going to have to say goodbye. So sentiment is really only one piece of the equation.

What's your solution? What do you think? I mean, I think there has been dialogue in the community about reducing the number of inputs into a health score, but would you say that that's not enough? Like we've got to go further. We've got to really simplify this thing. Yeah.

I think the solution is multilayered. I think first of all, if you have historical data, see what the data tells you. And I think most companies, they start their health score based on conjecture.

They assume this is what a healthy customer looks like, because maybe they don't have the data or they don't have complete data about churn. Like I work with some companies, they've never even seen a customer churn yet, because they're that early stage. And so their health score obviously is just based on total guesswork.

So first is checking the historical data against different silos that represent usage, the usage of metrics you're trying to track, also satisfaction metrics that you're trying to track. You might find too that there's more nuance to like executive engagement that you're trying to track, or whether the customer has left you a positive review or stuff like that. Although I recommend support tickets are another one, I recommend staying away from those nuanced ones to start.

So minimize the number of things you're looking at down to the bare essentials. Set up alerts across each of those rather than a conglomerate health score. And make sure that you're aligned with the outcome that you're trying to predict, because it's different if you're trying to predict churn, versus predict referrals, versus predict upsells, or that kind of thing.

Yeah, and don't don't measure effort metrics either. A lot of people are like, oh, our health score is based on number of QBR that we hold. Like that's a great effort metric.

But that's not a metric. That's not something that measures customer value outcomes. So I have one last thing to add on this, which is the best health score I think you can have is actually not a health score.

It's an ROI calculator. I worked with some companies on building an ROI calculator, which shows like you put $1 into this machine, you get $2 back. People pay for that all day.

But you know what I think the metric is? It's not the calculator itself. It's how often your customer looks at it. So if it lives in app, and they look at that calculator a lot, and the numbers are good, that's like a dopamine hit for them.

Oh yeah, this thing's working well. I would say that's the metric, is how often that's being looked at. Valuable too, because a lot of times customers don't, like they don't know what a healthy customer looks like.

They don't know what your other customers look like or what outcomes they're seeing. Well, I think benchmarking is incredibly valuable, both internally and externally. Your ability to benchmark your health score against other customers is key, but your ability to say, well, you guys don't engage with us in these ways and our most successful customers do, and things of that nature, trying to find a way to... I was also talking about this with Byron about benchmarking and how that... I think that's a key value prop to your customers, your ability to benchmark.

What I thought was interesting though, and this is kudos to whoever the customer is that's already building a health score and has never had anybody churn yet, that's key because now you will have the historical data. When your first customer churns, you can build theories as to why, and it's not pure conjecture. Like you said, you now have data to back up, like, oh yeah, they were 25% less in terms of logins or this tool's adoption rate versus our other guys.

Now look, you can't make that determination the first time somebody churns. You need to have some other churn occur before you can start to create a trend, but having that data at your disposal from the beginning is like, that's a goal, man. Absolutely.

At least something to look back on and say, okay, we could have seen this coming. Or we can start to build a theory of why you think it happened. Going back to Vitaly, they have that entire early warning system you were talking about.

I'm like, oh, this person dropped below this threshold or they went above this threshold. Here's a playbook for how to deal with that. You don't necessarily have to build a playbook for it, but they can warn you every time that sort of thing happens.

So to your point of, if you don't necessarily want to lump them all together yet, you can still have that ability. Again, they should really be a sponsor, man. Anyway, I think that's our time, buddy.

I love this topic. It's where I think we need to go, and we need to be spending more time crunching data and being super thoughtful in the way we structure these things so that we can unleash CSMs to just go and have the conversations based on those early warning systems or trying to validate the theories that we think that data is showing us. And this gets me excited.

Maybe we should do a round two. All right. Well, until then, Rob, we've got to say goodbye.

Take it easy. You've been listening to The Daily Standup by Lifetime Value. Please note that the views expressed in these conversations are attributed only to those individuals on this recording and do not necessarily reflect the views and opinions of their respective employers.

For all inquiries, please reach out via email to Dillon at LifetimeValueMedia.com. Find us on YouTube at Lifetime Value and find us on the socials at LifetimeValueMedia.com. Until next time.