Product Monetization

How to use product data for forecasting new revenue and churn

product-led-ideas
Get access to all our product-led growth courses and private community of growth experts.
Learn More
About
Transcript
Feedback

In the age of product-led growth models, your product engagement data can be effectively used to forecast both new revenue as well as churn.

Derek Skaletsky:
Hi, Derek Skaletsky, I'm the founder of Sherlock. And in this session, we're going to talk about how to use your product engagement data to forecast revenue in a product-led world. So the first question is, obviously, why should we use product data to forecast revenue? And the reason is because in a product-led business product engagement drives the entire revenue life cycle. When we talk about new revenue in a product-led business, you are offering a free trial or some kind of freemium version, or even a pilot, free pilot, of your product to users. If they do not engage with the product and see value by using the product, they're never going to convert into paid users. Then, obviously, with your existing customer base, you're already paying users, you're offering your product on a monthly or an annual subscription, so if they don't engage with the product over that timeframe, they're just not going to continue to pay for the product.

Derek Skaletsky:
And then, finally, a SaaS revenue model, a current SaaS revenue model, especially a product-led model, is really a land and expand model. So if you can't get people to expand the size of their account, increase their spend with you, the model breaks. And so if people don't engage with the product, they're simply not going to be a candidate for expanding their revenue. They're more likely to churn. So product engagement, as you can see, drives and is essential for driving revenue across that entire life cycle for a product-led business. It's just that simple. So that whole model, the whole product-led model, has changed everything about how we deliver software, how we build it, how we market it, how we attract leads, everything about it. So why hasn't it changed the way we forecast our revenue? In some cases it has, but in the most cases, it really hasn't, and I think it should. So let's look at how we could do that.

Derek Skaletsky:
So when it comes to new revenue and forecasting new revenue, currently, there's two ways we're probably doing it. The first way is to just apply some standard conversion rate across your spreadsheets. So you know based on historical data, you convert about 5% of your trials to paid customers. So you're just going to take that 5% and you're just going to apply it to all your leads in your forecast, and that's how you're going to forecast new revenue. So that is one way to do it, and some people are definitely doing it that way. And then the second way people are doing it is they're using what I call a sales-led forecasting model. And this is probably more common and you probably are doing this yourself right now. So a sales-led forecasting model is pretty simple. It's got a simple formula. It basically says, "You just take a deal, you have a deal in the pipeline, you take the size of that deal, the potential size of that deal, and you multiply it by its likelihood to close this month." And that gives you the sales forecast for that particular deal.

Derek Skaletsky:
So let's take a look. So you have ACME in the pipeline that came in. We're going to say it's a $2,000 potential deal. They're 50% likely to close this month. So we're going to apply $1,000 from that potential deal to our sales forecast for the month. So you just do that across all your deals, all the deals in your pipeline, and you come up with an overall monthly forecast for sales for new revenue for that month. The deal size is a factor in this formula, obviously. We're not going to really focus on it. How you determine a potential deal size is funny and it's an art, and it's actually mostly like this. But we're not really going to focus on that piece. The factor we want to focus on is this likelihood to close factor. So in a sales-led forecasting, likelihood to close is based on sales stages, and by sales stages we mean this. You've seen this before. You're probably doing this. In your CRM, you've got various stages for a deal to be in. Demo requests, you held the demo, they're in consideration, whatever that means.

Derek Skaletsky:
They've requested pricing, you've sent them a proposal. So based on that sales stage, that's how you're calculating or translating a likelihood to close factor for your forecasting. The question is, what if you don't have sales stages? What if you're a product-led business without these typical sales stages, most of your customers are not going through this sales process? Well, in that case, we're proposing that likelihood to close for a product-led business should be based on the product usage of the trial or freemium account, or more specifically, activation rate. So what is activation rate? Great question. So measuring activation, this is going to be a really important thing for forecasting and product data for sales. So activation rate, activation itself, we probably all realize what this means roughly. It's the series of actions that an account needs to take to reach an aha moment with your product, to get to first value, to get set up, to get onboard, whatever you might call it. But an account needs to do certain things to become activated with your product.

Derek Skaletsky:
And activation rate is a measure of how far along that account is for reaching that activation point. And we think it needs to be a rate because we don't think it's effective to actually measure activation in a binary way. It's not just they're activated or they're not. You're going to miss a lot of opportunities if you have that kind of binary definition. What we mean is, you should measure activation like this. You've got a new account that signs up, and you've got a bunch of things they need to do to become activated. Connect to G-Suite, create a template, invite two teammates, and then account is activated. And some accounts will complete two of those, so those accounts are 20% activated. Some accounts will get four of those done, and that's an 80% activated. And some will complete them all, and that's 100% activated. So that's how you measure activation rate for your product, and you should have that for every account. Every new account, you should be measuring activation rate, and you should have them in a way that you can look at those activation rates across all your accounts.

Derek Skaletsky:
Those activation rates are going to be analogous to the sales stages that we saw before on the sales-led forecast. It's going to work very similar. So what you're going to do is you're going to define a likelihood to close rate based on an activation rate for your accounts. So an account that's 0% activated, you could forecast is 0% likely to close this month. An account that's 50% activated has a 70% likelihood to close this month. 100% activated is a 90% likelihood to close. So the product-led forecasting formula is exactly the same as the sales-led forecasting formula. The math works exactly the same. You're going to take a potential deal size for that deal, you're going to multiply it by its likelihood to close, and that's going to contribute to your sales forecast for that account. The product-led forecasting model, like I said, is exactly the same. The only thing that change is how we determine likelihood to close, and for a product-led forecasting model, you're going to use activation rate as the factor to determine an account's likelihood to close.

Derek Skaletsky:
After that, the math is exactly the same and you're going to get to your monthly new revenue forecast that way. Hope that makes sense. Hope it seems logical to everyone and pretty easy. Now, let's do churn. So there's two ways of forecasting churn that we're doing it right now, and just like with the sales model, most of us are just applying a standard churn rate in our forecasts. So raise your hand if your forecasting exercise looks like this. You just drag that 2% churn across your spreadsheet, and you've got your turn forecasts. So if you've got a bunch of historical data and it's very consistent and you don't have a lot of fluctuation and you're changing nothing in your product or nothing in your offering our messaging and you're attracting the same customers and everything's going to be consistent over the next 12 months, applying that kind of rate is probably fine. It's probably working fine for you. But generally, overall for most of us, it's really not that helpful to do that.

Derek Skaletsky:
So what we need to do on a product-led method is to forecast churn at the account level based on engagement with your product. So it's going to look really, really similar here. In this case, however, you know all your accounts and you know their current MRR. You don't have to forecast deal size in this case. You're going to apply a likelihood to churn factor to that MRR, and that's going to get you a monthly churn forecast. Same exact formula. But just like with the sales side, the likelihood to churn factor is something you're going to have to determine, and that should be determined based on product engagement usage, product engagement, or lack of product engagement, to be more specific. So how do you build that model? Well, here are the four steps for building that product-led churn forecasting model. The first thing you need to do is understand turn indicators. So let's dig into that for a minute.

Derek Skaletsky:
Why do accounts churn? So there are a bunch of reasons accounts churn. Let's look at the most common. Account never got the first value before actually buying or they didn't get onboarded well. So they never really got that first value, really adopted the product. They're a bad fit in general. They didn't really have a high value use case when they bought the product so it's going to be hard to get them to see a lot of value, so they're going to be likely to turn. They were never able to figure out how to integrate the product into their business process so it never really became essential to them so they churned out. A key user left the company. Obviously, that's essential. They found a competitor that better solved their problem so they left. You have your own product issues. You have bugs or performance issues with your product and they just couldn't take it anymore so they canceled. They cut their budget. They just cut their overall budget and can no longer afford your product, or they just plain went out of business.

Derek Skaletsky:
Except for budget cuts, which can come out of the blue at times and you can be blindsided by that sometimes, all these other reasons for churn will be expressed in some kind of product engagement indicator. So those engagement indicators, these are ways that people are engaging or not engaging with your product, can determine that they're likely to churn. So low activation, they didn't adopt the product well, they didn't get integrated with their team, that will be expressed as a low activation rate for that account. They just weren't seeing value, a key user left the company, those kinds of reasons will be expressed as low engagement with the product or a significant drop in engagement. This happens all the time when you see a key user leave. You can just see an account, their engagement fall over a month tremendously. That's an indicator that they're a churn threat. General inactivity, just inactivity period. So low engagement is one thing, a drop in engagement is another thing, but they just simply aren't using the product. They're just not logging in. That's another thing that's a big indicator, obviously, of churn.

Derek Skaletsky:
And the last one is they've triggered some kind of, what I call, a farewell event. So if your product has some farewell event, something that people do in the product that is indicative of them leaving the product, so they shut off all their integrations or they exported data, or they turned off a bunch of automations, those things are indicative of them getting ready to shut down your product and move on. So those are all engagement indicators of churn. So then the next step is, obviously, you've got to track these indicators. You do a good job tracking your product data, being able to measure engagement, measure activation, et cetera, and then you're going to defy your that likelihood factor by identifying those top at-risk cohorts based on product engagement data. So what does that look like? What you'll have is a table like this to define your likelihood to churn factor. We could break this down a little bit.

Derek Skaletsky:
First is tenure. We do think, for most of us, you need to take tenure into account when you're determining likelihood to churn. So I usually say you can have three main buckets of tenure, new, which are less than three months old, young accounts, which are three to six months old, or mature accounts that are older than six months old. Activation rate, major factor in this calculation. So if you have accounts that are less than 25% activated or less than 50% activated, those will be factors. The activation rate is really mainly applied to those new customers. You don't really need to worry about activation for more mature customers. Low engagement scores, so if you're tracking engagement and you can quantify it, if an account has low engagement scores, those are going to be factors in your definitions here. As well as inactivity, so an account that's inactive for 15 days or inactive for 30 days will be factors.

Derek Skaletsky:
The trend in engagement, so if you see an engagement trend drop more than 20%, that should be part of your formula here. And then what'll happen is you're going to start to lay out your definitions of cohorts and apply a likelihood to churn factor to them. So, for example, if you have a new account with less than 25% activated, less than 25 engagement score, they're going to likely to churn at 10% rate. If you've got a new account less than 50% activated, less than 50 points of engagement score, but they've dropped 20% in their engagement, they're also 10% likely to churn. If you've got a new account, it doesn't matter the activation rate, but they've been inactive for 15 days after signing up, well, they're likely to turn at 20%. And what you're going to do is you're going to do this definition. These are just examples, by the way, but you're going to do this exercise and define those cohorts and apply a likelihood to churn rate against them.

Derek Skaletsky:
And once you have that, now you just plug this into your forecast and it works exactly the same as, again, the sales forecasting formula. You take that account, current MRR, likelihood to churn, you've got your MRR churn forecast. What if you have annual contracts? This is a really great question. It can get complicated with the annual contracts, but absolutely you should be doing this exercise the same, except with a few caveats. Number one, you don't have to worry about tenure. So they sign an annual deal, you don't have to worry about new or young accounts. You're really just talking about mature accounts here in this forecast. You don't have to apply this model to all your accounts. If you have an MRR model, you're going to do this exercise and apply it to all your accounts. If you've got annual contracts, you're going to only apply this model or do this exercise to those expiring contracts.

Derek Skaletsky:
You're going to make adjustments to your at-risk cohort criteria, that definition of likelihood to churn, because, obviously, you're not going to need it in the new and young cohort, and you may have some different factors that you may want to take into account. Human interactions, a lot of times with annual contracts, you're going to have your team members interacting with the customer. So those are obviously better predictors. If you've got a customer that tells one of your CS reps that they're not going to renew their annual contract, obviously, that's a better predictor than any other engagement data, so you need to plug that in. And so I would apply this methodology to contracts that are expiring three months out for the next three months. And it'll look something like this. So you've got your accounts lined up. Only your accounts, like I said, who their contracts are expiring within the next three months. You'll do your calculations for likelihood to churn, apply that there, and then you'll have columns for the accounts that are expiring in each of the next three months.

Derek Skaletsky:
Then you're going to apply those likely to churns across, and then that'll filter down into a forecast for each month over the next three months. Like I've said, if you've got a verbal that an account isn't going to renew, you would just plug that in as 100% likely to churn in your forecasts so you can capture that. Now, what about expansion revenue? Another great question. There are some challenges with forecasting expansion revenue in general, and some challenges with forecasting it based on engagement data. The first challenge is there are a wide range of expansion levers for different products. Volume is an expansion lever. So if people are increasing their volume and therefore are going to get charged more or need to go to another tier with your product so they've sent more emails or they've used more automations or whatever it might be, that's a lever for expansion.

Derek Skaletsky:
Seat licenses, they just add more people to the account, is a lever for expansion revenue. You might have a feature-based lever, so they want to use a part of your product that's only a premium part of your product. So that could be an expansion lever. Or a lot of us have a mix of some of these or all of these to determine expansion. So that just makes things a little bit more complicated and makes things a little bit more customed to your product. You have to really figure this out based on your expansion criteria for your product. A lot of times it's difficult to estimate an upgrade amount, how much the expansion is going to be. And for a lot of us, there's still human interaction involved with those expansions, so just basing that forecast off of product engagement data can be a little bit challenging. However, it can be done. So if your criteria is volume based, your expansion criteria is volume based, you can look at a signal which says, "Show me all the accounts that are nearing their volume limit."

Derek Skaletsky:
So if you say they can send 9,000 emails a month, and you've got all the accounts that have sent 900 emails this month so far, you can apply an expansion likelihood factor of 10% to those accounts and start to forecast off that. If seat licenses is an expansion lever for you, you can look at active user rate above 75%, "Show me all the accounts that have active user rate above 75% and they're highly engaged." Well, they're 10% likely to expand their seat usage this month. So you can use that for forecast. If you've got feature-based, you can do a couple things. You can have a hand-raising option, so you can have an explicit hand-raising option that people go to the part of your product that's premium only and they request a demo. So, that's an explicit hand-raising. Implicit might be they go to that section and they don't request a demo per se, but they watch a video that you've put on that page so they could see what that premium feature is all about.

Derek Skaletsky:
So if you're watching that, you can apply an expansion likelihood to those accounts. Also, if they're highly engaged lower tier accounts, you've got a bunch of lower tier accounts that have an engagement score of 90 or above or 75% or above, you can say, "5% of those are going to expand their revenue for the month." So there are ways to create this expansion revenue forecast based on engagement data. So, in summary, obviously, we need to rethink forecasting for a product-led business. Applying our traditional sales-led forecasting models to a product-led business, number one, isn't very accurate, and number one, is just counter to the nature and the fundamentals of a product-led business. And that's because the entire revenue life cycle is dependent on product engagement. Every phase of that revenue is dependent on product engagement. So for your new revenue, use engagement and activation rate as the main factor in forecasting a likelihood to convert factor and then use that in your forecasting.

Derek Skaletsky:
For your existing revenue, use engagement criteria to define a likelihood to churn factor or rate, and use that in your forecasting. And for expansion revenue, it's definitely model dependent, so you've got to play with this with your own product. But give it a shot. I mean, I think you'll be surprised how much visibility it does give to your team and to what the expansion revenue opportunity is for this quarter or this month. This may sound like a lot, especially if you haven't even started to think about it this way, but I don't expect you to go out and do all this in the next month. But I would hope that you start to implement some of this. Maybe start with new revenue or maybe start with churn and start to ease your way into this kind of model. To find that engagement criteria, start to apply it and see how it works out. Implemented a v1 and iterate over time, and eventually, hopefully, you get to a really robust forecasting model based on engagement that's accurate and insightful for everybody.

Derek Skaletsky:
So thank you. That's the presentation for today, and like I said, this was not something I didn't intend everybody to just go out and do this. But what I hope is that people start to think differently about forecasting in their product-led business and really start to think about how product engagement data is an essential piece of being accurate and insightful in that forecasting. Good luck. Thanks, everybody.

Contact Us

  • This field is for validation purposes and should be left unchanged.
Gretchen Duhaime
Derek Skaletsky
CEO of Sherlock
Helping SaaS businesses master the revenue lifecycle @Sherlock. Former founder @Knowtifyio & @traackr