In this presentation, Jonathan Hsu, Lead Data Scientist at Social Capital, explains how data science can be used to dissect and understand growth metrics for consumer and SaaS businesses. He first distinguishes between macro‑level growth analytics (high‑level topline numbers presented to investors) and micro‑level growth analytics (product‑level experiments such as A/B testing and email lifecycle management). He then introduces a framework developed at Facebook for breaking down monthly active users (MAU) into three components: new users, resurrected users (previously inactive but returning), and churned users (those who leave). Using two illustrative examples that both show 5% month‑over‑month MAU growth, he shows how identical topline growth can mask very different underlying dynamics—one driven by low churn and high resurrection, the other by high churn and heavy acquisition. He argues that the appropriate strategic focus (acquisition vs churn reduction) depends on which component dominates. Hsu extends the same decomposition logic to monthly recurring revenue (MRR) for subscription businesses, adding expansion (existing customers paying more) and contraction (existing customers paying less) alongside new revenue and churned revenue. He presents two real‑world SaaS examples with contrasting churn/expansion profiles and introduces the “quick ratio” (ratio of net new revenue to churned revenue) as a concise health metric, noting a 4.5× quick ratio for a portfolio company versus 1.6× for a non‑portfolio company. Throughout, he emphasizes the importance of looking beyond headline growth numbers to the underlying drivers to inform product and go‑to‑market decisions.
Chapters
OK, so how does data science fit into growth? OK, a lot of the stuff that I'm going to talk about here is really going to be also about how we think about companies when they come to pitch to us. So when a company comes and talks to us about their growth, typically what will happen is they'll show us some sort of top-level— some slide that looks really great. And they'll be like, oh, look, it looks awesome. But of course it looks awesome. If it didn't look awesome, you wouldn't bother showing it to us. But the question is, what's going on underneath? What are the guts of this thing? There's some sort of macroscopic growth analytics that we end up doing in partnering with our companies to help figure that out, to figure out what are the drivers of that growth. There's this whole sort of other area, which I'm terming here as micro growth analytics, which is microscopically optimizing your product. This is A/B testing. This is piecing together your email communication strategy with your users in terms of managing them through their lifetime and getting them to come back. I'm not really going to talk about that today. A lot of people talk about this stuff, so I feel like it gets covered plenty. Oops. OK, so here's a concept— accounting for MAU growth. OK, so the key thing here— MAU here is monthly active users. So the graph on the bottom is an example of a company whose MAU is growing at 5% per month. Looks great, right? 5% per month. Now, that's something that some company could come in and show us, and they'd be like, oh, look, we're growing super fast. So the notion here is we actually developed a framework at Facebook to break down this growth so that we can sort of understand the components of it, because this sort of top-line figure can be hiding all sorts of stuff underneath. So in order to give you a sense of what that is, this formula here actually kind of explains this. So this basically says the MAU tomorrow, or next month or something, let's say next month, is the MAU today, right, plus whoever was new, right, you added people, plus some resurrected people. So those are people that you didn't quite add. They were here before in the past, but were not here this month, but came back. And then you lost some people, which is the churned, right? Cool? The bottom one says the MAU today Well, if they're here today, tomorrow they either come back, they're retained, or they don't come back and they're churned. Easy definitions. Cool. Now, the thing is that here are 2 examples. These 2 examples down here actually both produce the same result on the first page. Both of these here will produce 5% growth month over month, but they're really different. The one on the right, on the left I guess, when you were facing it, the low churn, low acquisition case, Is the case where there's a lot of resurrection, that big red chunk. So it's like bringing back a lot of users who have been gone in the past. Very low churn and very little new. So it's clear what to do there, right? We should just ramp up the new users. We should buy them, maybe. Maybe we should spend some money on advertising. Maybe we should be aggressively pushing the users to refer their friends into the system, right? Because we're not getting a lot of new users here, and we're also not losing a lot of users. So, you're saying that the So we're not stretching our users sort of to their point of leaving in terms of aggressively trying to push them to recruit more people. So that's the notion on the one on the low churn case. In the other case, this case looks very different. So in this case, you have a whole bunch of new users every month, a whole giant chunk of them. But the following month, they all fall out. That's what that big churn thing is at the bottom. And there's that little bit of resurrected, which is that little bit of red. On top. Now, that looks really different. So in this case, we probably shouldn't keep buying users or aggressively trying to get more users, because if we just get all these users, they're all just going to fall out. Not very good. In this case, the thing to work on is not getting more users. It's clearly fixing the churn problem and not worrying about the new. So the thing is that the underlying— so remember that both of these produce the same end result in terms of MAU growth. And the point here really is just that it can be masking different things. And it's to your service. So it does yourself credit to understand your own growth. Once you have a product that sort of gets some users, you need to understand, OK, if that number is growing, is that it? Or is there more to it? Are they sticking around? Are they not? What are the sort of dynamics driving it? I just mentioned a thing here. The area typically— so I see a lot of consumer companies. This area here above the line divided by below the line is typically around just larger than 1. It's very normal for a consumer company to be losing some users, getting some users, and for it to balance out and be slightly positive. So there's some slight growth. If you think about why that's the case, the reason why that's the case is because in some sense, when you make a consumer app, Your ability to compel the user to come back is not so high. You can't force them. You can't go over there and be like, hey, knock on his door, kick him into your app. So there's a natural sort of churn. The contrast to that is something like subscription revenue. So you can actually use the same concept with regards to subscription revenue. So MRR is monthly recurring revenue. So if you look at a SaaS business, They make money by having a subscription product that another business pays them for on a monthly basis. Or if you're like Spotify, you have recurring revenue. You can do the same thing with recurring revenue. The recurring revenue in the next period is the recurring revenue in this period plus new recurring revenue, which is the new one. Expansion is like people who were already paying for your thing but are now paying more. So, if you're like Slack, One month the customer is paying you for 10 seats. The next month they're paying you for 12. So it expanded by $2. So that's expansion. Churned, the customer falls out. Contraction is the opposite of expansion. So that's, you know, I was buying 10 months from you last— buying 10 seats from you last month, and this month I'm buying 8 seats from you. So I contracted by 2. So that's contraction. And you can see you can graph the same thing here. So on the left is an example of one where there's a whole lot of churn, of cancellation, and contraction of subscription revenue. On the right, there's very little. These are actually 2 real examples. I won't tell you which companies they are. The one on the right is actually in our portfolio. The one on the left is not. But we saw them. But this is, yet again, one of these examples. When you're looking at your subscription revenue, you want to have a sense of, OK, OK, so I spend all this sales team and bandwidth trying to acquire customers. Do they stick around? Do they stay above the line? Or do they start falling out below the line? Are you losing them? Cool? So our general partner, Mamoun, actually made a term for this. He called it the quick ratio. The one on the right is 4.5x. The one on the left is 1.6x. Thanks, Tim.
