Churn is dead. Long live Net Dollar Retention Rate with Dave Kellogg | SaaStr Software Community

Good morning, everybody. Thanks for being here bright and early. I'm Dave Kellogg, and today we're going to talk really about understanding SaaS businesses in general, how to think about them, how to value them. We're going to do that through the lens of SaaS metrics, and today with a particular focus on understanding the healthiness of the install base of a SaaS company. So that's going to quickly bring us to talking about churn, a topic about which I've blogged a lot, and to net dollar retention, which I think is actually a more effective way to get at what churn is trying to measure. So let's jump in. Quick introduction about me. So first, who is this guy? So I've been a CEO of SaaS startups for over 10 years at 2 different companies. I took HostDynamics from $8 million in ARR to $50 million and MarkLogic from $0 to $80 million. In addition to that, I have 10 years as a CMO at 2 different companies, one, Versant, where we went from $1 to $15 million, and Business Objects, where we had kind of an epic run from $30 million to over $1 billion in revenue, not valuation, revenue, the old-fashioned thing. And then 10 combined years on 5 different boards, Aster Data and Granular, and 3 other boards today, Elation, Nuxeo, and Prophecy. I'm also an advisor and angel investor, work with a lot of companies, and that's just a sample. And I'm also a blogger, which we'll talk more about in a minute. So, you know, so that's who I am. Why do I care about this topic? Well, it's a 2-line proof, which is metrics are key to fundraising, and fundraising is key to to success, right? Any further questions? And certainly now, you know, in today's Silicon Valley, and this has changed over the last 20 years, in today's Silicon Valley, now more than ever, right? The ability to raise money is what separates the winners from the runners-up and the losers. So in the formula, you know, this is not a race, you know, VC-backed SaaS is not a race where a strong kick at the finish, you know, like a quarter mile wins the race. In my opinion, it's a race where you get in first early by driving high growth with the right unit economics. And if you've done that, if you're a leader with good growth at the right unit economics, you can then raise lots of capital and you set up a virtuous cycle, right? The more capital you raise, the faster you grow, the faster you grow with the right unit economics, comma, the more capital you can raise. So that's the name of the game. That's why this matters so much. The other reasons I care about it are, one, it's a big impact on valuation and ergo dilution. Second, I'm a huge believer in really trying to understand the guts or the drivers of a SaaS business, and looking at SaaS metrics helps you do that. I think every CEO should have a driver-based model where you play with sales productivity, sales turnover, sales ramp time, marketing as a percent of sales, cost per sale. You put these various drivers in and you can see what happens to your business, see their impact. And to understand driver-based models, you want to understand SaaS metrics. And finally, the last point is more of a pet peeve. I loathe the general lack of rigor I see when some people discuss SaaS metrics, where they talk about metrics without defining them or without knowing what they're counting. And there's a lot of that that happens when we get down to churn. So part of this is, if we're going to look at these metrics, let's make sure we understand how you're supposed to calculate them, how people game them, and how they're supposed to work. So onward. So that's kind of the intro. I do blog a lot about this stuff. My blog's at kelblog.com, and some of the topics I cover are SaaS businesses in general, kind of all aspects of SaaS businesses, particular emphasis on SaaS metrics, strategy, sales and marketing, which is my background, and more. So if you have a chance, if you like the material today, swing on over to Kellogg and take a look. Among other things, I'll be posting the slides to this presentation there. So we have 3 topics today. The first is understanding a SaaS business, and we're going to talk about why it's hard to understand the SaaS business. We're going to talk about how to think about a SaaS business, and then we're going to talk about kind of basic unit economics. Second section, we're going to talk about the trouble with churn, because it turns out, much as I love churn, I increasingly am kind of skeptical when companies talk about churn churn rates, and we'll talk about the 3 problems I see with churn rates that kind of undermine it as a good metric. And by the way, churn rates are used in other calculations, so if churn gets polluted, so do the downstream metrics from churn, which is another problem we'll talk about. And by the way, even the concept of churn itself can be a little bit elusive, so, so we'll talk about that in that second section, the trouble with churn. And then finally, long live, you know, king is dead, long live the king. Long live net dollar retention. We'll talk about this. This is actually more of a public company SaaS metric than a private company SaaS metric. And this is kind of, in my mind, the rare case where the public company SaaS metric is better than the private company SaaS metric. So for those of us who have spent all our time in the private world, I'll introduce you to what net dollar retention is. Then we'll talk about it as a valuation driver. We'll talk about how to calculate it. We'll talk about how to game it. It too is gameable. And then we'll talk about kind of a new twist on it, which is how to manage it. And then as kind of a preview, we'll end by talking about the metric I see coming after it. There's a request in the chat for defining churn, which I'll get to. I'm going to try and keep one eye on the chat. It's hard, to be honest. I got 3 monitors up. I'll define churn in a second. In short, it's ARR that leaks out of your bucket, and that will mean more in just a minute than it did right now. So that's what we're going to talk about. But first, let's just go look at the profit and loss statement. We'll start here. And this, you know, they say the average American family has 2.5 children, right? Like there is no actual average American family. This is the average SaaS company. For each of these lines, I took the median off Meritech and they have a great kind of public comparables website on their website, microsite on their website. And I took the median company. So this is the median revenue, the median COGS, the median gross profit. And I just laid it out into a P&L. So this is many ways kind of the typical public SaaS company. By the way, it actually looks a fair bit like Alterix, not in all respects. Alterix has better gross margins and other things, but if you had to say who was closest to this, kind of just eyeballing it in the data set, it's not miles away from Alterix. But in any case, this is not Alterix. This is the typical medium SaaS company, and the answer of where customer success fits into the P&L is sales and marketing, by the way. Good question. So trying to keep one eye on that. But in any case, the reason it's hard to understand a SaaS business when you look at the P&L, there's kind of— there's actually two different reasons. There's what I call the usual suspects here, and then in the next slide we'll talk about the more conceptual reason. But the usual suspects are first, the public companies don't disclose ARR, annual recurring revenue, right? They don't disclose it. And ARR is your subscription base, how many people are signed up, if you will. Revenue is basically a math problem. If you know ARR, then revenue is simply a calculation, and it is a lagging indicator, not a leading indicator, which is why financial analysts don't really look at revenue as much. They look at billings, quote unquote, and billings is a public company SaaS metric. It's revenue plus change in deferred revenue, and it's a way to try and impute what bookings are, but they aren't released, i.e., new sales don't get released, so you have to kind of triangulate on it. ARR needs to be implied in the same way. Churn is typically not disclosed. So you really, as a public company investor, are getting fairly limited visibility into the core engine driving a SaaS business. Now, if you think about it, it's kind of a one-way transformation, which is given ARR, I can in a deterministic way calculate a P&L. Given a P&L, it's non-deterministic to get ARR, right? I can come up with an infinite number of solutions that will generate that P&L. Maybe not infinite, but I can come up with a lot of different solutions that will generate the same T&L. So it is tricky to interpret public company SaaS metrics. In fact, I once met, ironically, a public company SaaS CFO who didn't actually understand SaaS unit economics because public company investors can't see them. So crazy stuff. The other problem with trying to understand a SaaS business is it's actually two businesses. You have this overall business that has $465 million in revenue. And it's composed of two very different businesses. One, the recurring business, all those existing customers who hopefully renew every year, right? And every year give you revenue. We'll call that the recurring business. And next to that is another business called the customer acquisition business, which is out getting customers. And, and look, I'll walk down the P&L. Let's just assume it costs the same to serve either type of customer. So 23% cost of goods sold. So each business has a 77% gross margin. Now that, that's where the similarity kind of ends. If we look, you could argue that 100% of sales and marketing in a SaaS company is to get new customers. It's not entirely true. Somewhere typically around 10% might be the cost of customer success to keep those existing customers. But, but just to make the math simple, um, I'd put all of sales marketing into this business that's out getting new customers and none into the business that's recurring, right? Keep the math simple. I split R&D. I just said half of R&D is to keep existing customers happy and renewing, half is to put in, you know, shiny new features that attract new customers. I split G&A on a similar basis. Probably should have prorated it by revenue, didn't do it, but I just put G&A 50/50. And when you do this exercise, what pops out is you see that a SaaS business is kind of a weighted average of this incredibly profitable recurring business with 55% operating margins and this incredibly unprofitable customer acquisition business. Now, the good news about the business on the right is that it is feeding people to the business on the left, right? And in many ways, you know, in order to make things simple, I tend to be very black and white and kind of do dichotomies, right? Just like putting all of sales and marketing into the, to the new business business. The other way to think about this is private equity cares a lot about the middle column, the recurring business, right? And lenders who lend to private equity care a lot about that because that's the annuity portion of a SaaS business. And venture capitalists care a lot about the right side, right? Because they want to know how fast you can grow this thing and how efficiently you can add new customers. Now, in the dark recesses of virtually every board member's mind, at one point or another, somebody thought, hmm, look at what happens when you turn off sales and marketing. Gosh, I have this incredibly profitable recurring business, right? And obviously you can't do that. As mentioned, you need some sales and marketing to keep the recurring business going, and then you'll Well, in effect, you'll stop growing, but will you? Right. That's what net dollar retention is actually measuring. What happens to this business if you kind of just leave it alone? Right. And the fantasy, and it is a fantasy, is what if it grew all by itself? Right. That recurring business, if it could just grow, if I could get rid of that other business, how much is that recurring business worth? And the answer is, of course, as mentioned twice already, it doesn't grow all by itself. I'm oversimplifying to make the point clear. But by the way, it does grow at 40% of the cost of new logo customer acquisition. If you look at the customer acquisition cost ratio, for example, in the KeyBank survey that I'm citing down below, you'll see that the expansion CAC, or the customer acquisition cost of expansion ARR, is about 40%, i.e., less than half of the cost of new logo ARR. So this thing doesn't grow all by itself, but gosh, it certainly does grow a lot cheaper than that very expensive new customer acquisition business. Now let's go back to private company land because I started us in public company land with the kind of average public SaaS company. But look, in private company land, which is where most of the people I think on this call come from, we get to see the underlying SaaS engine. And now for the person who asked about the definition, here's what I call the SaaS leaky bucket, which is I think of a SaaS company as a leaky bucket full of ARR. Right? Every quarter sales dumps more water in, uh, every quarter customer success tries to minimize the leakage out. Right? So we start with a certain amount of ARR, $3.9 million in this case, in 2Q of '19, we added $525K of new ARR, of which $400K was new logo, $125K was expansion, right? So new ARR comes in two basic flavors, right? From new customers or from existing customers. And then despite their valiant efforts, 100 units of ARR leaked out of the bucket, therefore leaving us with $4.3 million in ending ARR, right? And you could just keep running that thing every quarter. Obviously, ending ARR in one quarter is starting ARR the next, and you could get some good metrics just off this, right? Like, how fast is your ending ARR growing year over year? In fact, if you only had two numbers to give you the value of a SaaS company, it would be ending ARR and growth rate. I could then tell you roughly what that company is worth, kind of a first-order valuation. Another interesting metric is net new ARR. So if you subtract the churn from all of the new ARR, how much did the water level in the bucket change? There's a lot of ways to look at expansion. One simple way is just expansion as a percent of new ARR. People don't talk about this enough. I think it's a real simple-to-understand metric. To me, when it's less than 20%, maybe you're not exploiting your expansion opportunity much. As much as you should be. When it's over 40%, maybe you're not getting as much new logo as you should be, right? So I personally like it running in the 20s and 30s. And then there's net expansion, which is kind of all in— if we— not all in— if we take new ARR— sorry, if we take expansion ARR minus churn ARR, what do you have left? So that's kind of your net expansion of the pool. And you can see this company is not doing particularly well on that metric. It went up $25K, i.e., expansion offset churn by $25K into Q19, but then it it immediately went negative and offset the increase in 3Q19. So this company is actually struggling with that metric. But anyway, once you're looking inside a SaaS company's kind of leaky bucket and you can see the ARR engine, as I might call it, you can then calculate these kind of core SaaS unit economics metrics. So yeah, new logo to me is— there's a lot of different terms for it. Some people call it net new, which I don't like. To me, when you get ARR, you either get it from a new logo, i.e., someone who's not an existing customer, or you get it from an existing customer. So new ARR from somebody you're not doing business with is what I'll call new logo ARR. Some people call it new biz ARR. Some people call it net new ARR, but I think it's misleading. And when you get ARR, like if you were an existing customer paying me 100 units and you expanded your contract to 120, that would be expansion ARR, sometimes called upsell. And there are fine distinctions between those terms, but grosso modo, that's a good set of definitions. Acquisitions. So probably my favorite SaaS metric is the customer acquisition cost ratio, which is how much are you spending to put a dollar of ARR in that bucket, right? So it's last period sales and marketing expense divided by this period's new ARR. For SaaS metrics nuts, that's actually the inverse of the magic number which Scale Ventures came up with. I prefer it in this form because it's more intuitive to me to say how much did a dollar of ARR cost me. So I like the CAC ratio, a very important SaaS metric, you know, typically runs around 1.3, you know, below 1.3 is getting to be quite good, below 1 is really good, best I've ever seen was 0.4, anything between 1.5 and 2 people start to scratch their eyebrows, above 2 people get nervous, above 3 people won't talk to you, roughly. So lifetime value is 1 over the churn rate, and this is actually customer lifetime, I made a mistake on the slides, it's a little confusing, but the lifetime value is basically the ARR per year times the lifetime lifetime, and the lifetime is 1 over the churn rate. So if you have a 10% churn rate, we would say your average customer lifetime is 10 years. Now, this is where it gets interesting because there's this notion of, well, wait a minute, you know, is a CAC of 2 too high, right? If I'm spending $2 to get $1 of ARR, is that really too much? It's on the high side compared to averages, but for my business, is it too high? And to answer that question, the best way to do so is say, wait a minute, you paid $2 for something, what's it worth? Right? If you have a very high churn rate, just say you have a 50% churn rate, it's a terrible deal, right? Because you paid $2 to get $1.5, right? Because they went 1 year and half of them churned, you only got $1.5 back. So that's not good. But say you have a 10-year LTV. Well, gosh, if I have a CAC ratio of 2 at a 10-year LTV, then my LTV to CAC is 5, and that's pretty good. I'm getting back 5 times what I paid, which helps pay for all those other expenses, right? Like R&D and G&A and everything else. So LTV to CAC for me for a long time was kind of the ultimate SaaS metric. Like, this is it. This is the one that kind of explains everything. It's kind of a balance scale between how much you're paying for CAC versus how much you're getting back in LTV, right? And conversely, by the way, if you have a really low CAC, great, but if you have a very high churn, well, great, you can acquire customers who churn after a year very cheaply, who cares, right? It's not a great business. So another metric, which is very popular, and I mentioned it here because VCs look at it a lot, is CAC payback period. And this is kind of my unique way of calculating it. It produces the same answer, but if you just take your CAC ratio and divide it by subscription gross margin, you get in years the CAC payback period. What does that mean? If your CAC is 2, i.e., you're spending $2 to acquire— well, let's say your CAC is 1.5, make the math easy. You're spending $1.5 to acquire a dollar of ARR, and your subscription gross margins are 75%, then that means you get paid back your CAC in 2 years. 1.5 divided by 0.75 is 2. So how many years of gross profit does it take to pay back the sales and marketing costs of acquisition? A lot of people love this metric. You know, VCs in general kind of like all-in-one metrics because they're trying to make an investment decision. I always feel like operators don't love these metrics as much because if it's too high, I don't know which side of it is broken, right? If my CAC payback period is too long, is it because my CAC's too high or my subscription margin is too low? And as an operator, I want to go fix it. So in some ways, operators look at SaaS metrics a little bit differently than investors. So you think we're good. Hey, we just did SaaS Unit Economics 101. That's the— we looked at the public company view of a SaaS company. We looked at the private company view. We examined really what are the core key metrics, the leaky bucket, CAC, LTV to CAC, churn. We're good, right? Well, no, not so fast. Increasingly over the last several years, I personally, and I've blogged a lot about churn, I've become increasingly skeptical about churn rates. And I I almost entitled this section kind of tongue-in-cheek, "Why Private Equity Firms Recalculate All Your Metrics," because people play games sometimes with their SaaS metrics. And the VC side, frankly, I don't see as much recalculation, but when you meet PE people, they will literally take all your financials and just recalculate everything. You're going to say your churn is only 10%. They're like, "Well, we'll figure out what your churn rate is. Just give us the data and we'll calculate all the metrics." And the reasons we're going to talk about in here are some of the reasons why. So churn in my mind has kind of 3 fatal flaws, or not fatal, but you know, in reality, look, the title of the speech is somewhat dramatized. I wouldn't say churn is quite dead. I would say churn is wounded, and the problems with churn are threefold. First, there's too many darn ways to calculate churn, and that makes it ambiguous. Which one are you talking about when you tell me a churn rate? And by the way, you know, it could easily double the number, and any number that if you don't know how it's defined is off by a factor of 2, it's kind of a big deal. Deal. So it's a little bit too gameable, it's a little bit too misunderstood. Second, churn is used in downstream metrics like LTV to CAC. It's implicitly in there. And unfortunately, if you pollute churn, you pollute everything downstream of churn. So in my mind, churn inflicts kind of collateral damage on LTV. There's the question of, in a SaaS company, what goes into COGS to compute gross margin on the last slide? The answer of COGS in a subscription— COGS of the SaaS company is the cost of running the service. So in the old days, if If you had a data center, you'd have all your data center costs and your tech ops costs in there. In modern days, it'll be the check you write to Amazon in all likelihood for— and the people you have and the check to Amazon for running your SaaS service. Let me just see, there's a couple other questions up here. There's a lot of questions about new logo. I think we got that one covered, right? A new logo is somebody you haven't done business with before. Do I recommend a reporting system? We'll come back to that. So in any case, back to the question. Back here to the, to the content off the Q&A. Um, churn inflicts collateral damage on downstream metrics. And finally, churn, if you get into the details, there is kind of a dark rabbit hole of offsets and timing that even makes churn ARR, that one number, forget rates, just churn ARR itself, a tad ambiguous. Um, and let's jump in. I'll explain what I mean by each of these. So problem number one is there's actually at least four ways to calculate churn rates. You can either say before expansion or after So if I had 100 units of ARR in my bucket and 10 units leaked out from people who didn't renew and 15 units of expansion happened, did I have 10 units of churn or did I have -5 units of churn, right? Am I looking at churn before expansion or after? Gross churn is before expansion, net churn is after. And sometimes people reverse those two terms. It's not that uncommon. So another confusion factor in churn, what does that mean? Is it net of or or net. But in any case, by definition, gross is what leaked out of the bucket, net is what leaked out of the bucket offset by expansion, right? So you have which one are we talking about? So that's the numerator. And then in the denominator, we have, well, what are we dividing this by? Churn as compared to what? Either the two most common flavors are one, the entire ARR base, right? My whole subscription base or my available to renew base. Now, how could those be different, you might ask? Well, the answer is multi-year contracts. Contracts, right? Just say you do only 2-year contracts. If you have an ARR base of 100 units, only 50 comes up for renewal every year, right? If you do 3-year contracts, only a third of it comes up every year. If you do a blended average, as I did at my last company, of 1, 2, 3, and the occasional 5-year contract, some blended average of those contract durations come up every year for renewal. So, so, so this is a huge swing factor on churn, and it's, it's hard to argue which is right or which is wrong. If you're kind of trying to value the annuity as a financial person, I guess you should get the benefit of those multi-year contracts. So let's divide by 100 instead of dividing by 50, because mathematically they can't get out of those contracts. And if you have a history of renewing them and people not breaking out, then in reality, that's kind of the true churn rate. But on the other hand, if you're trying to understand the health of your business, you know, if 10 units leaked out against 100, that's 10%, not bad. If 10 units leaked out against 50, because only half the base was up for renewal, 20%, I'm getting pretty nervous. I don't know about you. So the issue is you make the matrix, you have 4 different ways of calculating this metric. And to me, it's a huge problem. And people often try to dress up one as the other. They'll kind of deliberately confuse net and gross. They'll deliberately confuse ARR with ATR. And, you know, obviously the most favorable metric is net, is number 4, the one— and by the way, I've ranked them in order of if I could only see one, I'd want to see gross ATR. If I could only see 2, I'd want to see gross and net ATR. If I could see 3, I'd want to see the but the same order on ARR-based churn rates. So, you know, my kind of quip on this is, you know, other than not knowing what the numerator is and not knowing what the denominator is, it's a great ratio, right? So there's a problem to me with churn rates. You can argue this is not a lethal problem. PE firms will just go recalculate this for you. A rigorous VC will kind of grill you to make sure you're actually telling them what, you know, they understand which box they're in when you answer. But that's only the first problem. Let's talk about the second problem. The second problem, as we alluded to, and there's really kind of two problems in here, is that churn blows up stuff downstream. So to the extent that churn is now less meaningful than we'd like it to be, now LTV is less meaningful, right, or lifetime really. It's not— this is where I made that same mistake again. This is really the customer lifetime, which affects LTV, is now blown up as well, right? So if your churn rate is 10%, 1 over 10 is 10-year lifetime. Oops, just kidding, it was actually 20. Oh, now it's a 5-year lifetime. Time. So this metric is now polluted by the pollution inherent in churn. So 4 different churn rates means you effectively have 4 different LTVs. So that's the first problem. The second problem is that very low churn rates break a lot of these ratios. If you have negative churn, i.e., you know, net churn that's negative, i.e., you expand more than you shrink, then you have infinite LTV. Well, I don't like that. I mean, I don't like a key metric that breaks when you're in a desirable condition because you want to have negative churn. So I want to know how negative, and now the metric's broken because I can't measure it. So That's not very encouraging. The other problem is more of a math problem, is when churn rates are very low, a lot of it ends up heavily in the future. You get a very high— you get a very long lifetime and a very high lifetime value, but like half the lifetime value is after year 10, right, or after year 25. And look, a 4-year-old company talking about a 25-year customer lifetime does make me at least uncomfortable, right? Like, how credible is that? So you're saying your customer lifetime is 6 times longer than you've existed. So that's a little dubious. This. The other problem with these very low churn rates is you get these very long lifetimes. So either you have the heavily future-weighted problem or you have the arbitrary cap problem. So one time I was working with a banker to sell my last company and I saw them calculating LTV and I saw that formula in the spreadsheet and I'm like, what are you doing? And the answer is, well, because of the long lifetime problem, if the churn gets too low, the lifetimes get too high, so I'm going to cap it at 10 years. So we're going to calculate it and we'll take the minimum of 10 years in the value. And as soon as I see that kind of stuff around a SaaS metric, it just makes me super uncomfortable, right? Because now I have 4 different churn rates. I have the low churn problem and the super long LTV problem. And now I'm inserting an arbitrary cap kind of silently, and it's just making me very skeptical. And by the way, as mentioned, if we've blown up lifetime value, we've then blown up lifetime value to CAC. So the metric that I once blogged as being the ultimate SaaS metric is now kind of polluted as well, because all these downsized downstream problems coming from churn. Now, um, let me just take a quick— so, uh, problem number 3: churn ARR itself can be non-obvious. So this is really in the detail, so I'm not gonna spend too much time here, but there are timing issues on calculating churn. I'll just give you a couple examples. Say somebody at renewal time drops 30 units of product A and adds 40 units of product B. Do we count that as 30 units of churn and 40 of upsell sell, right, which is going to drive up my churn rate and my expansion rates potentially, not net expansion, but my upsell as a percent of new ARR will go up, right? Do I look at it that way or do I look at it as just 10 units of upsell? And I wouldn't agonize over this one too much because as it turns out, most people, most finance and ops people, this is not a hard question. They just go, that's 10 units of upsell. Now, if I'm a product manager, I sure as heck would like to know which products are losing customers and which are gaining customers. So the product management perspective on this might be different, but the finance perspective what I call account-level churn. Hey, we care about this customer, they went up by 10, this counts as 10 units of expansion, and effectively that 30 units of what some people call product churn is effectively invisible because it's been completely offset. So in any case, that, that's one problem, which is at renewal time, how do we offset shrinkage and expansion? Not that hard, but it makes churn ARR a little ambiguous. The other problem is what I call upsell along the way, which is somebody starts out at 100 units, 6 months later they expand to 140, and they renew at 130. And the question is, gosh, is this 30 units of expansion or is this 40 units of expansion and 10 units of churn? How do I count that? And look, the company itself, I'm going to argue, is schizophrenic because when you're making the fundraising deck, right, you want to call this 30 units of upsell or expansion. But when you're doing CSM comp plans, you want to say, darn it, Mr. and Mrs. CSM, you lost 10 units on that contract. You couldn't renew it at full value. So I'd argue the company itself even gets a little bit confused. And then finally, we'll talk about this more at the end, but built-in expansion. So just say you sign a deal where it's a 3-year contract, $100 million year 1, $120 million year 2, $140 million in year 3. If you're going to think about like LTV to CAC calculations, this is more of a CAC thing than a churn thing, but how— what's the CAC? How much ARR did we acquire for that sales and marketing cost? Did we acquire acquire, some might argue, 360 units. Well, that's not ARR. We got 360 units of revenue, but of ARR, we got 100 units, the initial ARR, and then we counted as 20 units of expansion in year 2 and year 3, or do we flatten it out and say it's 120 across 3 years? Or if you want to be aggressive, do you say, "No, we acquired 140 units of ARR. We just haven't realized it yet." So which is it? Do a terminal ARR, average ARR, or initial ARR? If I'm doing LTV to CAC, I need to know the answer to that. So look, as was once said about cash and profit, there's a semi-famous quote that, "Cash is a fact, profit is an opinion." I feel the same way about ARR and churn, that ARR is a fact, churn and churn rates are an opinion. So, and that's what makes me uncomfortable with churn, and that's why I was kind of thinking, how can we solve this problem? Because it's just so detailed. Few people really want to talk about the 4 ways to calculate it, the 4 different LTVs, the timing and offsets. I mean, all that stuff kind of just, you know, bores people, confuses people. So how do we try and answer the question that we're trying to answer, which is how healthy is your installed base, without just getting kind of sucked into the weeds? And the answer is to zoom out, not in, right? And I have one word for you: cohorts, right? So let's talk about net dollar retention. So net dollar retention— actually, I'm going to just take a quick look at the Q&A here. So how does uncontrollable churn versus controllable churn fit into this? I mean, first, that's a great question. Those terms are super undefined in my mind, right? And even then, I think, like, what's the example? The example at my last company would be that, oh, they got a new CFO and they churned, or worse yet, they got acquired and then churned. So we're going to call that uncontrollable, right? We couldn't control that, so we shouldn't really count that kind of churn. But the issue is, it's not that simple, right? Because let's just say the reality was they bought the solution, they were not super happy with it, got acquired, and then churn. Oh, that's quite different now, isn't it? Because if they were super happy with it, maybe they would have fought to the death to keep the system rather than just use the system the parent company told them to use. And maybe that would have worked, or maybe it wouldn't have. So to me, controllable versus uncontrollable churn, it is a concept people talk a lot about, but, but I get super uncomfortable because I think it could easily be a cop-out, right? You know what, like my last company, they got a new CFO and they didn't renew. We can blame the new CFO, but if the new CFO asked the head of finance how they like the solution and that person was only lukewarm, the new CFO would say, gosh, just use my system then. I've got a great one, right? So I'm skeptical on the notion of uncontrollable churn. Do I care more about logo churn or ARR churn? I tend to look at SaaS businesses through an ARR lens, and I do that because, partially because the last SaaS business I ran was bimodal. We had a corporate sales channel that had a $25K ARR ASP, and we had a field sales that had a $75K ARR ASP, And therefore, if you looked at things like customer counts, they weren't that meaningful because we actually had almost no average customers, right? There were no 50K ARR customers. So because in my mind, the more your customers all look the same, the more you can just look at counts, right? Just counting the number of people who churned as opposed to looking at the ARR that churned. The more your business is 2, 3 different businesses all in one, right? The more to me you need to net it all out and just look at ARR. RR. So I don't do a lot of count-based metrics, but a lot of people do. They'll define churn as we had 100 customers, 90 renewed, therefore churn is 10%. I greatly prefer to look at ARR churn. So thank you for asking that. Yeah, usage-based pricing in SaaS with seasonality. That's a tough question. If we have time, we'll come back to it at the end. Where do you put the cost of retention, support, customer success, etc.? So you're right, I left customer support out of COGS. They do belong in COGS. Customer success, in my opinion, does not belong in COGS. It belongs in sales, and most companies will put customer success in sales. I have seen on occasion people put customer success into COGS, and the argument for that is if it's really a set of services to kind of keep you and get you up and running, it's a cost of running the service. If it's really about renewing contracts, it should be in sales in my mind. Some might argue G&A, and if it's about trying to drive upsell and expansion, then it definitely belongs in sales. Which, well, I won't go to the next level on that. So let me just go back to the Q&A. Do I see anything else here? Oh darn, I was looking only at the top of the Q&A, I'm sorry. What about customer support? We got that one. Long-term contract, what goes into acquisition versus recurring? Yeah, so the question, let me answer that one. If you do a 5-year contract for 100 units of ARR and I had that split SaaS business, where does the revenue go? And the answer is in year 1, it would go in the right-hand column. And in years 2, 3, 4, 5, it would go into the center column, the recurring business. That's where the revenue goes. The more interesting question, Randy, is where's the CAC go? How much, and especially if that contract's expanding, if I take that sales marketing cost to acquire that ARR, am I dividing it by the TCV? Am I dividing it by the starting ARR, the average ARR, the terminal ARR? By the way, the sales comp plan, what do you want to incent your sales force to do, is all tied up in that. So let me just go down here, sorry, the Q&A, I've been at the top of it. So good, good, good. How do you factor mid-contract churn versus contract renewal? Yeah. Oh, wow. You have mid-contract churn. Well, at first, that's not a good sign in my line. I've seen very little of it. Frankly, I'm always surprised that you don't see more cancellations of multi-year contracts. Like you sign a 3-year contract and people, they kind of just assume it's going to renew every year. And I'm like, well, just because they signed the contract, if they're really unhappy with the system, you may not get that. You know, it's not a renewal. It's a payment, but you may not get that payment in year 2. And maybe, in fact, if you want to optimize getting that payment, you might be better off treating it like a renewal, even though it isn't. So now, mid-contract churn means somebody is pretty unhappy. And the question is, look, if— let me pick the— I think that's what you're hitting on. Say it was a 3-year contract and they decide to stop paying in years 2 and 3. First, you could sue them if you want to, right? A contract is a contract. I don't recommend suing customers as a great strategy, but technically speaking, a contract is a contract. And if you're seeing a lot of that, to me, there's a big problem because frankly, I've seen very little of it. And I've worked with some companies where there are kind of serious product issues. People are not that happy, but they're still paying. So, but the short answer is it's not churn. It's actually, well, what is it? Let me think about that. So somebody refuses to renew. It wasn't up for— it's a great question, right? Because it's not in the ATR, base, right? So it won't show up in ATR because you're going to have 100 units of churn that was, you know, it's in the numerator, but it's not in the denominator. Technically speaking, it's actually bad debt, right? But, but you should take it out of the ARR pool. So, so personally, I would probably put it in adjustments. I mean, just to be clear, this is now the, like, the advanced level, but in that leaky bucket, I have one other row called adjustments, and I would put that into an adjustment. It's— I might put it into churn. I need to think more about that. I'd probably just put that one in churn, but go Those two adjustments is like currency, like just say you have some of your ARR in euros and some of it's in dollars and the size of the pool changed due to the currency rate changing, that would definitely be an adjustment in the leaky bucket. I guess now I'm convinced yours would actually be churn and I would count it in the period in which it occurred. And if I had much of it, I'd be very concerned quite frankly, because to a certain extent I'd also stop doing multi-year deals because if you're giving someone a price lock for a multi-year deal and they're not doing, you gave something and got nothing. So anyway, let's come off the Q&A back to the, back to our regularly scheduled content here. So look, this problem, trying to figure out how healthy the installed base is, is a great problem for cohort analysis where we just step back rather than stepping in, because when you step in, it's quicksand. We step back and say, let's just grab a cohort, typically all customers. What were they worth one time period to go? Let's just say a year ago, and now what are they worth today? Now, that's net dollar retention, today's value of their ARR, right? And ARR, there's some questions of that, it's just annual recurring revenue. If you agree to pay me 100 units a year to use my service, your ARR value is 100. So what were they worth a year ago? What are they worth now? And the beauty of this metric is 3 things. One, it's easy to calculate, it's easy to understand, and it's hard to game. And that's my biggest problem with churn ultimately, is it's easy to game. I got 4 different ways, plus some other things I could do to fudge it. So now there are a few bad apples who survivor bias this. Unbelievably, public companies actually do this. And what they're effectively saying is, well, excluding the customers who aren't doing business with us anymore, the cohort almost definitionally expanded by n%, right? So instead of taking a set of customers a year ago and looking at what they're worth today, they do it backwards. They take a set of customers today and look at what they were worth a year a year ago, and that's survivor bias because we've excluded everybody who's not a customer today. So the only way to game that dollar retention rate is to survivor bias it. And sadly, some people do, but most people do not. So look, the idea of cohorts is not new. Virtually every S-1— this is Datadog's— will show by year the value of cohorts. So the purple is all the customers acquired in 2013 and how they've evolved over time and value, etc., etc. So look, the idea of looking at cohorts in SAS is not new. The idea of making this kind of chart is not new. And this kind of chart is, in my mind, not only is it not new, it's quite standard. But what is newer is netting this out to a single number. Because if you get a single number, you can benchmark it. It's hard to benchmark this chart. I can't readily compare it to other charts. I mean, I can look at it, but benchmarks I can do off data. So, and by the way, net dollar retention is actually a very popular metric, right? So there's Martin Casado at Andreessen in a tweet stream about what metrics he likes to see. It's either in a board deck or investor deck, I can't remember, but you've got net and gross dollar retention. And by the way, remember I said some people define them inverse of how I do? He's one of them, right? So you can see he's saying that gross includes churn and downsell and net does not. But, but in any case, so a real example of where where, you know, people use those terms backwards relative to each other. But in any case, both metrics are important in his mind and in mine as well. You have Alex Clayton, a great SaaS blogger, who has done blog posts about net dollar retention rate in IPOs. Recently, this guy Zach Kukoff put a tweet up showing the net retention of IPOs at— of SaaS tools selling to developers, and you can see Snowflake 158, Twilio 155— those are really epic numbers, by the way— Elastic 142, and then finally John Ma, who I think writes Public Comps, I think is the name of his newsletter. You can see here's what it takes to go public in 2020. The average one had $330 million in ARR, 49% revenue growth, and 120% net dollar retention. So this is a popular metric, and I think it is popular for the reasons we cited. It's easy to understand, it's easy to calculate, and it's hard to game. So what is a good net dollar retention rate? About 115%, right? I think below 100%, people start to get scared, right? Because it's saying that your ARR pool kind of decays by itself. It doesn't expand. It kind of, if left alone, it shrinks. And that's not good. Not for lenders or not for private equity people or not even for VC people. Nobody wants to see the ARR pool shrinking. So I'd say 100 is a practical minimum. I think between 100 and 105, it's kind of considered very meh. 110 is, I think, kind of the minimum good, and 116 is median, and I think 120 starts to be very good. And anything much above 125, 130 is getting into the epic category. I'm gonna keep going here. So net dollar retention, you know, this was kind of an unpleasant surprise, which is I did some linear regressions and I said, hey, how much does net dollar retention drive SaaS multiples? If it's so important, I bet it's a big driver of SaaS multiples. And the answer is, oops, it's not. The R-squared is only 0.06 between net dollar retention and enterprise value to next 12 months revenue. So it's not a valuation driver. I now think of it as a financing enabler, right? Okay, so a higher one isn't necessarily by itself going to drive a higher valuation, but I can tell you if you don't have a high enough one, you're not going to be able to raise money. So it's still a super important metric. It just doesn't happen to be a good driver of valuation like ARR growth is. So the R-squared between ARR growth and enterprise value multiple is 0.5, and then rule of 40, which is the balance of revenue growth and operating profit. In effect, it's revenue growth minus free cash flow, or plus free cash flow, is the Rule of 40. And ironically, you can see here that the Rule of 40 at one point in time better explained multiples than growth, than just revenue growth. And now in these kind of frothy times, the better predictor of your multiple is just simply how fast you're growing, not how well you're balancing growth and profit. So that's kind of an interesting comment on the times, because for a long time, the Rule of 40 was a a better predictor of your multiple. How well you balance growth and profit matter more to your valuation simply than just how much you're growing. But right now, it's kind of just— what is it— just full speed ahead, grow as fast as you can. And that's a better predictor of valuation. Now, to move forward, let's start to wrap this up. Three pithy quotes to ponder. They're all basically saying, you know, what gets measured gets managed and various correlations. And when I was younger, I viewed that less skeptically. I was like, oh, if you want to manage something, you should measure it. And now I kind of have a negative connotation on the word manage, which is like— and this is expressed by the economist Justin Wolfers down there with Goodhart's Law— when a measure becomes a target, it ceases to be a good measure. I've actually blogged about this on pipeline coverage. I call it the self-fulfilling 3x pipeline coverage fallacy or prophecy, the self-fulfilling prophecy. So because you kind of ruin the metric by making it a target, because now everyone's managing the metric and it kind of loses its predictive value. It's kind of what Goodhart's law is saying. But, but to flip that over and to play the operator side of this, hey, if I'm running a SaaS business and I know that NDR matters a lot to my ability to raise funding, if not my valuation, why don't I just build it in? Why don't I, when I make a contract with a customer, say, hey, we think you're going to expand your usage over time. We think you're going to have 100 units of usage in year 1, 120 in year 2, and 140 in year 3. And how about I give you a price lock on those 3 numbers and I'll give you a 10% percent cushion if we estimate it wrong, and you're going to pay me those three numbers over the next three years, and wouldn't that be nice? You get price protection, you get some cushion on expansion, and I get built-in NDR. So right there, I have just locked in a near 20% net dollar expansion rate, right? I've traded away potentially some ARR now, but in my example, I actually didn't because the usage was naturally expanding, so I didn't really trade away anything other than the price lock, and it creates this notion of presold expansion. The first time I saw this, I was at a board meeting and there was a row in the spreadsheet called presold ARR, and I was like, what the heck is that? And the answer was, it was this. It was the $20K of expansion in year 2 and the $40K of expansion in year 3. I did a tweet a while ago about what do you call this, and the consensus average was initial ARR, kind of average ARR in this case, and then terminal ARR. So look, this complicates, as we talked about earlier, calculations and sales compulations, but the good news is you're locking in expansion upfront, and that's great. If you're going to get measured on this metric, why not build it in? There's a very similar thing that is bad in my opinion, which is price ramping, and the difference between the left column and the right column is in the right column, the usage and value is not changing over time. All we're doing is tilting the price. We're actually in effect saying you really owe me $120K a year, but it's kind of a financing strategy. I'm going to tip that line and, uh, it will drive on an ARR basis epic net dollar retention, but on a revenue basis, ASC 606 will flatten it out. It will see right through this trick and you will just end up with $120K per year in GAAP revenue. And this also has all the downsides of the left-hand column, uh, and in my mind it's playing accounting games. So there's a real shade of gray here, guys, between what I think is good and what I think is bad. But the fact of the matter is if the usage is actually expanding and you lock it in, that's smart. If you're just tilting the price and it's really a financing game, it's probably not good. So to finish this up, the next frontier is— I think the next metric we're going to see private companies start to look at is now kind of second billing in public companies. RPO, remaining performance obligation. It's the amount of contracted ARR you have under contract that is not kind of currently either been billed, which is revenue, or been paid in advance, which is deferred revenue. So if you're not releasing RPO, there is no other way to see it. The only way a company can— you can know a company's RPO is if they tell you, because it does not appear on the balance sheet. People sometimes confuse deferred revenue with RPO, and they're not the same thing because it only ends up as deferred revenue if they've paid you, in which case it ends up in your balance sheet as a liability. So this is not tracked much by private SaaS companies, but I think it's the next public company metric the private companies are going to start to look at either as they get bigger and start to think about going public or just because investors, private company investors, are going to want to know because a SaaS company that has booked a bunch of 3-year contracts that largely get paid is arguably worth more than one that hasn't. So that's remaining performance obligation. Let's wrap this up here with takeaways. Look, a SaaS company is really the sum of two businesses, a recurring business and a non-recurring business. Unit economics are great, but churn can be problematic. Problematic. When this churn gets polluted, it pollutes downstream metrics. The solution to all this is to zoom out, do a cohort-based analysis to say, how big was the base a year ago? How big is it today? That's called net dollar retention. A good rate is around 115%. Hey, if this matters so much, why don't we just build it in so we don't have to make it kind of a guessing game? And then going forward, little preview for the future, I think startups are increasingly going to look not only at their NDR, which is effective public company metric, but now RPO. So, uh, thank you all for coming. I will post the slides, and, uh, you can reach me at my blog or on Twitter @CalBlog if you have further questions. Again, thanks for coming. Good luck and happy expansion.