Gross Revenue Retention

Live from Schenectady, New York, it's SaaS Talk with the Metrics Brothers, Growth and CAC.

And I'm Growth, better known as Ray Reich, the founder and CEO of Benchmarkit.

And I'm CAC, better known as Dave Kellogg, independent consultant, EIR at Balderton Capital, and author of Kelblog.

And together we are the Metrics Brothers.

Indeed we are, Ray. Today, what, we're on episode number 13? We're in double digits here of the SaaS Talk podcast.

We are in double digits, but you know what I did to you? We moved this episode today, which is a follow-on to customer lifetime value. We're going to be talking about gross revenue retention. I moved it up to episode number 11 on us, Dave.

Okay, okay. So we got 2 more to wait for number 13, which you have to commit to not do on a Friday.

Promise you that. But boy, I couldn't believe we were talking about customer lifetime value, and at the end of our conversation, it was very apparent that we had to dig deep into gross revenue retention. So I am looking forward to that conversation.

Yep, I think we need to do it. It sounds like a simple metric, gross revenue retention. You'd think it'd be hard to talk about for 20 minutes, but I know we have a full plate of content. And there's a lot more kind of ambiguity and things to discuss than meets the eye.

But before we do that, Dave, let's take a quick break and jump into a quick word from our presenting sponsor, Maxio. The Metrics Brothers is proudly sponsored by Benchmarkit, the B2B AI and software industry's largest, broadest, and most contextualized benchmarking index. Dave Kellogg and Ray Groth Wright have put the Metrics Brothers podcast on hiatus for the remainder of 2026. All episodes we have ever recorded are still available on your favorite podcasting app. You can reach Dave at Kellogg and Ray on LinkedIn at Ray Reich. Thank you so much for listening to the Metrics Brothers podcast. Okay, Dave, let's get started. Let's start at the easiest macro level. What is Gross revenue retention.

Sure, so gross revenue retention, also known as GRR, is a SaaS metric that attempts to measure how much leaks out of your install base in a year. And it goes by several names. In the old days, I used to call it gross dollar retention. I've heard it called dollar-based gross revenue retention. Do you know any other names, Ray, that I'm missing there?

No, those are the two. I can tell you, 'cause I guess I'm not a very global person, but when I first started talking a lot about benchmarks, I would call it GDR, gross dollar retention, and my European and Asian friends were like, could you become a global citizen and start using GRR? 'Cause we're not dollar-based. So I think it needs to be GRR, Dave.

Correct. I had the same thing. I used to call it the net equivalent, NDR, all the time. And I discovered really that it's a little bit like the Pulp Fiction problem where the, you know, why do they call it a Burger Royale instead of a Quarter Pounder? And the answer is, you know, they don't know what a quarter pound is. So yeah, we shouldn't burn a currency, i.e., the US dollar, into the name of a metric. The intent here is clear that we're trying to say what it's not. You know, it's not logos. We're not saying we had 100 customers and ended with 90 customers, therefore we have a 90% logo retention rate. And by the way, if you wanted to count logos, that's personally what I would call it. The intent here, whether you say dollar, which is really a mistake, or you say revenue, which is the right answer, is to say this is a value-based metric and not a count-based metric.

I think it's a really good, clear delineation. But then we have to go to the next level and it's like, well, how do we calculate that? And that's where the nuances start to really explode. So how do you calculate GRR?

So look, I believe that both NRR— and I believe this passionately, Ray— that both NRR and GRR should be defined in terms of cohorts. And when I mean defined, I mean like there is one and only one correct answer. And you get it by using cohorts. There may be shortcuts or alternative ways that work under certain conditions that produce the right answer under certain conditions, but to me, we have to kind of decide which is the definition and which is the shortcut. And my argument is NRR and GRR are both defined in terms of cohorts. So specifically what I mean for GRR, we would put in the denominator— Basically the year-ago value of the year-ago cohort of customers. So a year ago, I have customers 1 through 10. They're each 100 units in ARR. I have 1,000 units of ARR. That's the denominator. In the numerator, I take those exact same customers, 1 through 10, and I add up their current ARR, including any shrinkage, but excluding any expansion. And that's the difference between GRR and NRR. Right. NRR includes both. GRR specifically excludes expansion. So we're basically looking at the shrinkage of a cohort of customers and say their current value of that year ago cohort is 95. I would, or sorry, 950. I would take the 950, divide it by the 1,000 and say we have a 95% GRR.

So that's one way, and that's using a cohort, which really simply means Let's say it's those 100 accounts from a year ago, and you're looking at those same 100 accounts today to do the calculation, correct?

Correct. And there's 2 things that can, well, there's a lot of things that could go wrong, but the one that bothers me the most is when people do it backwards and they say, hey, let's look at 100 customers today and see what they were worth a year ago. And that's what, you laugh, that's what's known as survivor bias, right? 'Cause we did it backwards. We were supposed to say, hey, let's go back a year in time, Count all the customers. Oh, there's 100 of them in this case. And then how many of those do I have today? If I do it the other way, it's inherently survivor bias, which will give you kind of an optimistic view on the metric.

Yeah, because you're basically eliminating all the churned accounts.

Yeah, excluding, as I always say, excluding the people who stopped doing business with us, what's our churn rate?

Okay, so that's the cohort method. Now I've worked with several companies who use more of a mathematical formula model. And that is you take your beginning ARR, you subtract the churned and down-sold ARR, and then you divide that by the beginning ARR. That gives you a decimal, you multiply by 100, you get the percentage. So why do you like the cohort model better than that formula model?

So, and this is a, I would admit, a Dave K idiosyncrasy, but, but I'm gonna say that there should be one definition That kind of define the semantics of what it means. And then you may have other ways of calculating this. There's a— basically in relational databases, my tech background, the definition of a join is the Cartesian product of 2 tables followed by applying the WHERE clause that defines the right answer. Nobody actually calculates it that way. But if you want to know what the right answer to a query is, you go back to there and say, this is the correct answer. Now let's find faster, easy ways to calculate it. This is kind of the same question. I'm going to argue that the definitional answer has to be in terms of cohorts.

Yeah.

And if you have a shortcut that can calculate it using, for example, what I would call a bridge— what you're calling the formula method, I would call the bridge method, right? You're kind of using an ARR bridge. Starting plus new minus churn equals ending. And in this case, you're excluding the plus news. It will produce the same answer in many, many cases. But the problem is it won't produce the same answer all the time. And this, Ray, is why I believe as would-be custodians of SAS standards— and you're more than would-be, you run the SAS Demetric Standards Board.

Yeah.

That we should say there's one answer. That when people come to us and say, what is the definition of this metric? We say one answer. And to me, the answer is define them in terms of cohorts. Can you calculate it using bridges as you described? Yes. Will it usually get the right answer? Yes. Will it always get the right answer? No. The simplest example of a case that'll miss, Ray, and I'm going to contrive it a little bit, but just say you do month-to-month contracts and I start with a cohort of customers in January.

Right.

I signed somebody in February, they're not counted because they weren't in the cohort, but then they churn in March. And if you're just looking at churn ARR or shrinkage ARR, they'll count. So, and that's always the problem when you lose the cohort. And most people don't separate, you know, expansion from customers who were in the cohort a year ago from expansions from post-cohort timeframe joining customers.

Right.

And that's where the bridge breaks down. Look, in a year of 1-year contracts, It won't happen. So if all you do is 1-year contracts, then everybody in the starting— it will work, right? Because everybody in the starting cohort will have been a customer a year ago and it'll work. But in that scenario, if you're doing monthly contracts, it won't, in my opinion.

And one other simple test. I am the product of a public school education, Dave. So I always say the other thing is make sure that the ending ARR is never bigger than the beginning ARR. I.e., you cannot have a GRR over 100% because if you do, that means you're adding some level of expansion revenue into your calculation model. Would you agree with that as a simple never should be under 100%? Over 100%, yes.

Never should be over 100%. Yeah, absolutely. GRR definitionally is maximized at 100%, and if you're getting an answer greater than 100%, you're doing something wrong. And as for the bridge method versus the cohort method, I honestly, let me get practical on you, Ray, because I like the theory of it too. But the practical answer is I think it's simpler. I just got to divide number 2 by number 1. And I never have to ask, what is churn? When does churn occur? And it may not seem like a hard thing because you think we're tracking it, but we're not separating it, right? I don't even have to ask. I have to say, hey, what were those guys worth? How many customers do we have a year ago? 1,000 or 100 in your example. Okay, how many of those people are still customers? 95. Great. What's their total ARR today? Great. Let's divide that by their total ARR a year ago. Boom. I divide 2 numbers and I get an answer. As soon as you go to the bridge, I think it has many downsides, including the fact that you can get the wrong answer. The other one I'd actually argue is it tends to put you into a quarter-over-quarter mode. You start looking at it quarter over quarter, and it's supposed to be a year-over-year metric, right? Like when we talk about GRR, implicit in the definition That is true.

Another thing that's implicit, it's gross revenue retention. I have a question for you, Dave, and this is the thing that I get into debates about this, and it's like I feel like I'm being too precise, and that is gross revenue retention rate is not the same as churn rate. What do you think about churn rate versus gross revenue retention rate?

I mean, I'll do my short stump speech on churn because I don't want to bore Frequent listeners, but I think there are 4 churn rates, as you know, that it's basically a little 2 by 2 matrix. And the question on the x-axis is, is you going to use the whole ARR pool, i.e., starting ARR, or are you going to use only that portion of it up for renewal in the period? So if we're looking at churn over a year, are we looking either at starting ARR for the year, or are we going to look at the sum of all the contracts that renewed in the year? And again, for some people that will be the same thing. If you just say you do only 3-year deals and it's year 2 of your company, you're actually going to have no ATR, right? There won't be any because no contract is up for renewal yet. So for some companies, the difference is dramatic. For some companies, there's no difference at all. At the last company I ran, when I ran it, Host Analytics, now called Planful, we ran about half 1-year deals, half 3-year deals at the time. So it made a big difference in our rates whether you picked The whole ARR base or whether you pick the ATR base, the available-to-renew base. So that's the x-axis. The y-axis is, are we doing it on a gross or a net basis? And so now I— so when you say, is GRR equal to churn? I have to say, which one? And I think it is. I think GRR is equivalent to what I would call gross ARR pool-based churn, i.e., I'm going to add up all the churn and I'm going to divide it by the starting ARR. And I think that's going to produce the same number as GRR.

Now, this, this is really kind of geek city, but for me, retention, you know, good GRR numbers, 88%, 90%, 92%. 1 minus that is my gross churn rate on an ARR basis. Because if I have a 90% gross revenue retention, I have a 10% gross churn rate on an ARR basis. That's the way I think of it.

Yeah, it's not a bad way to think of it, Ray. I guess I'm gonna be more literal and sometimes to a fault, as we know, but because I've found churn a little bit hard to count sometimes, and because churn rates can be ambiguous 'cause of this little 2x2 matrix, I don't love to count churn. And that's one of the things I like about retention is I don't have to ask. I never have to ask what churn is. I just say, what are they worth now? What were they worth then? Boom, I have an answer. And I feel like it lets me kind of jump over a potential swamp. Now, if you're asking me what a churn rate is, I have to calculate churn.

Right?

I mean, you're saying at what rate does churn, you know, does ARR churn out of the base? So I'm going to be forced to count churn. So I'm much more willing when I'm asked for a churn rate to go count churn. But I do believe your math is correct, that 1 minus GRR will produce the churn rate that I call gross ARR pool-based churn. So I think we agree on this one. Right.

Okay, now I think this is the most interesting, but once again, I'm geeky. And that is this whole concept of how to use available to renew, ATR, and your retention rates. And you are a fan of saying, hey, I just use, whether I have 1-year contracts or 3-year contracts, I put all that ARR in there and that's how I calculate gross revenue retention rates. And I'm saying as an operator, when I do have contracts coming up for renewal, where I actually have to get them to renew. So it might be some negotiation. I want to know for only those accounts that have subscription agreements ready for renewal, what percentage of them on a gross ARR basis renew? Do you think it's a good second way to calculate GRR to give you that insight, Dave?

So look, I do look at what I would call ATR churn rate. So I think the answer is going to be yes. I think they just tell you different things. Look, if you're trying to say, hey, my customers are really happy, 90% of them renew or 100% of them renew, then you're clearly trying to use the metric as kind of a proxy for customer satisfaction, right? You might say, hey, look, our NPS is 45 and our churn rate's only 3%. 97% of our people renew, right? Look how happy our customers are. When you're trying to use it as a kind of CSAT proxy, I think you should use ATR churn. Because maybe the reason, you know, 98% of them renewed was because only 2%, you know, had a contract up for renewal in the period and both of them didn't, right? So in order to get rid of that problem, I think we should use ATR of basically what's the percent of people who had a chance to not renew, right? What percent of them renewed? And that's what I'm trying to figure out if your customers like you, if they're happy, CSAT proxy, all that stuff. Conversely, if I'm just going to go into like, what's your installed base worth? Like what's the financial value of your installed base? I like to use the whole ARR pool because whether or not somebody would have renewed if they were given a chance is moot. Because if, you know, 3 quarters of my customer base is on multi-year deals, they don't get to renew. They're effectively auto-renewing. And what I'm trying to say, how much is this base worth? I think I need to include that. So the answer to me is there's a place for both these metrics. In just kind of GRR by default, I think you have to use the whole ARR base because it's almost built into the cohort-based definition, right? There's nothing about whether or not they have a contract up for renewal in the cohort-based definition of GRR I gave you. So to me, it's tricky, but in my little churn rate matrix, there is a box where you look at basically gross and you look at ATR-based, and that's the box I look at. for trying to figure out if the customers are happy.

See, this conversation between you and I is so important why executive room tables around the world should be having these discussions and debates. Because when this first came up a couple of weeks ago, man, I was an ATR bigot. I wanted everything to be ATR. And you were telling me why you wanted to use all ARR. And what's interesting, after we had that discussion, I bet we had it for an hour, hour and a half. We came to the same place, I think, which is there's a place and time for ATR versus just aggregate ARR.

Absolutely. Yeah. I mean, our listeners should know that Ray and I just don't turn on the mic, talk about this and not talk to each other for a week. We do sometimes, you know, get into long and extended conversations about, hey, wait a minute, something just came up and what do you think about it? And I think we both learn in that process. You know, I'm in the middle, I'll preview an upcoming episode. I'm in the middle of a deep dive into LTV to CAC that got Kind of prompted by one of our last episodes. So we do this a lot. And yeah, I think it's important that people understand these metrics. I mean, that's why we do this podcast, because while it may seem really geeky, if you're going to ask what a churn rate is, which in general is 1 minus the retention rate, I got to know what you're counting. Are we looking at the whole base? Are we looking at those people up for renewal? Are we doing it on a net or a gross basis? And it may sound really basic, but by the way, you're going to use that metric to build other metrics like lifetime value. So you really need to understand, right? If you're going to put churn in the denominator and put 1 on top and say that's your lifetime, It'd be really nice if you understood what that churn rate was.

By the way, I have to tell you, Dave, as much as I appreciated that discussion, we got to a good place and it's coming out today. My wife asked me to please stop having those conversations in the public part of our house.

Yeah, she's like, this is what you do for a living?

Oh my gosh.

Yeah, my wife's mad because we have to crate the dog during the podcast because he'll start barking.

Well, let's, let's go to the— I think one of the most important things. So we talk a lot about the metric, how to calculate it. We talk about all these nuances, but I think what's really important, especially to a lot of the, you know, earlier career people in this industry, why, how do you use GRR to really inform your decision-making?

So, so there's 2 things here, Ray. And unfortunately, I'm gonna have to bring up NRR, which we'll probably have to do another dedicated episode on. But, but to me, the first metric I go to in this little R family is NRR, because I think NRR is an attempt to value the install base. Like, okay, you have an installed base of customers worth $100 a year ago. What are they worth today? $112. Great. So kind of, it's like flying a plane with a 12-mile-an-hour tailwind. Hey, this is great, right? I've got it. And conversely, right? If on a net basis, I say, hey, what was the customers worth a year ago? $100. And what are they worth now? And you say, you know, $88 on a net basis, then I'm like, holy cow, we've got a 12-mile-an-hour headwind. So the first metric I actually look at is net revenue retention. But then it immediately begs the question, especially in tougher times: Do we have what in French they call one train hiding another? At French railroad crossings, I love this sign. There's there's always a sign at a railroad crossing in France that says "Un train peut en cacher un autre," and it means one train can hide another. So so if you're looking at one train and it clears, you start driving across the track. You may get hit by another train coming in the opposite direction. And this is the way I feel about NRR. which is one train can hide another, meaning really good expansion can hide really terrifying churn. And that's why I like GRR, 'cause I'm gonna say, hey, what's your NRR? And you go 140%. I'm gonna be like, wow, that's flipping amazing. And then I say, what's your GRR? And you go 70. I'm gonna be like, holy cow, you guys are amazing at expansion, but you've got one heck of a churn problem. And that's why I like GRR, because it doesn't let you hide behind great expansion. Ultimately, if I wanna value the base, I wanna know the net. But if I want to look for trouble, I want to look at GRR.

Yeah. You know what's interesting? We're starting to get a lot of people reaching out to us, Dave, and asking questions, sending it to our mailbag. And Tom Lawton, who's the chief revenue officer at Chally, he asked a question. So let's, let's end today's episode with his question. That is, should we be excluding variable revenue in a usage-based pricing model from our GRR calculations?

Okay, well, 2 things, Ray. One, I gotta finish answering the prior question, then I'll answer that one.

Okay.

'Cause I do want us to become a call-in show like Car Talk one day. So, but the second answer to your question is it's the one train hiding another issue. It's why I like GRR, 'cause it's kind of a trouble detector. The other reason is you can find troubled segments, right? We can look at GRR and NRR by segment, SMB, mid-market, enterprise. We can look by vertical, we can look by product. So it's a great way to look if you're looking at NRR, what's particularly healthy. And if you're looking at GRR, Where are the potential trouble signs? So that's the other use of the metric, either kind of look at the whole base at once or to look at pieces of it. Now, to answer Todd's question.

By the way, Dave.

Yeah.

Let me double-click on that though, because if you've grown and you've got multiple target markets, you know, like you said, SMB and enterprise or mid-market enterprise, or you go global, if you're not calculating GRR or NRR on a segment-by-segment basis, I think you're missing 80% of the value. I'm making that number up, but I think that's critical to do that.

Yeah, look, I would argue that the PE formula for acquiring a semi-troubled software company is to, during diligence or during the sale process, go analyze all the customers by segment, find the good segments, figure out what the company's worth based on those segments, buy it, shut down the bad ones, invest in the good ones. So it's an integral part. Potentially, if you're going to sell your company to PE, they're going to do that exercise for you. So you may as well do it yourself. Go find your good segments and your bad ones.

Well, do you think we can pivot to Tom's question?

Yeah, I think we can. So unfortunately, this opens up a really big can of worms for me because, because by Xloone, he means variable revenue. He's talking about a model where there might be a base subscription plus fees on top. So we have what I would call monthly varying spend, also known as, or it can happen with usage-based pricing or consumption-based pricing models. I genericize that to just talking about monthly varying SaaS, that we have 2 types of SaaS. You spend 120 units a year, which is 10 units a month. It's kind of flat. And there's monthly varying SaaS where maybe either with or without some baseline fee, you're getting some usage fee on top. So your monthly spend is varying, right? And that's a consequence of the pricing model. And obviously usage-based pricing is super popular these days, as are these hybrid pricing models. So this is going to come up a lot. And my answer, and I think we may disagree on this one, is the question is, what do you use for ARR? Right? Because I just defined a metric in terms of ARR a year ago and ARR today. Well, shit, we don't have ARR, right? We've got this baseline fee plus some varying overage charge, right? So the answer is, what do you use? And there's 2 different approaches. I think you take one, I take the other. I'll tell you mine. Mine is to say that basically I want to get a proxy for ARR, which is trailing spend. So I'm going to do it kind of the Snowflake way and say, okay, if you want a GRR calculation, I'm going to go back 24 months. And I'm going to look at, you know, year 2 to year 1, what did you spend? And then year 1 to now, what did you spend? And I'm going to exclude any expansion from that, which is going to be a little tricky when we— because how do you define expansion versus normal fluctuation, right? This is the problem that showed up in the Klaviyo S-1, if you followed that debate. But that's what I would try to do. I would use trailing spend as a proxy, and trailing spend does include both base and variable fees. I would just want to say you spent $100 with me a year ago.

Yeah.

You spend 95 this year, therefore you're down 5 units. Now the question is, of course, I'm not even sure GRR is meaningful in this case. NRR is, but on GRR, and I've opened up a real rat hole where this is literally the debate I'm in the middle of on Twitter, is what does GRR mean with a subscription-based pricing model? So I could do it for NRR. I would say in my case, the NRR in my example would be 95%. For GRR, I'm not even sure what it means to exclude expansion because I'm not I have no expansion. I'm just looking at spend. How much did you spend?

Yeah. So, but to your point, if all your accounts are growing, right, you're going to get a GRR greater than 100% using your model.

Well, I won't because I'm going to exclude expansion, but I don't know what expansion is. So I think I refuse to answer the— the witness will not answer the question on what's GRR. I can give you an answer on NRR, but you're right. But I wouldn't even know how to calculate it, Ray, to be honest, because they say now go exclude expansion. I'm like, well, what the heck does that mean? I would have to say if February is greater than January, is that expansion? I don't know, but I'm just looking at spend. I'm not sure, Ray. I'd have to chew on it more. But the bottom line is, by the way, what Klaviyo does is they exclude shrinkage. So what's weird, we may want to do an episode of the Klaviyo NRR/GRR calculations, but I could say I agree with you, GRR can never be greater than 100%. I will say that I think it's a big problem/challenge to take SaaS metrics built around annual flat subscriptions and translate them to a kind of monthly varying usage-based world.

Dead on. I say GRR if you're a more traditional subscription or a hybrid where you have a minimum subscription plus usage. I exclude the usage from my GRR calculation. And for those who debate me on that, then I say, then let's just use NRR. And that's the most important retention metric I would use in that hybrid environment.

Yeah. I mean, frankly, I wouldn't even know how to calculate it not doing it. At first, I knew you were going to say that because I know how you do it. And I think it's striking me in real time as a reasonable way to do GRR because I can't even calculate GRR using my method. So yeah, I see where you're coming from.

Well, today's been a great episode and sorry, I think we ended it a little bit on a, we don't have all the answers, but that's the reason we do this. Nobody has all the answers and it's these discussions and getting alignment and knowing that you're measuring your business by having a common definition. of the metrics you're using, right, Dave?

Right, Ray. Measure on.

Measure on. Okay, thanks for today, and we'll be looking forward to our next episode. Bye, everyone.

The Metrics Brothers: AI, SaaS, and More is a production of The Metrics Brothers Growth and CAC and a member of the Benchmarket Podcast Network. By accessing this podcast, you acknowledge that The Metrics Brothers make no warranty, guarantee, or representation as to the accuracy or sufficiency of the information provided or the humor content of the jokes told. Ray, information, opinions, and recommendations presented are, according to our spouses, probably wrong and provided for information purposes only. This podcast should not be considered professional or, for that matter, unprofessional advice. We disclaim any and all liability for any direct, indirect, indirect, misdirect, incidental, special, ordinary, consequential, inconsequential, or other damages arising out of any use of or God help you rely upon the information presented here. Ray Grothreich is based in New York City and available on X @rayreich. Dave Kellogg is based in Silicon Valley and available on X @kelblog. Schedectady, which is French for unspellable, is not our actual production location. Thanks for listening.