Net 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 Kellogg.

And together we are the Metrics Brothers.

An unlikely pair at that, Ray. Lucky number 13 is today's episode. Am I right?

You are right. And I will tell you, Dave, you were like a stallion, an unbridled stallion. The last 2 episodes we were talking about gross revenue retention and the Klaviyo metrics, and I'm even saying Klaviyo right now. And you wanted to get to NRR, net revenue retention. and move away from gross revenue retention. So I thought, why not make today's episode all about NRR? Dave, what do you think?

I think it's a great idea. And just to be clear, I like GRR. To me, it's the second question in this category. The first one is, tell me about your NRR. And then to make sure you're not hiding something, I say, now tell me your GRR. So they're both important. But yeah, I would typically lead with NRR. That said, before we jump in to those details, I think we should take 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 @kellogg and Ray on LinkedIn @rayreich. Thank you so much for listening to the Metrics Brothers podcast. Okay, Dave, so maybe we start this like we did last time, and that is let's start with the basic definition of what net revenue retention is.

Got it. NRR, also known as NDR, net dollar retention, is a measurement in my mind for kind of valuing the installed base of a SaaS company. Specifically, it says, how did your installed base expand over the past year? There are 2 ways to calculate it. One is the cohort method, where you basically look at the year-ago cohort of customers and you take their ARR a year ago, put that on bottom, take their ARR today, put it on top, divide the 2 to get NRR. That's the cohort method. What I'll call the bridge method is where you take starting ARR, then you add all new ARR from existing customers. So kind of new expansion ARR, basically starting, plus expansion minus churn divided by starting. And that's called the bridge method. I have some detailed issues with it, but a lot of people calculate NRR that way.

Yeah, I wanna make sure on the bridge method, we take the starting plus new minus churn. We also are gonna subtract any downsells from that, right? That's not really churn, that's kind of that downsell. I go from $100K to a $50K contract, correct?

Correct, it should be included. Some of it's semantics. I did just like I defined new ARR as having 2 flavors. You know, new ARR, where does it come from? Well, from 2 people, right? From existing customers and from new logo customers. Where does churn ARR come from? Well, right, from 2 things. From downsell, i.e., shrinking customers, and from lost customers. So, I would definitely include both downsell and lost in churn. To me, they are the 2 types.

Okay. Now, I actually don't have a major issue with the bridge method or the formula model as I call it. But I know you've even written a long blog on this. So, can you just at a, maybe a conceptual level, what your concern is with that bridge formula model?

So look, at the conceptual level, I have an issue with it, which is it's always dangerous to have 2 definitions for something, right? Because then you have to kind of do a proof to ensure they'll always produce the same result. And in my mind, it's much simpler just to have one definition and then say, hey, if you've got a shortcut way of calculating that metric, then good for you. Example, CAC payback period. I have a shortcut that gets the right answer. CAC ratio divided by subscription gross margin, right? That will produce CAC payback period in years. So times 12 to get months. That's a shortcut. To me, it's not definitional, right? The definition we covered in another episode. I don't want to raise it. But to me, the bridge is a shortcut method for basically usually getting the same result as the snapshot method.

Got you. And even though you talk about the cohort method being your favorite, The other thing I would say is if you've got a multi-customer segment environment, you have SMBs and enterprise and mid-market, I strongly recommend that you do a segment-based calculation of NRR also and not just look at your overall customer base because you might find that some customers just aren't as attractive over time as they are over the first year or so, right?

Yeah, separate point, but for most of the metrics we talk about, Ray, it's a great idea to segment them in your customer, whether it's by vertical, by customer size, by number of employees. A big part of SaaS is to try and identify what's working. And certainly segment-based SaaS metrics is a great way of doing that. But back to your initial question, I think there should be one way of calculating it, one way. I think it's definitional that the answer is by cohort. And if you have a formula that produces the same answer, good for you and go ahead and use it. Now, specifically, my issue with the bridge formula is you can basically lose the cohort. Because if you look at most companies, let's just imagine a quarterly bridge where we have starting plus new minus churn equals ending. That new ARR is going to include 2 subtypes, expansion from existing customers, right? And new logo from new customers. In that expansion number, The cohort gets lost. I.e., if somebody buys in January and expands in April, they're going to show up in the Q2 expansion number, but they were not— well, they bought in January, so they bought in February. They were not in your cohort. They weren't in your January 1st cohort, at least. So the same thing can happen with shrinkage. Somebody buys in February and they shrink in May, then that shrinkage is going to show up in churn ARR in Q2, May. That's a problem. Because they're not in the cohort, right? So to me, the issue with the bridge method is you can lose track of the cohort. So that's why I separate what I call net customer expansion from NRR. Net customer expansion is literally just new ARR from existing customers, i.e., expansion ARR plus downsell churn ARR, basically, right? That's telling you what's happened to your base. I think it's a very useful metric just to say what happened to my customer base in terms of raw size.

Yeah.

But for NRR, we're supposed to be doing a cohort-based calculation. In my mind, that was like the point of NRR, that we could step around a lot of messy math and definitions of churn and other crap and just say, a year ago, what were these guys worth? What are they worth today? Boom, I have a number. And when you use the bridge method, you kind of step into that swamp.

Now, we didn't talk about this at all, so I'm going to throw a curveball here to you, but hopefully you're, you're in the batter box and you're ready to go. If I have an NRR of 120% and my GRR is 90%, would you agree that my expansion rate is 30%, just NRR minus GRR? Is that a very dangerous, too simple of a model?

That would not be the same as net customer expansion, but that would— people in my mind don't talk about an expansion rate. But yeah, that would be— I understand your math and I think it's correct.

Okay. But that's an important point. So let's just throw out, what is net expansion?

Net expansion would be as a metric for a period. And it says in that period, let's add up 2 things, the new ARR from existing customers, and then subtract the shrinking ARR from existing customers, right? Which is definitionally from existing customers. That's net customer base expansion or shrinkage. It's not a cohort-based metric. It's a period metric, right? So what was it in Q2? What was it in 2021? Right? You can look at it for a period. And the point I'm trying to make, and I think we should move off it after this, is just be careful if you're using the bridge method, particularly if you run on monthly SaaS, because people are expanding and shrinking all the time. Right? So somebody buys in February, they shrink in March, they grow in April. In pure annual SaaS, you can expand across a year. It's actually impossible to shrink, right? Virtually impossible to shrink because you have a contract saying you won't in pure normal annual SaaS. So this is why I'm calling the bridge method a shortcut. It works under some circumstances.

Right.

If the circumstances are that people never expand or shrink within the year of a contract, then it'll work. If they expand only, then it's actually going to, well, it will include that expansion where it potentially shouldn't, right? It could include expansion of people not in the cohort. So anyway, that's why I view it as basically you have the definition, which is based on cohorts, and then you can use the bridge as a shortcut, but make sure in your situation, it's going to produce the same answer. Easiest way to do that, Ray, is calculate them both. And if you ever get a difference, like one of those must-be-zero cells you put in a spreadsheet, right, then you're going to go, what's going on here?

Got you. It's a good idea. Okay. Everyone's talking about how does usage-based pricing or consumption-based pricing impact how you calculate NRR. And it's interesting because I don't calculate NRR on a monthly basis. I believe it's an annual or annualized So either I look at an entire fiscal year and I look at current fiscal year versus previous fiscal year on a cohort basis, or if I want to do it monthly, then I just look at 12-month trailing spend. So I don't have to be concerned of if it goes up month 1, down month 2, back up month 3. Does that make sense as far as that's how you'd calculate NRR in a variable SaaS model?

Yeah, I'm not sure I need to characterize the metric as specific to a certain type of period, but But yes, short answer is I like using spend. Look, Ray, the problem, as we've talked about before, is all these formulas we use were invented in a world of pure annual SaaS, right? ARR, and did ARR go up or down, right? How long does it take to pay back the sales and marketing expense of a dollar of new ARR, right? They're all based on this notion of ARR. In usage-based or consumption-based pricing, You don't have ARR, right? You may have a base fee and an overage fee, right? Some people like to do the metrics only on the base fee and ignore the overage, which I think is weird. I can understand the motivation for it, but I still think it's weird. There's a reality of what your bill was every month, right? And I like working with that reality. So to me, the problem is we have formulas built for a world where there was a thing called ARR. And in monthly varying SaaS, which usually is a result of usage-based or consumption-based pricing, we don't have any real concept of ARR. Therefore, we need to find a proxy. Some people use trailing spend. I'm with you, Ray. Some people might use trailing spend for a quarter and then annualize it, right? They'll take Q4 trailing spend and analyze it and compare those. So you're really comparing Q4 to Q4 to do NRR. Highly seasonal business might make sense. Some people will annualize a month and say, oh, we'll take 12 times this December and compare it to 12 times December a year ago. Klaviyo, as we saw last week, will take 12 times the expected next month in a rather bizarre twist of fate that you can hear about on that episode. Some people will do 4 times the most recent quarter. So the trailing's better, 12 times most recent month. So the problem is we need to find a proxy for ARR. People use a lot of different ones. And I think you and I both Prefer trailing spend in some ways because it's simple and definitive. There's no opinion, right? I would say, what is it? Profit is an opinion. Cash is a fact, right? I feel like this is not an opinion. Spend is not an opinion. We know how much they spent in December or, you know, in Q4.

Well, let's talk about another opinion of the way to do it in these hyper-land and expand models. And I always like to talk about The Snowflake, and that's the 2-year lookback process where, you know, their enterprise customers may take 9, 12, 18 months to fully ramp till they get to a steady state. And even then, it's not a steady state because they continue to expand. So they actually do something very interesting, Dave, and correct me if I'm, I'm not wrong here on the 2-year lookback period. They look at a cohort of customers from 2 years ago and then they compare That kind of months 13 through 24 versus the most recent 12 months. And that's how they calculate their NRR. Is that correct?

That is correct. Yeah. And they're doing that because they're using trailing spend, right? Because if you're not using trailing spend, if you actually have ARR, I could just go back one year, get the year ago cohort of customers, ask what their ARR is. By the way, that ARR is a forward-looking commitment, right? I signed a contract that I'll spend 100 units with you in the coming year. So, we go back a year kind of looking forward, and then we compare that to right now and say, okay, right now looking at the forward commitments, what do we have? Let's compare ARR today to ARR a year ago. So, that's the classic NRR formula. If you're going to use trailing spend, the first time I saw Snowflake's calculation, it freaked me out. I was like, what the hell is this? But if you're going to use trailing spend, the road leads to this calculation, which is, wait a minute, I need to go back 2 years and get that cohort of customers.

Yeah.

I'm going to take their trailing spend for months 24 to 13, and then I'm going to put that on the bottom. And then on top, I'm going to put the most recent year trailing spend on top to get MRR. I think it's a completely valid way of doing it. And I think trailing spend kind of leads you nowhere else. Yeah.

And one caution I would give to the listening audience, especially if you kind of use that 2-year lookback, Snowflake made a little bit of a mistake once where they looked at the cohort of customers from the last 12 months and then compare that to those customers that were on contract 13 to 24 months ago. But inherently that eliminated all the churn customers. So you got to start with the customers from 2 years ago and then compare that to the customers over the last 12 months. If that made sense what I just said.

Well, it made sense to me. I didn't think they did it that way, but maybe at some point they did or they made a mistake at some time.

Just one quarter, it was a mistake and they corrected it.

Okay, got it. Because their definition, I remember the definition from the S-1, or maybe it was 10-K, but the definition was pretty clear that they were not supposed to do that. It sounds like somebody, a new guy in FP&A, new gal in FP&A.

Now, so one of the other things I hear all the time, and maybe not as much in the last 6 months, but 12 months ago, oh wow, NRR should be 140%, 160%. Look at Snowflake, look at Datadog, look at Twilio. So Dave, kind of where's your perspective on Good, better, best.

Yeah. So you're the benchmarking guy. So I'm gonna kick this back to you very quickly. But in my opinion, 120+ is very, very good. 120 and 120+ are like very, very good numbers. Aspire to 120+. Like 150, 160, 180 is now, you know, that's epic, but they're not getting those anymore. Even Snowflake is down from 180. I don't know their latest. I'm guessing it's 130 or 140. So I would say for the rest of us, right, if you're not Snowflake, 120+ is very, very good. 120 is very good. Then 110 is getting to be more typical. 105 is, I think, around your benchmark at median. You can update that in a second. And probably most importantly, less than 100, to me, that's a bright line and that's bad. But before we talk about the less than 100, why don't you tell us what the data says, Ray?

You know, thank God you've been reading benchmark.ai. But seriously, it depends on the ACV. That's such a high corollary to what a Good, better, best is. But across the entire population and private companies, our most recent is 105%. But that less than 100%, Dave, I will tell you, if you have a less than $2,500, especially a less than $1,000 product and you're not multi-product, you might see 85%, 95% NRR because you don't have a lot of upsell, cross-sell opportunities. But once you get to $10K and above, you're exactly right. It should be 100%, should be the minimum threshold, and you're looking to get up from there.

That makes sense. I mean, look, from an investor perspective, my belief is that 105 would be like kind of a yawner, like, okay, great, you got 105. 110 is like, oh, it's respectable. I'm now just going to look at the same numbers through an investor view, whereas 105 is like, okay, you know, below 100, I'm like, whoa, yellow light. Like, what's up? And you're right, small ACV, maybe that's the reason, but I'm going to be asking, right? Why is it below? 105, I'll yawn. I won't get excited. I won't walk out of the room. 110, I'll be like, okay, respectable, solid. 120, I'll be like, hey, it's pretty good. Anything above 120, I'm going to be whipping out my checkbook. So I think that's kind of the investor view. Let's talk a little bit more. I mean, look, below 100 scares me because it means your bucket leaks on a net basis. And it means if we just stand, like, look, the 110%, 120% NRR is kind of magical because like in theory, we could just stop everything and look at our bucket and every year it'll grow 20%. Now we know it's not quite that simple, but that's kind of the underlying flawed logic. Whereas if you're less than 100%, it means if we do nothing, this thing bleeds out, right? That it means, oh, it's going to be worth 90, then 81, then 72, then 63, right? And after 9 years of looking at it, it's going to be down to almost zero. And I think that's a bright line that freaks out investors. And you're right, I know companies that have an NRR less than 100, but personally, I find it scary.

Well, I must admit, I don't have a lot of deep experience with low-value products, and I'm talking $500, $800 a year being sold to the SMP or the SOHO market, kind of the single-owner home office. So that's where I see it, Dave.

And I must admit, I don't know enough about it to have There's a metric we touched on last time that we won't drill into here, Ray, but we called it kind of the GRR/NRR gap. And somebody had done a little research on that. I can't remember. I emailed it to you and we talked about it. And in the end, maybe we should do an episode on it. I think we thought it was only quasi-episode worthy, but certainly the lower the number gets, I do want to know, because you're basically arguing that when NRR gets to be below 100, it just might be because there's very little gap between NRR and GRR.

That's right.

That gap is super tight because there is no upsell or expansion. And that's probably a good thing. Like if I were to have a low GRR, I would want to have a tight gap. I would never want to have upsell of 30% and a GRR of 95, or sorry, an NRR of 95, right? Because it means I'm absolutely bleeding out on a GRR basis.

Yeah, no, totally agree. Hey, you know, we always talked about, we wanted to make this podcast with some audience participation. You think we could do a couple of those today?

Yeah, let's do it.

Okay. So the first one, and I really love when people do this on LinkedIn and Twitter, but it was Jonathan Corey, the CEO of Precursive, and he said, what role should professional services play in influencing NRR? And more importantly, at what cost?

So, and Ray, I'll ask you the next question. So Jonathan, how you doing? Hope you're listening. Look, the role to me is to make it high, as you know. I mean, my personal mission statement for PS in a SaaS company is maximize ARR while not losing money. So, how can we maximize ARR? Well, maximizing NRR certainly helps. Happy customers buy more. I actually think mathematically, the bigger impact would be saves, right? Like, helping CS grow somebody from 100 to 110 is nice, but saving a customer from going 100 to 0 is really nice, right? You have to do 10 of the former to get 1 of the latter. So, I've always viewed just like relief pitchers in baseball, I've viewed PS as kind of the save guys, right? Like put them on the mound in the 9th inning to get the save. So I'd say first and foremost, be willing to break even, maximize ARR without losing money, focus on saves, and then focus on expansion. Phil, let me go to the next question, Ray, and I'm going to kick it over to you. This one comes from Tom Krueger. Looks like he's a B2B SaaS executive and CFO. His question is, How would you handle product swaps, ARR moving from one product to another, comma, in a multi-product company when we're looking at NRR, GRR, et cetera?

That can get very messy very quickly. And a lot of time, even in multi-product companies, the contracts don't even always spell out you're paying X for product A and Y for product Z. So I try to do it holistically at the account level, Dave. And I honestly haven't spent a lot of time looking at product swaps from an NRR perspective. Now I do, I have done NRR calculations by product. So if I have a multi-product portfolio, I will look at those customers only with product A and what's their NRR versus companies with product B. And then I'll do, oh, companies with product A and product B, what is their NRR? But I don't do it on a product swap basis, like, if you know what I mean.

Yeah. So you're basically saying you're staying out of the allocation business, which is you're very happy to allocate expenses, right? But it seems like you don't want to allocate revenue across multiple products on a PO. Like if you have a pure product B customer and a pure product A customer, then we can analyze those independently. And I'm just poking fun at you there, but I'm more willing to do allocations here. I just need to remember that we did them. That's all, which is, you know, and the more product B is thrown in free to sell product A, the less meaningful this math is going to be, right? The more there are 2 different products that each independently exist or independently sold to the market, they each command their own value, the more meaningful this analysis is. So we need to understand where we are on that spectrum. But the short answer is on the overall financial metrics, it would be invisible because we're just looking at NRR of an account. If that account went from 100 units to 110 units, its NRR is 110. And if there's a lot of, you know, like what they call the duck where the feet are moving under the water, right? If there's a lot of switching back and forth between products underneath it, that will be invisible. And I think at an internal kind of management accounting quarterly business review level, I would definitely be tracking that. And I've run companies where we're deliberately, you know, product A is, let's just say, dying or running out of gas, and we're trying to move people to product B as a way to, you know, protect us, protect NRR. So bottom line is you can calculate it if you're doing allocations, understand that and be most wary of kind of line item discounting, right? Where I'll give you 10% off on the core product, but 90% off on the other product. Hopefully somebody will come along and reallocate that according to list price, but even that might be misleading. So let me go to you, Ray, for the final question we have today. Phil DaCorte, a friend of mine, worked together at— was it Host Analytics back in the day? He's the founder of ClearPath Ops and former VP of RevOps at RingCentral. Phil would be curious to understand our perspectives on the best comp strategies to optimize NRR.

Okay. Number one, I like giving the NRR goal to all the go-to-market leaders— marketing, sales, and customer success. But that's a lagging indicator. Then I like to identify what the leading indicators and input metrics are to get that NRR result. And that's where on a comp plan perspective, I want to really maximize incentive comp for CS for retention. So that GRR number, I will pay them handsomely for that. I would pay customer success for customer success qualified leads, identifying those upsell and cross-sell potential opportunities, and then handing that over to whether it's an account management team or the AE. The other thing that I think is really important is if your AEs are responsible For that upsell cross-sell motion, which is a critical input signal to NRR, you've got to make sure their quota has a unique line item quota for both net new ARR and expansion ARR. So you may compensate them a little bit less on expansion ARR. Let's say it's 6% or 8% versus 10 or 12 for new, but you've got to understand how you're motivating people to drive their responsibility, whether that's retention, identifying upsells, or closing upsells. That's my feedback.

Hey, Ray, my answer is similar to yours, but not identical. First, I like comp neutrality between customer success and sales on the theory that it lets work flow to the right place. So, if an expansion customer, if an existing customer places an expansion order, I'd like sales and CS to make the same amount of money as they always do. So, that means that the work will, in theory, flow between the two of them. They can look at each other and say, hey, you want to do this one? You want to do this one? Who's busier? Is it within your capabilities? And work kind of flows to the right place because they're comp-neutral. So, I want the two people on the front line deciding who should work on what. Second, I do think CS should be paid on cohort NRR, as Ray talked about. So, give me a little cohort of customers. They're worth 100. My job is to make them worth 110. I do think they should also be paid. On expansion opportunity identification. So, I always think of CSMs as SDRs for expansion opties. And then finally, I just want to make sure this is one thing that often goes wrong with CSM comp is you take our best CSMs and we put them on the hardest accounts and they produce a terrible NRR and they get punished, right? Because we've basically given them a dive that has a degree of difficulty of 3.5 and everybody else is doing like a regular dive, like a 1.5 or 2.0 degree of difficulty. And we've made it very hard for them to make their comp plan. So, there's There's a number of tricks to get around that, but the main thing is don't punish your best CSMs by having them take on the biggest accounts. You want them on the biggest accounts, so we need a way to reward them for that kind of hazardous duty.

Now, I can tell we grew up in different functions, right? I grew up in sales where I like meritocracy, right? Let me make, you know, the money on what I do myself. But you wanted to have a socialism plan where the CSMs and the AEs make the same amount of money.

Yeah, I wouldn't call it socialism at all. I would call it collaboration just because I don't think of it as socialist. I mean, I would call pods a socialist model where we have a shared quota amongst 5 reps, right? That would be socialism in my mind. This is just, I want the CS and salesperson to say, let's each work on what we should work on to be most efficient. And the last point was just about, let's not, that's more about perverse incentives. Let's not punish the people who are doing our best work.

Right. And I totally agree with that. So, you know, I, I never want to cross over into that political lines with you, Dave, but I thought, ah, that was kind of a good opportunity to poke you a little bit. But you came back and collaboration, not socialism. I love it. Well, we're going to have to make that the end of today's episode, Dave.

Hi, thanks for listening.

Okay, thanks everyone.

The Metrics Brothers AI, SaaS, and More is a production of The Metrics Brothers Growth and CAC and a member of the Benchmarkit 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 consequences consequential, inconsequential, or other damages arising out of any use of, or God help you, reliance 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.