Expansion ARR and NRR in a Variable Pricing Environment - Part 2

Live from Schenectady, New York, it's the Metrics Brothers, Growth and Cat.

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

And I'm Dave Kellogg, EIR at Balderton Capital, independent consultant, and the author of Kellogg.

And together We are the Metrics Brothers.

And we go together like revenue and invoices.

Oh, revenue and invoices. Aren't those 2 very different concepts, Kek?

Well, they're different things, certainly, Ray. And the reason we're going to talk about them today, and we're going to talk a lot about, uh, well, that today, is we're going to do a special one-time follow-up episode on the topic we did in our last episode. And the title for today is going to be Thoughts on Expansion Percentage and NRR in a Variable Pricing World. And I think we need to do this, Ray, because as you know, Ray, we didn't get through all the material we wanted to do yesterday. We both thought it was a fine episode, but there's a lot more to say on the subject.

Totally agree, because there's a lot of complexities and nuances here, and I think we have a very structured approach to go through today.

Okay, so let's do this. This is going to be a little bit more like a class than our normal conversation, although we are allowed to talk in class. Are you ready to begin, Professor Ray?

I am ready. 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.

All right. So the topic we're going to talk about here is what I call variable pricing, which is my capstone term to include pricing models where things aren't the same every month. Because in the world of SaaS, traditional SaaS that we grew up in, It was pretty simple. You sign a contract for 120 units a year. You, in effect, you'd get revenue of $10 a month. You'd probably prepay the contract upfront. So the invoice would be, to get to your thing, right? The invoice would be for 120 units, right, in the first month of the deal, but the rev-rec would come 10 units a month over the time period. Everything was pretty nice and pretty simple. Today, of course, we have variable pricing models. First side point, but usage-based pricing can be variable. I work with one company that, hey, how many users do you have last month? And the bill changes. Much more commonly usage-based pricing, i.e., consumption-based pricing. How many SMS messages got sent? How many tokens got processed? Consumption-based pricing or outcome-based pricing, which is coming with AI. How many support calls got handled? How many S1 opportunities got delivered? What problem does this create, Ray?

Well, first of all, being a SaaS metrics guy, you know, I think about, you your session at SaaSmetricspalooza last year, SaaSmetrics rests on Atlas's shoulders of ARR, annual recurring revenue, which was much easier in a subscription model. But Atlas didn't shrug, but his business model sure did change. It's less about annual fixed contracts today. We have new pricing models which introduced the concept of your base, which might be your platform fee, And then your variable usage or outcome-based. And now you even sometimes have credits and contract terms might be a little bit different. And all this means that the invoice and the revenue is not the same every month anymore.

Correct, Ray. And the impact here on SaaS metrics. So, so basically the, the, The revenue and the invoice, and they're different things, aren't the same every month. And that kind of breaks all the SaaS metrics because they kind of assumed they were. So, so we have all these formulas built around this thing of ARR. ARR is based upon a fixed annual fee amortized over a period of months. And now we don't have that.

But Dave, not only, not only do we not want to give it up for our metrics, Investors still want to talk in ARR. Correct.

So, we've got a problem because we have a whole world of SaaS metrics built around formulas. A lot of those formulas rely on ARR. ARR kind of doesn't exist anymore. So, what are we going to do? So, people are finding proxies for it. And I think that the most honest and the simplest conceptual way to think of this is ARR actually is dead, right? If you don't have annual contracts, and basically fixed fees across them, then you're not in an ARR world, but that's okay, right? ARR is dead, but it's okay because we can create proxies for ARR. And we should be aware that we're doing that. So investors have a proxy they call implied ARR, which is basically 4 times quarterly— I'm going to call it software revenue. Snowflake calls it product revenue, which I like. I want to call it subscription revenue, but I won't because that often connotes to some people, oh, that by subscription you mean base, you mean the base fee, right, Dave? And no, I didn't. I meant to say software revenue. So I'm not going to use the word subscription because it can— some people will take it as base only, some people will take it as all software revenue. So, but in any case, we have this metric that's used by public financial analysts called Implied ARR, which is 4 times software revenue. Some people create other proxies. They might do 12 times monthly revenue. I've heard people do 4 times average quarterly revenue. I don't know if you've heard of other ARR proxies you want to share here, Ray.

No, those are the primary ones. That's going to cover 95% of what I see, Dave.

Okay. So, and in the end, I refer to the way we're going to look at things in this episode today as spend is truth. Right, where spend to me is basically revenue, right?

Oh, hey Dave, I have to interrupt you because you know there's Ray Growth Reich has another thing I love to do. I love to call it my base revenue plus my variable revenue, and together that gives me my software revenue.

Correct. Base and variable is good. Yeah, those are the terms we're going to use today on the episode. I think they're descriptive terms and we're going to use them. So, But either way, back to spend is truth. I think it's a much better way of looking at this because what it means is how much you spent is truth. Like you spent this much money in Q4. And yes, the invoicing may have been timed funny. You may have paid some of the money upfront, but kind of imagine a world where all the payments are 30 days in arrears and the base fee is amortized across the months. In that world, your bill in that world, right, was your spend. That's how much you spent that month, Ray. This much was your monthly portion of your annual fee. This month was your usage. Here's your bill. That's what I'm going to call spend. And if you just take spend as truth, it's what you spent, not 3 times what you spent in the last month of the quarter or 4 times what you spent in the first quarter, right? It's just your quarterly spend is truth. I think it's the best proxy for ARR. Thoughts, Ray?

I totally agree. And I'm going to maybe jump ahead a little bit because when I think of ARR, number one, it was most often used as a forward-looking metric. How much am I going to get over the next 12 months? And I could usually determine that pretty easily from looking at the contract.

Correct. So let me build on that, Ray. In the past, if you wanted to know what ARR was, all you had to look at was the contracts. You never had to look at the revenue system. or the invoicing system, you could just say, grab me all the contracts and I could tell you what ARR was. That's not true anymore, right? The contracts will have kind of pricing model terms specific for every customer in them, but you have no idea. You can't make the bill if you don't know the usage, right? So today we're going to need to look at basically the sum of the bills, which in a perfect world would be equal to software revenue. Well, perfect being the world I described a minute ago where everything's amortized and everything's done 30 days in arrears. So not perfect, but in that idealized world, Revenue is the same as the sum of the invoices. And that is what I'm also going to call spend. And that's what we're going to use as kind of truth. And the question is, where do we get those conceptual invoices from? Like, how do I generate one? And the answer is you start with usage data, right? Then we apply not the pricing model, but your pricing model. Okay. What was the deal that Ray got? How many base units did he get? What's his price for an overage unit? Any other discounts? What does Ray have? So, we take Ray's usage, kind of run it into this black box, and out pops this invoice, right? This bill, your spend, whatever you want to call it. And that's what you need now. So, in the old days, you could just look at the contracts. In the new days, I need to look at your revenue or your bill to start figuring out what ARR is.

Yeah. The only thing I would add to that is that's why The entire product analytics and usage data infrastructure you build is so important in this model too.

Yeah. So, one of the things we should say while we're at this point. So, Ray also talked about ARR being forward-looking, which it was. You know, one VC I worked with said his favorite definition of ARR was, if they don't screw anything up, here's how much revenue you get over the next 12 months, right? If we sell nothing and you don't churn anybody, But you just kind of let the machine ride along. If I had $12 million in ARR, I'll have $12 million in revenue in the next 12 months. It was a forward-looking indicator. And going back to those first principles is important because we need to understand how ARR changed before we start shoving these proxies for ARR blindly into formulas. But I think Ray and I's favorite proxy is simply spend, i.e., trailing revenue for either a month or a quarter or a year, right? That's the thing we should use as a proxy for ARR. And if we do that, we think we'll get the best answers. So, I want to do a couple of examples though first, Ray, before we jump in. So, let's do 2 examples of what we're talking about here. I want to talk about what I call a humpback customer. And I'm going to pick these examples on purpose to make things simple. But their monthly spend Goes 100, 200, 100, 100, 200, 100, 100, 200, 100 forever, right? So, that means I'm getting $400K per quarter from this customer, right? 100 plus 200 plus 100. I get $400K per quarter. So, I'm getting— and because every quarter looks like every other quarter, I'm getting $1.6 million or units per year in revenue. Implied ARR, would be $1.6 million if I did it off a quarter, right? Because every quarter looks the same at $400K. So 4 times $400K is $1.6 million. Um, if however I did a different formula for implied ARR, this is a non-standard one, but if I took 12x the last month, I'd actually get $1.2 million, right? Because the month— the, the every quarter ends in a month of 100 units, right? So, which is correct? And the answer is 1.6. And this is back to what Ray was saying. ARR is supposed to be a leading indicator. So, if whatever ARR my proxy produces doesn't kind of closely approximate the revenue I'm going to get in the next 12 months, it's not a very good indicator. So—

Dave, I got to jump in here because you know how I have these really bad jokes sometimes. When I saw humpback customers, that's what you called this one. I'm like, oh, that's the type of deals I always like closing, the whales, the humpback whales.

Okay. Thank you, brother, Professor Growth, for that. That was awesome. Okay. So let's do another customer, Ray. If you like the humpbacks, maybe you'll like the sinusoids. I'm gonna do a different customer. I call them the sinusoidal customer and they go 100, 200, 100, 100, 0, 100. So it's like a sine wave. January, 100, 200 in February, 100 in March, 100 in April, 0 in May, and then 100 again in June. So these people—

Even if I looked at that quarterly, it's 400 in the first quarter and only 200 in the second quarter.

Correct. In fact, Ray, more generally, it's 400 in every odd-numbered quarter and 200 in every even-numbered quarter. Now, so, and basically it oscillates with a half-year frequency, right? Because that's a 6-month cycle. So it's quite stable across halves. Every half, this customer gives us $600K. Every year they give us $1.2 million, right? It's stable, but it's vibrating. And the reason we're doing this, guys, is I'm trying to pick examples where things move around because I'm trying to get our head around what life is like in a variable pricing world.

Yeah.

And I just find it really easy to pick these kind of cyclical customers, assume they run forever, that I can just drop anywhere and look around me. And what I'll see with this customer is basically, if I look at quarters, they're kind of bouncing around. They're at 200, 400, 200, 400. If I look at halves, they're stable, 600, 600, 600, 600. If I look at years, they're stable, 1.2 million, 1.2 million, 1.2 million. So, so NRR, depending on how I calculate it, If I do a 2-year lookback, which is the snowflake method, right, where I take what do they spend in the last 12 months and compare that to what they spent in the 12 months before that, it's going to be 1.6 divided by 1.6, which is 100%, which is the correct answer. If I use change in implied ARR, I also get the right answer for the wrong reason. So this is a funny example because in the final quarter of the year, it's a $200K quarter, right? It's a 100, 0, 100 quarter. in Q4. Um, so that's only $200 in revenue in that quarter. Implied ARR would therefore be $800, which is, which is not a good proxy for ARR. But I accidentally get the right NRR because I'm dividing in the prior year, it was also $800, right? So this is the kind of weird stuff that happens. I, I think the moral of this story is that longer periods damp volatility, and that's why I picked one with a 6-month Frequency as opposed to a 3-month frequency. So we could see, or really period, so I can see that get damped out.

And my recommendation to every company I work with is that a 12-month period for NRR is something you should always be calculating, not just a quarterly or biannually. Okay.

So, and let's go back to what NRR was supposed to tell us. And this goes back to a point you raised a minute ago. So once again, folks, what's the definition of ARR— sorry, NRR? It's supposed to be ARR today divided by ARR as of a year ago for the year-ago cohort. Okay. So example, you bought in year 1 for 100 units, you renewed in year 2 for 120 units, your NRR is 120%. Now, a couple of things go wrong in this formula commonly. Before we go variable, let's just talk about in the regular world what goes wrong. First, survivor bias. People do the math right. They go ARR now divided by ARR a year ago, but they pick the wrong cohort. They pick the people who are customers now, which conveniently excludes anyone who stopped using us over the year. That's called survivor bias. You shouldn't do it. Another mistake, in my opinion, this one might be more controversial, is using a bridge method where they, they say, well, let's calculate NRR by doing starting plus expansion minus churn divided by starting. And that can be wrong because it can— is one of— and these are all kind of corner cases. But first, I like— I don't like it because it loses the spirit of the metric. It's supposed to be a snapshot metric in my mind. But more importantly, you could include expansion or churn from people— well, just to get expansion from people who weren't in the starting cohort.

Right.

And technically, you could actually get churn from people if they didn't have annual contracts. I joined in month 3. I leave in month 6, or, or, or, yeah. So, so then I would not be in starting cohort and you'd get both my expansion and my churn. So, so, so that's not good. Or, or you could get the double expander as an easier example. I start in Q1, I buy 100 units, I expand in Q2 to 200 units, expand again to Q3, and then a new guy drops in, buys in Q2 and expands in Q3. His expansion should not be counted because he wasn't in the starting cohort. Mine should be, right? And I don't want to have to worry about all that shit, Ray, which is why I say people should just do the cohort method, right? Just get all the people who were customers a year ago. What did they spend 2 years ago, basically? And ARR now divided by ARR a year ago for the year-ago cohort. That's the best way to say it. And if you just do that formula instead of the bridge, life is much simpler and you get the right answer.

Dave, you know, You used one word about 6 times in the last 2 minutes, and that's expansion. So, for those of you listening, we're going to talk about what is expansion revenue or ARR in a few minutes. So, wait for that one.

Okay. So, before we go to expansion, Ray, let's talk about GRR. So, the answer here for NRR in a variable pricing world, I think Ray and I both agree, is do the Snowflake method.

Right.

You have 2 choices. You could do change in implied ARR. That's one choice, right? Take the most recent quarter, multiply it by 4, and divide it by the year ago ending quarter, right? The year ago quarter, multiply it by 4. So you've got 2 implied ARR figures and you divide one by the other. That loses all notion of a cohort, right? There could be new people entering. Right. There will be actually new people entering, right? It's not a great way to do it because you lose any, at least the bridge method. If people don't expand into a year, it's cohort E. But this is not even that. You're going to be including new people in there. The new ARR, right, will be in there, not just the expansion from new customers will be in there. So the right way to do this is to use the snowflake 2-year lookback method. which I won't say in English 'cause it's really hard to say, but you're basically comparing last year's to the year before that for the group of people who were customers a year before that. And that's the big disadvantage of the Snowflake method because it only includes people who were customers 2 years ago. So it's really a backward-looking metric. It's not gonna tell you, and by the way, if you wanna be looking forward, be looking at sequential expansion, right? This is another reason that in a, variable pricing world, you need to be focused on sequential because to get some of these metrics like NRR to work, you're now in— you're going back even further than NRR does for SaaS.

Yeah. So Dave, as we were talking about this, you mentioned this word that NRR was really a straddle metric.

Yeah.

Can you talk a little bit more about that?

Yeah, I think, I think we talked about it, but, but I always think of it as a straddle, which has got one foot in the past and one foot in the present. Sorry, one foot in the past, one foot in the future, right? So if it's January 1st and you've just renewed your contract that you signed January 1st last year, when I calculate NRR, I'm saying, hey, what did Ray sign up to give me in the coming year? That's future, right? He signed up to give me 120 units. Hey, what did Ray give me last year? He gave me 100 units. Therefore, Ray's NRR is 120 divided by 100, which is 120%, right? So that's why I think it's a straddle. I got one foot in the future and one foot in the past, whereas we just talked about with the Snowflake version, which we think is the right way to do it, you've got one foot in the past and one foot in the distant past, right? So, so you're really looking backward and that does affect interpretation. But let's now go talk about GRR, because if you think variable makes NRR hard, I think it makes GRR really hard. Ray, it looks like you want to say something before we do that.

No, not at all.

Okay. All right. So GRR, gross retention rate, what is it? It's supposed to be kind of like ARR, but you can't count expansion.

Right.

It only counts shrinkages. Ask most people, that's the definition. So more formally, it would be the ARR now divided by the ARR year ago for the year ago cohort, excluding expansion. Right. That would be the definition of GRR. And if you think about this like kind of a calculus, an integration problem, you're kind of going along this curve and you only get to integrate the area under the shrinkage.

Right.

You only get to integrate when it goes down. So if we took our sinusoidal customer, the guy who's going 100, 200, 100, 100, 0, 100, that guy, that guy drops twice in that cycle, right? They drop from 200 to 100 and they drop from 100 to 0. And I don't get to offset those because I don't get to count the increases. So I would count that as 400 units of shrinkage if I don't get to count expansion. And a lot of this will come down to the time period. But if within the time period, and the time period's a year, I— well, is that true for this one? Let's just put it this way. If I don't get to count expansion and I only count shrinkage, I get to say they shrunk by 400, which is clearly wrong because we'd have them starting out with kind of $1.2 million in annualized ARR. And then saying somehow they— well, they shrunk by 400 because we had 400 downsteps and they end up with 800, which is completely wrong because this customer is flat and gives us $1.2 million every year.

Right.

So, so GRR gets very hard and it's actually the time period that helps save you because within the time period you get to let extraction, shrinkage and expansion offset, right? If I'm counting within the time period, but when I'm going across periods, they don't. I'm not sure I said that right, Ray. What do you think that's hard about GRR?

Well, number one, there's no way in a sinusoidal kind of variable That you can even do GRR on a monthly or a quarterly basis if you have that type of high variability. You've got to have that longer period. And I think the only way you can look at it is the true spend of a 12-month period. So I think you can only do GRR using 12 months in this example.

Yeah, basically what I was describing with all the downsteps would be looking at it every month. So up by 100, down by 100. And yeah, so that would be the answer. And because basically the idea with GRR is, well, what is a shrinkage? And the answer is, because this stuff's volatile now, it's moving around, right? It's bouncing every day. I can't only count the downward movements, right? Because I'm just integrating all the negatives. So what we want to do is make the time periods longer so they have a chance to offset, right? And say, Hey, what did Ray do 2 years ago? In this case, he did 1.2 million units. What did he do last year? 1.2 million units. So, in this particular case, your GRR and your NRR would both be 100%. And that is the correct answer if we look at this curve. So, that's how Ray and I both get to this feeling that long periods are good because when something's bouncing around a lot, you want to take a longer look at it to basically— To avoid kind of— it's not oversampling, but to avoid looking so granularly, you get the wrong answer. So I think now as we start, we got 2 more topics to hit, right? As mentioned, this is going to be a longer than usual episode and we just kind of crossed our usual length. I think a lot of the stuff we talked about was interesting, but somewhat academic because in a variable pricing model, sequential change matters more than year over year. Once again, in a SaaS contract where you only have one chance to renew or change your contract once a year, it's great to be doing all those year-over-year metrics. Hey, what did Ray sign one year ago? What did he sign today? Right? In variable, it's how much did you use— Ray use yesterday? How much did he use the day before? Right? We're literally going to be looking at Ray every single day. So I think people are even going to forget that a lot of these metrics were defined as year-over-year. And we're just going to see more and more focus on sequential change because that's what matters in a variable pricing world.

In fact, we're going to be pivoting from looking at the contract and contract value to the product utilization trends. That's going to be so critical for anyone responsible for minimizing churn and maximizing expansion.

Correct. And I do think the first place— let's just talk about that for a minute, Ray— that basically in the variable pricing world, you've got usage data. You run that through a black box called Ray's pricing model, out pops Ray's conceptual bill. And I think you need to look at both sides of that equation. Like, if you're in customer success doing churn prevention and churn detection, I'm gonna just look at the raw usage data, right? I think that's the best way. I don't care how good we were at monetizing your usage, right? That's kind of the translation between usage data and bill, right? So, how good are we at monetizing usage? Is comparing the usage data to the bill. Is this customer potentially in trouble because they're not using the system? I don't need to know the billing transformation to do that, right? I just need to look at your usage. So, and obviously, these companies now all have data science teams that do that. They forecast usage to the extent that the usage forecast becomes input, right, to a black box that pops out a revenue forecast. That's important too. Right? So, this is not just an FP&A exercise anymore, right? This is now a data science team and finance exercise to do this work. But one of the things that also comes up here, Ray, because another popular investor metric is expansion ARR. And let's define it. What percent of your ARR this quarter came from existing— of your new ARR this quarter came from new Versus existing customers. That's what expansion ARR is. It is different from sequential net expansion, right? Sequential net expansion is I'm just looking at, for any given customer, did Ray go up or down in this time period? Ray spent 110 units in Q3, spent 100 units in Q2. So, his sequential net expansion was 10 units. That's pretty easy. I can do that for the whole base as well. Hey, the base spent 100 units combined. The base spent $110 million. The base went up. But expansion ARR is kind of a different viewpoint. It's saying, no, no, no. Of all the ARR, or I should say revenue, right? Because there is no ARR, right? Of all the software revenue this quarter, how much of it came from new versus existing customers? Which of course begs the question, what's a new customer versus what's an existing customer?

Right.

Thoughts, Ray?

Well, I cannot tell you how important the point you just made is because when people say, hey, my, of my total new ARR, 40% was from expansion. That is not, not the net expansion rate. So how much does somebody grow every month? It's that entire cohort of both customers, old and new customers. And what contributed more to your growth ARR. So important.

So, did you answer the question though, Ray? How do you distinguish between a new and existing customer?

I did not.

So, would you like another swing at it, Ray? Interesting stuff. Are you— have you become like a politician in Washington, D.C., where you never answer the question?

Yeah. Well, I have been listening to a lot, way too much political news. But anyhow, so, I, as a former sales leader who had a contracted commitment plus variable usage, I always used a year. Whatever that new logo customer did for the first year, I put that into new revenue or new ARR. So that's—

I'm going to call that new logo ARR just for Kelloggian taxonomy. Okay. So anything you spend in the first year, In your mind is new logo ARR in that year and that year only.

That's my traditional model. Now, what I'm seeing with a lot of variable pricing model companies is they're establishing this ramp period. Some people may call it how long does it take me to onboard a customer to get to some normalized usage of the product and thus the revenue being generated from that. And let's say that's 6 months. I'm seeing companies saying, okay, that first 6 months I'm putting into new, but everything after that, month 7 and after, I'm calling expansion revenue. That's because they've said, I've got a standard ramp and then a standard expansion motion period.

Got it. So, they're going to define the definition of new with a time period. And the time period is determined by how long it takes them to kind of get to steady state.

Usage, correct? That's correct.

So, if it takes me 6 months to get to, I'm going to say steady state in double quotes, right? Where it's not exactly flat, but it's kind of bouncing around. That makes sense to me. I just want to talk for a minute. So, what that means is if you're calculating expansion percentage, that person shows up as new logo ARR in Q1 and Q2. And then thereafter, they will show up as expansion ARR. Right? Because expansion ARR is all about who bought it, right? I define expansion— basically, new ARR to me has 2 flavors. You say, who's it from? Is it from an existing customer? Is it from a new customer? Okay, the new logo versus expansion. And then basically, what's the— in the old days, it was easier. What's an existing customer? Well, somebody who already signed a contract. What's a new customer? Somebody who hadn't. And now, because we have this ramping behavior Ray talks about, we kind of have this notion of kind of steady-state customer. Which is I want to call them new until they hit steady state. And then after that, once they've hit their steady state, which takes around 6 months, we'll count expansion on top of that. Interesting way to do it. I do think it also affects your CAC ratio. And the only thing I'm going to raise here, Ray, is that you need to have consistency because sometimes there's a tendency to, like, when you're— Like, I'll tell you about a bad tendency for the past. A company does a 3-year deal. The ARR is 100 in year 1, 200 in year 2, and 300 in year 3. Right? Just say they spent 200 units to get that deal. Well, what's their CAC ratio? Some people would go, well, there's 200 units of sales and marketing to get the deal divided by 100 units of year 1 ARR. Therefore, the CAC ratio was 2. Some people would go, well, there was 200 units of ARR, but we spent it really to get the average ratio of the contract across the time period. And it's 200 units per year if you take that contract. So the CAC ratio was 1. Some people, the most aggressive types, would say, well, the terminal value of that deal is 300 units. So we're going to take the 200 units of sales and marketing spend divided by 300 units of ARR and get 0.6. But where— I see your head nodding. Where did you land on this issue? What's your take?

I think it needs to be the first year revenue. And that's why one of the other things I've seen some larger SaaS companies do, they use this 1-year exclusion model to say, hey, everything in the first year is new. And all my CAC efficiency metrics use the first year revenue, just the first year.

Got it. So, the question here is going to be, so I think I generally like your answer. In that example, I could also argue, For the average. The one to me that's unconscionable is the year 3, but with a straight face, I could argue we should take the average value. Normally by default, I would just take the year 1 value and just say, hey, our CAC is high. And by the way, that affects— I mean, we're back to the same issue, which is that I guess you can't count it as expansion ARR as it expands. It creates a lot of problems. So I think— I want to come back to variable though to start wrapping this up. I'm trying to figure out the CAC on this because you're saying it takes 6 months to get to your steady state. In theory, the ARR you bought, right, the annualized revenue you bought is kind of— and let's just say it happens by month and you hit it in 6 months. And this is going to be horribly controversial, but if you're saying it's basically 6 times month 6, right, that if you want to know what steady state is, you just told me it takes 6 months to get there. So this is a case where I'm not gonna take a long time period 'cause I say, voilà, at 6 months they've hit steady state. So I, it's actually 12 times month 6. The annual value I bought was 12 times their month 6th. And then I'm gonna compare that to what I paid for it, right? 'Cause CAC ratio is how much did I spend, what did I get? So I don't know if you have thoughts on that. It's a little crazy, I just made it up.

Yeah, that's why I personally recommend and I see in these environments, That the first 12 months is new ARR and my CAC efficiency is based upon those first 12 months of revenue divided by the sales and marketing.

But how do you calculate it in the first year? You're basically saying I have to wait a year to know last quarter's CAC?

Exactly. Now, now I'm going to take you down a rabbit hole because what I see a lot of companies saying is for that size customer, I have a predictable, they're going to be doing $200,000 in total revenue in the first year, and I do my estimated CAC ratio, but then I go and I reconcile it at the end of the year with my actuals.

Yeah, because I think, look, for historical financial, you can clearly do the CAC based on year 1. You could do it. The problem is it's too delayed. I like the notion of an estimated CAC. That's effectively what I just did. I just estimated it using your ramping model to say, you know, hey, if we're pretty sure at 6 months that's a good predictor for their for their annual spend, then why not just take the sales and marketing expense divided by the 6 times— sorry, 12 times that month. So, okay, Ray, we've covered all the material I wanted to. Do you have anything to say? I'm going to try and summarize this after, but do you have anything you want to say?

No, I had something I thought was funny, but it's going to complicate things. So please do summarize.

All right. So here's the summary. In today's kind of one-time special drill-down episode, We tried to take a more structured approach. Hopefully we succeeded to talking about expansion percentage and NRR in a variable pricing world. We started by observing that variable pricing is effectively the big change. And variable pricing is any pricing environment where the bill, the bill, the conceptual bill we talked about is not the same every month. If the bill starts varying, then ARR starts to lose meaning. And SaaS metrics are built on ARR. So what do people do? They try to find some proxy for ARR to plug into the metrics. We talked about several of those proxies today. Public companies use implied ARR. We both showed where that kind of works and we showed where that didn't work. And it also is a function of how you define it. Again, most people do 4 times quarterly software revenue. And it can work okay. It failed in some of our NRR examples. The philosophy I like is spend is truth. When you're trying to build a proxy for ARR, just tell me how much someone spent and maybe I'll annualize that. Maybe I'll quartize that. But I like starting with spend. We observed that in the past, to know what ARR was, you used to call legal and ask to see the contracts. And now, you got to call rev-rec, right? You got to call accounting and say, let's show me— basically show me the conceptual invoices. Show me the bills. For this customer. I defined this notion of a conceptual invoice, which was basically if we lived in a world where we invoiced everything 30 days in arrears or at the end of the month, and if we included the base fee amortized in those monthly invoices, then that's this kind of conceptual invoice. And then I could just get the sum of those across months and I'd know your spend, right? So that's when we say spend or we say invoice, we say product revenue. Those things are all supposed to be the same. The reason I steer away from revenue is there are other things in RevRec that can change it that I'm trying to exclude. So a lot of different ways to look at it. You could say RevRec excluding weird stuff. Then you could just say it's software revenue. But in any case, how do you get these invoices? How do you get these bills? How do you get RevRec? Is you start with usage data, you run it through this black box called the pricing model. or the customer-specific pricing model, because it is these days. What's Ray paying per unit? How many units did Ray include in this base fee, right? So, you take the usage data, you run it through this black box, you get a bill. And I really like that. It is a simple conceptual model. And both the input and the output are super useful, right? For churn prediction, I think you want to be looking at usage, right? We don't care how usage is monetized. We care a lot— well, in churn prediction, we don't care how usage is monetized. But as soon as we talk about running the business, we sure as heck do care about how usage is monetized. And that's the purpose of the pricing model, right? To monetize usage and turn usage into bills. We talked about 2 odd customers that I think are useful for thinking about this stuff, which is the humpback customer who goes 100, 200, 100, 100, 200, 100 every quarter. And we did some SaaS metrics for the humpback customer. Then we did the sinusoidal or oscillatory— the sinusoidal customer who went 100, 200, 100, 100, 0, 100. And looked like a sine wave. And we talked about how these metrics would be calculated for them. We then kind of switched off NRR because during that section, we were pretty much focused on NRR. We reminded ourselves what NRR was supposed to be and how NRR used to have a foot in the future and a foot in the past, and that's now gone. We talked about how GRR can get weird over short time periods because you're aggregating every negative fluctuation and you get basically the wrong answer. So, you need to take longer time periods when looking at GRR. Then, we pointed out, I think the biggest thing, which is a lot of these metrics are year-over-year metrics. It's fun for Ray and I to talk about this stuff. But if we were running a company again, we would be focused on sequential change, month over month. What did this person— maybe even week over week to analyze customers. We then dodged back to SaaS metrics and talked about expansion ARR. We reminded people of 2 things. One, you probably care, well, One, you definitely care about sequential net expansion, which is a different metric. You're just looking at what happens to a customer or the base and how they contract or shrink over time. Expansion ARR is more of a time period view where you're saying of all the new ARR in a time period, I should say the new software revenue in any time period, how much of it came from new versus existing customers? So, the question is, what is expansion ARR? First, there isn't ARR anymore. And then second, expansion is going to say, well, how much do we get from new versus existing customers? Which in turn begs the question, what's an existing customer? Ray made the argument. Ray made 2 arguments. He felt you could define existing as everything they do in the first 12 months. Or his second choice, which I like better, was you could also use a ramping model and say, hey, if your average customer takes 6 months to ramp, why don't we call it new for the first 6 months after that? And then once they hit steady state at month 6, we'll then consider an expansion after that. I like that one a lot. Ray, did I miss anything in that summary?

You did not. The one thing I would say is spend is truth. Product utilization is imperative because that's going to help you with predicting both churn and identifying where you have the most expansion opportunity.

And just because Ray's a fancy guy from Schenectady, he says utilization instead of usage. But yeah, we got to look at usage data. I absolutely agree.

I don't get to use 5-syllable words much.

I know, I know. It makes your day when you get to. Okay, Ray, and somehow in this 42-minute episode, have we managed to miss anything you wanted to say?

No, I think this was really good. And I actually want to go back and maybe even think about how do I actually document this and put it in some type of writing.

Okay, will you go ahead and do that? Thanks, Ray. We'll see you next time.

Okay, bye everyone.

Take care. 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, 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 at Ray Reich. Dave Kellogg is based in Silicon Valley and available on X at Kellogg. Skedectity, which is French for unspellable, is not our actual production location. Thanks for listening.