Usage-Based Pricing and CAC - The calculation impact is significant

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

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

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

And together we are the Metrics Brothers.

We are indeed, Ray. How are we doing content-wise? Last week, what did we talk about? Was it PLG?

Yeah, we talked about product-led growth and how it impacts customer acquisition costs, which I thought was a really fascinating conversation.

And I think at some point we got Axel wrapped on PLG versus usage-based pricing, right? As I recall, I can't remember if that was the live episode or the draft where you kind of busted me for conflating them.

Yes. Yep. And usage-based pricing is so unique and different, and it's not always combined with PLG. So I thought this week, let's talk about that topic, how usage-based pricing impacts customer acquisition costs and its efficiency derivatives.

I think that's great. I want to, I want to look at the other side of that coin too and look at CAC and churn and expansion because, because you're right, it affects not only CAC, but depending on how you handle usage-based pricing, And I would say normal fluctuations in usage, it's going to change how your expansion and churn metrics look as well.

Hey, we're CAC and growth, not CAC and churn.

I got to be churn today because I have some experience on this one where there's a double counting issue that we'll hit later in the episode.

Well, that sounds like a great discussion. And, you know, we've only got 20 minutes in these podcasts. So let's take a break and get a quick word from our presenting sponsor, Maxio. Right before we get into it, Dave, 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.

So, Ray, before we dive into the CAC side of this equation and user-based pricing, let's first talk about user-based pricing itself. There are many different models out there, not just one. If you have to group them together, how many models do you see?

Well, you can get into multiple derivatives, but I categorize it in 3, Dave. Number one is pure usage or per— pure consumption base. That means I have no minimum, I pay 10 cents per unit, and at the end of the month they add up all my units, multiply it by 10 cents. That's per usage. But then we have the minimum commitment with pure usage. So that says you're going to use a million of these units over the year at 10 cents per unit. So your minimum commitment is $100,000. So you have a minimum commitment and then anything above and beyond that, you have a per unit price. So that's the second. And then the third, I call the hybrid model. It's where you're paying an annual subscription. Maybe that's for access to the platform. Maybe that's some other reason it's a subscription. And then above that, you have usage-based variable pricing on top of that. So you're paying $50,000 for access to my platform and you're paying me 10 cents a transaction above that. So that's the 3 I look at, Dave.

Okay. One question. What do you think the norm is on the second model when you don't hit your minimum? Do we do rollover minutes or not? That's always the issue with the minimum pricing model to me. I sign up for $100,000 minimum. I spend— look, if I spend $120,000, it's easy. I get 20 units of overage at a higher rate. Yay. Everyone loves that except the customer. But what if I'm under? What's the norm in that case?

Yeah, it really depends on the pricing model. I've seen 2 primary structures. One is it's an annual commitment. You've got to do a million units, right? And at the end of the year, if you use 900,000, it's gone. Now, the other model that I actually prefer to eliminate some of the friction upfront, it's 83,000 units per month, and you can roll that over to the next month. And then at the end of the year, it's if you don't get to that million at $83,333 per month, you still lose it, but you don't lose it on a monthly basis.

Interesting. Interesting. How does this translate to ARR? Let me just cut straight to the chase. Would— because most of SaaS metrics are built on some notion of ARR, right? Like CAC is sales and marketing divided by ARR, or churn is current ARR minus last year's ARR. So I'm gonna argue that the concept of ARR is central to SaaS metrics. How do you map— I'm sorry for getting ahead of ourselves, but this is the big question for me. How do you map these models to ARR or how do you just deal with that issue?

Yeah, it depends on the model. So if it's 100% usage-based, I don't turn it into ARR. To me, I look at that as revenue and I don't even define it as GAAP revenue or recurring revenue, but I'll look at my last 12 months and say, how much revenue did that customer or that cohort of customers generate. Now, if it has a commitment, right? And then you have overage, I use that commitment as my contracted ARR, Dave. So if it's $500,000, that's $500,000 of contracted ARR. And then on the usage above that, I don't turn it into ARR. I just call that variable revenue.

Got it. I think the way I see it, Ray, look, I think you're right that there are 3 different models. Let me just give my answer to that question. I tend to look only at trailing spend because to me it's the only truth to be found. Because while the pricing model has different ways of arriving at some fee for what you pay on the month or some fee and what you pay across a year, at least from the investor point of view, and I think this is maybe the heart of my argument, I think about these questions differently as an investor than I do as an operator. But as an investor, truth is, how much did you spend last year? How much did you spend this year? And the whole purpose of those pricing models is to try to optimize those numbers. As an operator, as a CSM, I want to know what pricing model you're on. I want to know when I can upgrade your baseline fee. I want to make sure you don't get mad at me for too many overage fees, right? I care about all that operationally. But as an investor, I feel like I just want to know, hey, what's your NRR, right? Like if you had a cohort of customers worth $100 in year 1, how much are they worth in year 2? And I can rise above this segmentation of subscription versus usage or baseline plus overage. I can kind of sidestep all that. So, so I don't know how you feel about that, but that's the way I try to look at this. I'm always trying to find a proxy for ARR and I'm always trying to keep it simple because I feel like it's quicksand once you go in.

Well, honestly, I don't believe that usage-based pricing environments are best for traditional SaaS metrics. I really don't. Things like net revenue retention. I think it's way too hard to determine in a 100% usage-based pricing model. What you can do is say, okay, in 2024 I had $90 million coming from my previous customers from last year and I had $10 million of new. Now of that $90 million, I can tell you $70 million was what they did last year and $20 million with the growth rate. So I use growth rate on the previous annual revenue, Dave. That's what I do.

Yeah, I mean, that's a lot like trailing spend though. So I'm not sure. I mean, look, we may have a rare Metrics Brothers spat here, but I think you're doing what I'm doing using different words. I'm not sure. Look, the first time I saw this is when I read Snowflake's K-1s and S-1, and they do, as you know, what I call a 2-year lookback in ARR. where they just say, hey, how much did you spend last year? How much did you spend the year before that? Let's divide the first one by the second one. And that's our NRR number. So it's a 2-year lookback. It's based all on actual spend. Like, I don't need to know the pricing model, right? I just need to know the output of the pricing model, which is how much do we bill them? So I do think, and by the way, the world does calculate NRR for Snowflake, right? I mean, in fact, they're the poster child for the world's best NRR, right? So to say they don't have an NRR is going to be a tough position. I don't know. What do you think?

I don't think we're going to get to 100% agreement here because once again, for me, it depends on the model. So if it's a hybrid model where you have subscription, right, plus a usage base, I do 2 things. I do 2 CAC and CAC derivative equations. One is, okay, what's my minimum subscription? What's my cost to get that? And then I take my sales and marketing expenses to get that. I divide it by my committed revenue, my subscription. That's my subscription CAC and CAC payback or CAC ratio. And then I use subscription plus variable revenue, Dave. And then I do my CAC payback on the combination. I call that my total revenue CAC payback or CAC ratio. That's the way I do it. That way I have 2, the minimum And the minimum plus usage.

Yeah. And the issue here, what's happening, Ray, is you're focused on CAC because my system breaks down totally for CAC now that I think about it. Like, I don't know how to calculate a CAC ratio given what I just said. It works really well for NRR. And I guess the place that I've been sucked into this stuff is on NRR calculations and GRR calculations. My method works fine. I've also been sucked into it on the definition of churn, right? Where to me, we have what I call the sinusoidal customer. And let's just say every month they go 10, 9, 8, 9, 10, 11, 12, 11, 10, 9, 8, 9, 10, 11, 12, right? So they're just sinusoidally oscillating between 8 and 12. If you're not careful and you just kind of integrate the area under that curve and you say that every increase is expansion and every decrease is churn, it also, it all offsets to zero in the end, but it adds a lot, a big number, a lot of noise in effect to both your expansion and churn numbers. And that's what I'm trying to avoid. And trailing spend, by the way, eliminates that. I just say, how much did you spend last year? I didn't care if it was shaped as a sine wave. How much did you spend this year? I don't care if it was shaped as a sine wave. The answer is it didn't change. But I would concede that mine, I don't really have a good answer on CAC.

But I don't think we have a spat here. I think we have a definitional issue, Dave, because we agree. I use something called a 1-year lookback period for NRR or GRR and usage base. So I look at the revenue from 2023 And I compare that to 2022, i.e., 12 months trailing spend, right? I do it exactly the same way, even in usage-based. And what I don't get hung up on is, am I using recurring revenue or GAAP revenue? However you define that in your company, right? Year over year, that's what I use for GRR and NRR.

So how do you get the coming 12 months? Are you just gonna use the subscription commitment? Like, like if I look in the, the regular ARR world, if last year you gave me 100 units and on January 1st you sign a contract to renew for 110, I haven't received 110 yet, but I've got a piece of paper that says you're gonna pay me 110. So I would take 100 and you've committed in 2024 to do 110 and in 2023 you gave me 100. So therefore you're 110% NRR. If you're only looking back one year, that implies you have to look forward, no? Or are you actually looking back 2 years? Like, I couldn't get the read on that, right? Because that's the difference. So the usage-based model, if there is a commitment, then I can see what you're using. But if there's no commitment, if it's pure usage-based, I don't know how to get the coming year. In fact, there's no way to get the coming year.

There's no way. That's where forecasting becomes the topic. So I have another metric. I can't believe I'm doing this here, but I have something called annual projected revenue, APR. I wrote a very long LinkedIn article on this about a year and a half, 2 years ago, because where I grew up, I actually had subscription plus usage-based pricing 30 years ago. And we, as both salespeople and finance, we had to get really good at projecting for the next year how much revenue every customer is going to do, including their minimum commitment plus overage. And we did APR, annual projected revenue. So we got into forecasting models.

And that makes sense to me, Ray, because basically you, the way you, here's the answer to my question. You forecasted it. You didn't get the forward-looking year by looking at a contract that says you're going to give me $110,000. You built a sophisticated model presumably and forecast it. I actually think that's fine. But to me, the question is when you, and by the way, your method is at least comparable because one of my beefs is if you're comparing a real SaaS company to a usage-based pricing, In the SaaS company, you're doing 1 year forward and 1 year back, whereas the usage-based using the Snowflake model, you're doing 1 year back versus 2 years back, right? You're not— it's not the same metric. You're looking at different time periods. So if I wanted to be comparable, I'd rather use your formula. And by the way, as a company gets bigger and has history, I think you're able to forecast that pretty accurately. And by the way, to defend my point of view, I don't actually care how you made the forecast. I just care you get the right revenue number, right? And you're going to say, hey, the pricing model has a huge impact on how I make the forecast. I'm going to say you're right.

And we did 2 other things, and this is really kind of 301-type stuff. We looked at macroeconomic trends and projections on an industry basis because we had a lot of customers in different industries. So we would have a different prediction for our automotive industry customers versus our healthcare customers because they had different macroeconomic variables, right? And second, we looked at growth of the unique company over the last 12 and 24 months. Because if we saw that they were growing and that impacted our transactional volumes, we would build that into the next year too. So it got very sophisticated, Dave.

Yeah, that's awesome. And by the way, I think it's actually really good to go back and look at history because, you know, everything old is new again. Uh, the fact of the matter is, in timesharing or other such models, we, we had to deal with these issues. So I think it's cool to go back and look at that. Ray, what else did you want to hit here today? Because I feel like I drove you off into the weeds there. What's next on our topics?

Yeah, you know, the key to me is though, because I love— 2 of my favorite metrics, and I know one of them is yours, is the CAC ratio. So you're looking at how much sales and marketing spend, right, compared to how much new revenue. And then there's the CAC payback period. So one of the things I really want to talk about is, can we really do CAC ratio and CAC payback period As a meaningful metric for either investors or operators in a primarily usage-based pricing model?

Yeah. And the question, look, to translate that question in the way I think about it, I paid a dollar, what did I get? Right? Because that's the essence of what the CAC ratio is. Hey, how much did you, well, actually, how much did you pay to get a dollar of ARR? Or in this case, I'm saying, hey, if you pay a dollar, what do you get for it? Same question. And the issue is in a purely Usage-based model, you don't know, right? With the minimum commitment, you do, right? So you could use the minimum, right? Or you could use a prediction. I tend to think the right answer, and I'm going to kind of maintain my investor viewpoint here because I think as an operator, I'm going to answer these questions very differently. But as an investor, everywhere I want to put ARR when I don't have it, I can use implied ARR. And implied ARR is just last quarter subscription revenue times 4, right? And that's what they do with public companies. You go to Maritech public comps, you see implied— Snowflake will not use the term ARR in any of their documents, but you go to Maritech, tech, you'll see they have an implied ARR figure, right? And they have a pure usage-based model, not pure, but they have a heavily usage-based model. So I think that's my opening volley is at least as an investor where I'm trying to understand the operations and finances of a business, I can use implied ARR. And that would be my opening volley because the alternative is you have to somehow guess what I just bought. So I just signed up Acme, I spent 2 units of, you know, sales and marketing to get Acme. What did I buy? And I don't know, how do you answer that question? What, what you're gonna say, it depends on the model, I'm pretty sure. But if I just signed up Acme and it cost me 2 units, how do you answer the question? What did you get for your money?

So this is where I look at 12 and 24 months history of ramping, right? So I look at it based upon whether it's what industry or what segment Right? They're enterprises, SMB.

Yeah.

I look at what's my average ramp process? Is it 3 months of ramp? Is it 6 months? And for that cohort on an annualized basis, how much revenue is that once it's ramped? And I will use that annualized projection of a fully ramped customer in that cohort. And that becomes my ARR proxy that I divide into sales and marketing expenses.

How much time do you give them? Are you gonna say—

Depends. If you have an average 3-month ramp, I only give them 3 months. But if I've got a business model where it's a 12-month ramp, then I have to look at it over 12 months. I look at month 12, I look at the MRR and I multiply it by 12.

So there's 2 things you did there, Ray, and I think they're both important. One was you kind of refused to answer my question for Acme, which is you said, I don't do it for Acme. I do it for the cohort that Acme's in. Which I think is a, is a, makes it much safer in my mind, right? Because now you're doing across a group of customers, but you're basically predicting based on some onboarding ramp how long they're gonna be. So I think you, you put part 1 was I'm not gonna do it for one customer. I'm gonna do it for the cohort. And then part 2 is I'm gonna make a prediction of what they're worth based on history.

Yes.

So yeah, I mean, look, great. I think ultimately this is all pointing to the fact But I know you agree with me on this one that we're kind of shoehorning SaaS metrics into usage-based models, right? They weren't built for usage-based models. In my opinion, they don't work particularly well for usage-based models, right? You get axle wrapped like we're doing and we'll continue to do if we talk about this. You get wrapped around the axle very quickly because the metrics weren't built for this world and we're really trying to shoehorn this world into metrics built for a different world. And that's why you're making a prediction of what the ARR is gonna be. In original SaaS, there was no prediction. I had a piece of paper that said you're gonna pay me 100 units over the next year, right? I know exactly what I got for my money. And you're saying now it has to be probabilistic.

Exactly. And that's why, and I just saw some research and I know we talked about this right before we recorded today's episode, that when you look at primarily usage-based pricing, if you look at the growth rate of those B2B SaaS companies over the last 6 and 12 months, it's significantly lower than your pure subscription base. Why? Because people could throttle down on usage of a product very quickly. But if they had a subscription, it was still 9 months longer before they could cancel it. They couldn't throttle back. So there was a lot of risk with this usage-based pricing model also, David. I took you another direction now.

Yeah, look, I think as we went into this, I said we're going to find out there's 2 types of CROs. purists and pragmatists. And purists would go with a pure usage-based model and get hit the second they missed. I'll tell you the quote. This goes back to back in the day when I was officer of a public company. When a bad, you know, when the industry, when the economy hit a bad quarter, we'd say there's 2 types of companies, those that miss guidance and those that make guidance but reduce their forward guidance, right? Like it's going to hit everybody. But do you have enough cushion so it doesn't hit you on this quarter? But you push it out to next, or are you not carrying cushion and it hits you instantly? This is the same question, basically, which is there's 2 types of companies. The purists get hit now. They get it instantly because they have no ratchet. They have no safety net built in. Whereas a pragmatist CRO, and I greatly prefer to work with pragmatists, will have said, yeah, we love usage-based pricing, but go get minimums and be very careful about the rollover policy because that way we're going to buffer the impact. I mean, the ultimate buffer on the impact is not to be usage-based at all, but then you don't get the upside. So I do I do think, look, my frustration here is that, you know, make hay when the sun shines. Everybody thinks usage-based pricing is the best thing in the world when the economy's growing and money is free and up, up, up, up, up. And as soon as we hit a downturn, we're gonna see if people retreat from it. But basically, because I'm conservative, even when the sun shines, do usage-based models, but put in minimums. And I think after this downturn, everyone's gonna be a pragmatist. Yeah.

Totally agree. If you've got a year of history, that's the time to say, hey, let's give you an annual commitment, but you're going to save X percent per unit. Then it's like, oh, okay, then maybe I'll go to it. Because on the flip side, finance CFOs, what they hate about usage-based pricing is someone projected they were going to spend $200,000 as the buyer and they spent $350,000. That doesn't last very long either, Dave. I think we got to wrap up today.

Yeah, I agree. By the way, one of my— before we wrap, Ray, one of my soundbites was, you know, everybody loves Snowflake's 180% NRR except the customers. Right. And maybe that's a little harsh because if they're really expanding usage and value proportionately, maybe, or even faster, they might be happy to expand their average contract at 1.8x per year. But what excites Wall Street isn't always what excites customers. And what excites investors isn't what excites customers. And look, on the receiving end of that, Like just, I'll tell you about my cable bill at my house in Oregon, right? It goes up with an incredible NRR and the net result is I want to find a new cable provider. So there's always two sides to all this stuff.

And one of the things everyone says, if you want to think about what's going to happen in the future, go study history, right? Whether that's politics, monetary policy, et cetera. So 30 years ago when I was doing this usage-based pricing model, I was so excited. I had one customer that went from $300,000 to $1.8 million. The next year. And they came back to me and said, Ray, we gotta try, we gotta go a different direction because I can't afford this variable pricing model that you gave me. So our highest churn was from our highest growth customers from the previous 2 to 3 years. That's the risk.

Newton's second law, for every action, equal and opposite reaction. Maybe not always equal and opposite.

Well, CAC, I'm glad we talked about this today. And, you know, next week I think I'm gonna take us into another one. How about if we talk about customer lifetime value next week? What do you think of that?

That's one of my favorite topics. So yeah, look forward to doing it.

Okay. Talk to everyone later. Thanks, Dave.

See you. 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. Schenectady, which is French for unspellable, is not our actual production location. Thanks for listening. Right.