Very True by Verissimo

Alex Oppenheimer (00:02.656)
Okay, so today I'm gonna talk about ARR. There's been a lot of talk about ARR lately over the last 15 years, but especially over the last six months as it's come into focus and question again with a lot of AI companies. And the question is, why has it come into focus? What is it? Why does it matter? Why won't everyone shut up about it? And I have been, I'll take it back to the first time I ever heard of this concept.

It was actually when I was at Morgan Stanley in 2011 and I was assigned to the coverage team for the company Synopsys, which is now a famous, very large company in the EDA chip design software space. And the VP, very, very smart guy, he's gone on to be very, very successful, was explaining that they were restating their revenue and now it was being recognized ratably. And I was like, okay.

And then, you know, from there, I, when I was at NEA, we had companies at that period that were doing the same thing that had gone from on-prem and or, you know, kind of appliance styles, old school software pricing, and we're switching over to the SaaS model. And I have been deep, deep, deep in that since then. Never stopped. I am a finance nerd. I'm a math nerd. I'm an engineering nerd. And I've just been focused on this term and this idea for a very, very long time now.

And I had this realization after my old boss, Harry at NEA told me to, he said, this SaaS thing is taking over. This was early 2014. He said, the SaaS thing is taking over. Everything's going to be a subscription business. And I want you to be the expert because you're really, you know, finance and you're an engineer and I want you to just take everything and learn it. And, and that's what I started doing back in early 2014. If I'm being honest, it took me about a year to figure out what the crux of it is.

A lot of it has to do with my mechanical engineering view of the world is that everything's integrals and derivatives. And my biggest realization was that ARR is the derivative of recognized revenue. Now there's a lot of terms in there and if you don't know what all of them mean, then what I said means absolutely nothing. But for me, it's a very, very helpful idea. The key is that it is a function that is a leading indicator.

Alex Oppenheimer (02:23.52)
of what's to come of something else, which is recognized revenue. So I figure I'm gonna start, I did a quick lookup on the definition of revenue. To take it back a little bit further, this concept of revenue, which if you don't understand revenue, you can't possibly understand what ARR is. And by the way, I haven't said what that acronym stands for completely yet, but I assure you one of those Rs is revenue. Then it's hard to figure out what ARR is, again, if you don't know what revenue is.

Revenue is an invention. you think back a thousand years ago, there was no such concept as revenue. it's an invented concept to bring economic reality closer to what the books represent for a business. Business used to all be done in cash and that was it. That was the only thing that mattered was cash. And so what I say is, you know, some of the biggest innovations in accounting have actually been the invention of this concept of revenue. And then the invention of this concept of depreciation, which is

actually pretty much the equivalent thing on the cost side. So I am just going to share what a Google Geminis AI overview definition of GAP is GAPs revenue is, which is GAP defines revenue as the gross inflow of economic benefits during an accounting period from an entity's ordinary activities recognized when goods or services are transferred to customers for an amount that reflects the consideration the company expects to receive.

This is not recognized when cash is received, but rather when earned and realized through a five step model, contract identification, performance obligation identification, transaction price determination, price allocation, and revenue recognition as performance obligations are satisfied. The last part I think is makes things more complicated and confusing. but let's just do the core principles that they lay out, which is that one is that revenue is earned. couldn't agree more.

And then it immediately says revenue is recognized when it is considered earned. And I think that those verbs of recognized and earned are absolutely critical to understanding what revenue represents. When it is earned and realized, meaning the company has transferred control of the promised goods or services to the customer and has a reasonable expectation of receiving payment. So the key idea here is it's goods and services being

Alex Oppenheimer (04:47.138)
being delivered and made available to those customers. Now that made available portion opens you up to the entire world of digital goods. Again, when this term revenue was invented, there was no such thing as digital goods, but there was services, obviously. And so most services were delivered by humans, not digitally, but the same concept here applies. So there's this continuum of definitions, which I'll add it in the description.

of something that my dad and I actually sat down and wrote a couple of years ago, which was revenue related definitions in SAS. And there are so many different things. I'm not going to run through all of them now. I'll the link below. But it's a table that has billings, bookings, revenue, backlog, ARR, MRR. I mean, all the different, TCV, ACV, all these different things and what they all mean and what they can mean and what they don't mean. But let's now.

hit the nail on the head, is this concept of ARR. What is ARR? Now, I am not the arbiter of the rules, but I will share what I believe strongly makes sense. And then I will talk about the two things that are implied by ARR, which is why it is coming into question now. So what is implied by ARR is annualized recurring revenue. So

We just talked about that first, that, that second R, the revenue part. If we don't understand what that is, nothing else matters. Let's talk about the first R, which is recurring. Now, what is recurring revenue? my, again, working definition, which is by no means the rule is that it is revenue that you can reasonably expect to recur. Now, sometimes that means contractual obligations, three-year contract.

you know, great paid upfront, paid upfront is the best way because there's not a lot of refunds in the software world. So if the customer has access to it and they continue to have access to the product and they have paid you upfront, you will almost certainly recognize all of that revenue 11 months from now. But usage based, you know, is that ARR again? I think it depends on how volatile it is.

Alex Oppenheimer (07:10.36)
Frankly, you have to look into that and say, what is the recurring, I'll make up a word, recurring-ness of that revenue and how reliably can we count on it being there next month? And that's a really, really key idea. And then annualized. That's a really important idea as well. We are annualizing something and you can annualize whatever you want. The example I always give is ice cream cones sold.

You could annualize ice cream cones sold in July in one hour. And you can extrapolate that, which is annualizing is an extrapolation. You could extrapolate that out across the entire year. Now we generally think that's crazy, right? Because in January you sell less ice cream than July and, and by the way, the store's closed for at least 12 hours a day. So you've got to be clever with how you really think about what does it make sense to annualize something. And so the reason that we

Run rates are not a new idea. The reason that we feel comfortable annualizing this revenue is that it is recurring. So these two terms really go hand in hand to make it mean something. And going back to what I mentioned earlier, this kind of calculus, I've got this analogy in a lot of my presentations where I talk about position, velocity, and acceleration. And then the distance traveled is the integral of velocity over a period of time.

So in that exact same way, the recognized revenue over a period of time is the integral over that period of time of the function of ARR. Now, ARR is not often a mathematical function, but it kind of is, right? If you think about a term or a business mechanism, the term would be the SAS magic number, which is simply the net new ARR divided by the associated CAC. That is your.

magic number. So if you have a machine like that and you know what that function is, it's, you know, if it's just a simple multiplication, let's say it's 1.5. So we put $100,000 in, we're going to get $150,000 of net new ARR. We don't necessarily know exactly what the shape of that curve is, but you could probably figure it out. Oftentimes, you know, salespeople are hitting their numbers at the end of the month or the end of the quarter, and you could with enough data, you actually could draw that curve.

Alex Oppenheimer (09:36.438)
with a decent level of precision. Would it really matter? Probably not. The concept matters more than the calculations by a lot. So again, this idea of annualized recurring revenue, the reason we care about it is that ARR is a leading indicator of recognized revenue. Recognized revenue is the first line on the income statement. People call it turnover. They call it sales. That's what it is.

And it is the number one globally accepted. Topline operating metric that allows you to benchmark a railroad company and a social media company to each other, which is very, very powerful. And we have to give credit where credit's due to the invention of revenue. But again, there's this idea that if we know how much revenue we're going to have 11 months from now, that's worth something.

And so what ARR really is, again, it is a leading indicator. And the implication of that leading indicator is if nothing changed in our customer base in the next 12 months, we will earn and recognize this much revenue, which is our ARR over the next 12 months. I have slides that have integral symbols and visualizations of shapes, but effectively if you say we have

$1 million of ARR, which again, I use the wrong verb there. If we are at $1 million of ARR, it's sort of like at being at 60 miles per hour. What you're saying is if over the next 12 months, nothing changes, we don't go faster, we don't go slower. We don't have churn, we don't have upsell, we don't add any new customers, we will recognize $1 million of revenue. It's same thing as if a car is going 60 miles an hour. The implication is that

If nothing changes in the next hour, then we will cover 60 miles of ground. And that's helpful for me. think a lot of people kind of intuitively understand cars moving and the concept works the same. But again, it's that reliability of the revenue that we care about. Now, ARR and SaaS have been conflated for many, many years. SaaS is a very powerful business model, enterprise software.

Alex Oppenheimer (12:00.002)
You know, the guys in at, over at acquired talk about this all the time, like the enterprise software business models, the greatest thing ever. Right. get annual high margin annual contracts paid upfront, with multi-year contracts and opportunities for expansion. There's never been a better, better business model ever in human history, especially one that's legal. So it's extremely powerful, but I said something in there that's super important, which is that it's high margin and.

high margin like matters a lot. It gives you a huge margin of error. And when you think about the accumulation of ARR and really that as a result of revenue for one of these enterprise software businesses, the power lies in the fact that to go from let's say zero to a hundred million of revenue, let's say you have to spend $150 million, which would you put you at a magic number less than one. Now,

and a loss just on sales and marketing, let's say of $50 million. Now, let's say you do the same thing next year with 100 % retention from the previous year. That means that you're going to grow from 100 million to 200 million, and you're also going to spend $150 million. Now, let's say you do that again, you go from 200 to 300, you also spend $150 million. Assuming that stays relatively consistent, you're not necessarily trying to accelerate growth. This would actually be decelerating growth on a percentage basis.

then that's extremely powerful because your margins will naturally improve over time. Now, obviously we see that they kind of don't, but they also kind of do. The way that margins are reflected for software companies that are public, I think is very inaccurate and has to do with legacy definitions that are used in the accounting world and bookkeeping and the less to do with the economic reality of those businesses. I'll give you an example. If you put customer success people in sales and marketing,

that gets reflected in CAC when in reality that is an ongoing variable expense. The more customers you have and the longer you keep them, the more money you spend. Whereas a pure account executive is paid to bring in new customers. Yes, they have a base salary and a commission structure, but that is what they're paid for. So that's a really critical element here. And I'll go back, you know, when lemonade here in Israel,

Alex Oppenheimer (14:28.928)
announced that they had hit a hundred million of ARR, people were not happy. They got ripped in the media and Twitter. I mean, it was brutal. And I didn't know anything about insurance at the time. I sat with them and I kind of, and I wrote about this as well years ago, maybe five years ago. Like it actually kind of is ARR. You could argue that insurance premiums are actually more reliable from like a churn perspective historically than a lot of SaaS revenue.

So like, what's the issue? And the issue is that it's the shape of the revenue, we'll call it, and what hits the bottom line and how it hits the bottom line is very, very different. And so what they got hit for with the wrong, in kind of the wrong way was trying to pretend to be a software company when they're like an insurance brokerage, basically, and an insurance company, have multiple models. But I thought, frankly, that was super unfair. And it actually, on the recurring aspect of ARR,

was even better than the...

a lot of SaaS companies, but the margin profile is obviously very, different. So again, this concept of ARR, annualized recurring revenue, like why is everyone up in arms about it now? I'll use another example. A couple of years ago, or guess a year and a half ago, maybe one of my portfolio companies said, we're going to start using Toast's definition of ARR.

So I was like, hmm, let me go open Toast 10K and see how they define this. And so I will bore everyone with reading this. They define ARR as annualized recurring run rate.

Alex Oppenheimer (16:19.224)
They put a hyphen between run rate. Otherwise, I always say then it would be R, which is strictly for pirates. And here's their explanation written in their 10K. 10Ks are notorious in these sort of definitions for being super, super wordy. So we'll see if it meets expectation. We monitor annualized recurring run rate as a key operational measure of the scale of our subscription and payment processing services for both new and existing customers. To calculate this metric,

we first calculate recurring run rate on a monthly basis. Monthly recurring run rate or MRR is measured on the final day of each month as the sum of one, our monthly billings of subscription services fees, which we refer to as the subscription component of MRR, and two, our in-month adjusted payments service fees, exclusive of estimated transaction-based costs, which we refer to as

the payments component of MRR. MRR does not include fees derived from Toast Capital or related costs. MRR is also not burdened by the impact of SaaS credits offered. MRR calculation includes all locations on the Toast platform and locations on the legacy solutions, which have a negligible impact on ARR. ARR is determined by taking the sum of

12 times the subscription component of MRR and four times the trailing three months cumulative payments component of MRR. We believe this approach provides an indication of our scale while also controlling for short-term fluctuations in payments volume. I'm gonna just pause there. It's super wordy, it's complicated. First, I'll give credit where credit's due. I am all for companies, especially ones that have like real operating data, inventing metrics that suit their business model.

I would never be opposed to that. I think it's super, super important so that you can benchmark yourself against yourself, which is the most important thing to be benchmarked against to figure out if you're running efficiently, how you optimize that, how changes impact the top line and the bottom line of the business. That's great. The problem is here, and again, I was a banker. I understand how these things come to be. I've written some of these things like

Alex Oppenheimer (18:39.522)
They're just throwing it all together and trying to then smooth it. And one of the issues is that the recurring nature, again, I'll use this made up word, recurring-ness again, of the subscription component and of the payments component is different. It's just the payments component is never going to recur as reliably. If you think about it from like a DCF perspective, the discount rate is higher because it goes up and down.

year round, there's seasonality in most of their customers. And then the second issue is that, so they say our in-month adjusted payment service fees, exclusive of estimated transaction-based costs, which means they're not netting like the gross costs, but for payments, the gross, like the cogs is super, super high. Like they have to pay a payment processor. And so they're taking that bigger number. Now, if you

compare that to the subscription component of their revenue and that ARR, I imagine that's a very high margin ARR. And by the way, the CAC is associated with both of these things. Now I'm all for, we want to represent the scale of the business and how we think it's totally growing. But the problem is, if you think about it again, I'll bring back the DCF idea, right? If you were to do a discounted cashflow analysis,

on the value of this revenue. Obviously we want to get the cashflow and then we want to figure out how reliable is that cashflow. Those are the variables that matter. I'll probably talk about this in another podcast and my kind of oversimplified startup DCF for directional valuation modeling. But the key variables in that are one, the short-term growth rate, like how fast is the thing growing, which we're getting from this. Great. The second thing we need is the free cashflow margin.

Here we're kind of putting everything together. And so it's very, very different, which again, it's not a problem. The company's free cashflow margins, the company's free cashflow margin, it's not a big deal. You know, it is what it is. But where we run into an issue is when we start talking about R, which is the discount rate. And because the discount rate of part of that cashflow is very different than the discount rate of another part of that cashflow. So just the very fact that they're admitting they have to smooth it,

Alex Oppenheimer (21:04.64)
from month to month to represent it means that it's more volatile, means that it's less reliable, means that it has a higher discount rate. Now, I don't know who else nerds out this hard on this stuff as I do, but if you see all the pushback now and a lot of these AI companies and is it really ARR and the death of ARR and da, da, da, da, I think part of the issue is that these were the same people that flipped out when Lemonade said it had ARR. Like, well, it kind of does and it kind of doesn't.

If the gross retention is super low, is it really ARR out? And at what point do you draw that line? At what level of volatility on an up and down basis? And at what level of volatility on a, like a kind of a customer churn, like just customers going to zero basis, do you draw the line? You know, if you have a consumer SaaS company, like look at Netflix or Spotify and they've got subscriptions and people are staying on those.

platforms for years and years and years. I don't know what the churn is off top of my head. The way that Spotify, as I recall, didn't, you they did direct listing, but in their S1 equivalent, defined their churn was like the most wordy, complicated thing. was like a, it was like a quarterly churn rate taken on a monthly basis and then averaged over the year. I'm like, I don't know, but it looked nice. It was like 10 % or something. So if you're losing 10 % of your customers a month,

If you annualize that, it's not really just 10 % times 12, which would be 120%, which would mean that you're churning your entire customer base every year. You really have to look at it on a cohort basis. So it's 0.9 to the 12th power and then one minus that, which the bigger that number is, the bigger the differences between the simple multiplication. If you're doing 1 % churn a month,

and you do that same like 0.99 to the 12th power, and then you do one minus that, it's gonna be like around 12%. But if you're at like 10 to 15 % a month, it's a problem. Anyway, if it's 20 % a month, people will be like, is that really SaaS? Like, okay, you you could very clearly show like retention numbers and as a direct, you know, causal correlation to multiple. Now, usually we don't know the direct.

Alex Oppenheimer (23:29.366)
You know, retention numbers, sometimes they'll publish NRR, which again, there's another, another post for that because when you're using a combined metric and it is based on other things that are subjectively defined, you just are getting to a point where it almost means nothing anymore. I've seen eight different ways of counts calculating NRR and they might all be wrong. I don't know. So, you get into a lot of questions there. And so.

When people look at some of these SaaS products that everyone's going to try right now, we're sorry, one of these AI products, everyone's trying them. Everyone's happy to try an AI product. It's why not? Right. But then when they can't make it work for them or they run into issues or they find hallucinations, they quit and they stop paying. So churn on these things as high because it's basically just a paid trial. Now they can access central question. Are you going to knock a company for

paid trials, because they could also just give three months for free and let people turn off that. And then they would never have had the up and down. The other way to do it when you get into these companies at scale is you just don't really count it towards your ARR number until they're like three months in. And that's fine. You just have to denote that in your definition. But again, this is a monthly thing. That's revenue.

You delivered something to them. used it. It doesn't matter. doesn't nothing. doesn't say anything about it, whether they liked it or not. They had access to it. Doesn't even matter if they used it. Frankly, that's recognized revenue. So that's all to say this whole death of ARR movement. I think it's just as in most cases, a term got bent out of shape, overused, misused, not well understood. A bit of a cliche, but if you understand the power.

of this metric and why it came to be, you know, really did it exist before 15 years ago? Maybe, I don't know, we could look up when ARR, the term ARR was invented. Some of the early researchers on kind of this math were like, you know, I think it was Joel York and Michael Scott, they, you know, invented some of these original terms like SAS metric number and stuff like that. anyways, that's all to say that I think NRR is useful.

Alex Oppenheimer (25:51.802)
Sorry, ARR is a very, very useful metric. When it's used properly, there are two aspects to it that are implied by it. One more definitionally, one more kind of colloquially, definitionally, it implies a certain amount of retention. What exactly that is and what those dynamics are, the world has not agreed upon. And then it also implies a certain type of business model and a certain margin profile of that business, which...

Again, it's a separate thing that often gets ignored. We're talking about revenue. You don't have to tell me about gross margin, contribution margin. That's not what anyone claimed when they talked about ARR. So I will leave things there. Happy to continue the discussion. We'll probably dig deep on NRR and these concepts and everything else in podcasts in the near future. Hope everyone enjoyed. As always, feel free to reach out with questions, comments, criticisms.

or if you want to jump on with me and discuss one of these ideas live.