I am not an accountant and am still working my way through the post, but I think two things already
1. I think you might be conflating gross margin and contribution margin. The latter is about variable costs. You seem to argue that COGS is the place for variable costs and it's actually the place for production costs. Unsure, but wanted to get that out there.
2. Agree that SaaS took us to a place where costs of the running the software are eaten by the vendor and put into COGS. Hence the glorious days of 98% GM are long behind us. In effect, SaaS vendors are in two (inextricable for now) businesses: building the software and running the software.
3. I've always believe that software companies are worth a lot not because they are profitable but because they have the potential to be at scale. Most trade-off profit in favor of acquriing new customers to grow (and that gets rewarded). But if you look at at-scale software companies, they throw off a lot of free cash flow (e.g., CRM $9B, ORCL $8.5B)
What are the consequences for the existing VC model? If stage 3 is recognized as myth, that makes the VC's job to sell investments of multiples of X returns that much harder.
Seems the myth lives to serve, and as long as we don't come up with a better, more lucrative story, we may yet be trapped in the fantasy.
Maybe end of ZIRP will lead to a rude awakening, but we still need a new narrative to act as cornerstone for the next generation of more sustainable (dare I say rational) companies.
Any bets on what that new cornerstone belief may be?
Yeah, I mean, I think that's the ponzi-like scheme in all of this. The VC model isn't trying to figure out how much cash a business will eventually generate; it's trying to figure out how much some future investor will eventually pay for the company's shares. There's no requirement that the "business fundamentals" need to match the share price. People could just collectively decide it's worth a lot, and it'll be worth a lot.
If there's a new cornerstone though, I think it'll be more about lowering the cost of "design" (ie, making the software). Like, the zero marginal cost of delivery is a thing and is real, so the model works great if you don't spend so much money building it. And I think we probably can make that a lot cheaper. I'd guess that a lot of software companies are pretty bloated, and their development costs are higher than they need to be, in part because of this assumption that they can be.
(Fwiw, I think that bloat isn't that engineers are overpaid or lazy or whatever; it's that companies do a lot of things that aren't valuable. Today, you can build a lot of stuff that doesn't have any impact, and kind of handwave it away. I suspect over time, that's what changes.)
We can only truly grow once we embrace the reality instead of excusing it. Really hope more people get into the habit of digging themselves out of the bullshit holes we ourselves have been digging for such a long time ...
Basically, you're arguing that the stock market doesn't know how to value software companies and I think that's incorrect. The reason public SaaS companies have good gross margins (median 77% per Meritech) and only modest operating margins (13%) is, in a word: sales & marketing (34%). The shit is expensive to sell. We thought PLG would help, but it didn't (perversely, I think TT showed that PLG companies spend more on both R&D and S&M). The reality is any SaaS company could print money if decided to stop acquiring new customers (that 13% operating margin could shoot to 43%). But they get valued for growing so they make the investment to grow. And once they hit scale, big scale to your point, then they start to repeat the benefits. It's one reason why they often sell for nice prices in M&A -- because if I can leverage my S&M to sell your software, then I get real synergies.
Also, as per my other comment, I think you're conflating COGS/GM with a different concept (contribution margin) that measures (all) variable costs and only variable costs. But ultimately profitability isn't about GM. If you put R&D into COGS or OPEX it doesn't change the bottom line -- cash flow or net income. I agree that the market appears to value software companies on their potential, which to me assumes both large TAMs (true) and the ability, one day, to print money.
For fun, you might invert your perspective and see what you come up with. I learned this from a friend when we were arguing about the second-serve in tennis. He said to me, "Dave, you're arguing that the established practice is wrong. Why don't you try it the other way. Given that virtually all professional tennis players have a second serve, try to understand why." That is, do a similar post from the POV of trying to explain/defend software valuations. I'd love to see what you come up with.
(Combining my response to this comment and the other one)
I'm not an accountant either, but I think you're right about gross margins vs contribution margins? That said, I used gross margins because it was much easier to find data on that than contribution margins. Data on these things is pretty messy (or at least for me), and finding SaaS vs S&P GM and operating margins was fairly straightforward, but not so for contribution margins (or net margins, which is probably a little better to use than operating margin). I wouldn't think that distinction would make much of a difference though, and software companies would have high contribution margins relative to other businesses too.
Still, amateur accounting liberties aside, I think the blunter question remains. No matter how you do the specific math, software companies have a huge potential edge over other industries, in that the thing they can "manufacture" a high-dollar product for almost free. And yet, as TT showed in his post, the biggest and best software companies are only slightly more profitable than everyone else. That's the real oddity to me here; given that giant advantage, shouldn't the gap be bigger?
My best guess we're both to some degree right here. Part of it is that production costs (be it all variable costs or COGS, I don't really care about the specific accounting) is a lot higher - and persistently higher - than people assume. And I think you're right that software is expensive to market and sell. Though that's a very interesting point, because if S&M expenses eat up the huge gross/contribution margins of software businesses, that implies that software is a lot more expensive to sell than other products. Which I've never seen anyone actually argue explicitly, so now I wonder.
I agree with you that a lot of valuations are based on potential (and potential scale, more than margins, probably), though I'm not sure that doesn't imply to me that things are overvalued. Like, I think you can put software companies in four rough buckets:
1. Private companies, which mostly lose money
2. Small publics (eg, the company Jamin tracks, which are roughly break even (bad margins, better FCF, etc)
3. Big publics (eg, S&P 500 companies, which per TT's analysis, are probably slightly better than the average S&P 500 company)
4. The whales (eg, MSFT, Meta, Google, which print infinite money)
It seems to me that people look at the whales - "tech companies can make unlimited amounts of money!" - and roughly assume that companies in the other three buckets can either get to that scale, or can capture the same economics around margins. But neither of those things seem particularly true to me. If you get that big, sure, the combination of scale and good economics will make you insanely rich. But most companies can't get to that scale - and without that scale, you don't get the economics either. I don't think markets discount the former - the odds of getting that big is hard - but not really the latter - if you don't get that big, the economics are kind of meh.
Which I think is my answer to your last question (though I need to think more about that). It's that people see the best software companies, believe those economics are available to smaller companies, and don't fully discount how unlikely it is for a company to achieve it. (That, and tech is sort of Ponzi-ish, where valuations are to some degree set by people's expectations of other people's valuations. And once that baseline gets inflated, it's not that easy to push it down.)
Great description of the VC-funded SaaS model, really insightful.
Adding to the Veeva example, another category of software businesses with lower on-going development costs are very niche software applications. Think of those that hold cos love to buy, e.g. software for orchestra management, kindergarten management, CRM for metal workshops etc. The type of customer that wouldn't buy software in a box because they don't know how to host it. The sales argument "buy our software instead of hiring a part-time secretary" works really well there and also precludes any expectation of continuous improvement because the comp (the secretary) doesn't improve every quarter either
I think that's true, but most of the niche products I know like that run into scale problems. If you're selling to orchestras, your market isn't that big, which makes it a lot harder to cover your initial development costs. This seems like a huge problem in things like ed tech - the products are good and the markets are stale, but nobody can pay anything for them.
I think part of the poor margins has to do with the sequencing of things required to succeed - which you touch on.
So you can’t effectively hire a team to build your product and then layoff the entire team and have sales sell it - then bring back the team that built it to improve it when you really need them - then layoff the entire sales team when you are hitting a nice organic growth stride - then bring back the sales team when organic growth starts to dwindle… etc.
My point being even if everyone is “staying busy” the one crucial thing you need to do next won’t usually involve everyone at any given point in time. So therefore you end up with resources that you needed last quarter or may need next quarter that are getting paid this quarter.
I suppose this is where I could pitch fractional work and agencies - which sure that can be helpful, but I honestly think this is mostly unavoidable. You need a large number of people that are bought in when you are scaling a startup. Unavoidable side effect is a good number of people at any given time are idle on the current “growth blocker.” Oh and not to mention we don’t often actually know what the next unlock is going to be (often is just time). Which means the entire company may just be staying busy waiting for any significant growth to occur.
Yeah, so there's another way to look at all of this, which is basically that software companies undisciplined. They overhire, build a bunch of stuff that people don't really want, do a lot of busywork, and so on. I think that's probably true, at least in terms of how you'd ideally want to run a business. But I think your point is right, which is part of that is because hiring is a ratchet, and if you hire a bunch of people to build the initial thing, you're very unlikely to scale down the team after that. So, you keep building, expand more than you should, and so on.
Which, might be the real impact AI has on all of this. It's less about replacing engineers exactly, but more about making your production capacity more variable. We do the same for our AI bots that we do for cloud compute, and scale teams up and down as needed.
The AI bot comparison to cloud computing thing is really interesting. If that can turn into scaling departments up and down I think it would have a significant impact on the economics for sure.
Ha - I just thought of this. The other terrible thing about all of this is the statistics on what percentage of features people actually use in a product. It’s 20% or something like that based the last stats I read.
So - borrow money to constantly improved the product and keep adding features all of which 80% isn’t going to be used. Yikes. I think modern product analytics and product practices help with this - but still…
Yeah, if anything gets "corrected," I think it'll be that - that software companies spend a lot less money building things that they don't need to build. It feels like a lot of software development is a kind of busy work, that has all the appearances of being productive, but doesn't really mean much to anyone.
You're right that the software customer is buying a TRAJECTORY of technology, not the current features. As the customer's need changes, as compute capabilities change, as user preferences and budgets change, so must the solution evolve or it will die. The software business will incur costs to accommodate product evolution.
To the extent conventional accounting fails to capture it, an investor might assume that some of the costs of future innovation will be recoverable in annual price increases or through breaking out the new features and selling them for additional fees, and then add back what's left to appropriately burden the valuation.
I agree that that's how investors often think about this, but I think the same assumptions are still embedded in that. Like, I think SaaS businesses are actually more like magazines than product companies. Magazines can't release new articles and thing we can sell them for additional fees; they realize the new articles are what people are paying the existing fees for. When software companies sold software in a box, customers expected to buy what was in the box. Now, they expect to buy something that's like a magazine - a steady cadence of new releases. And you can't charge for that; that's the expectation.
Yeah, there are definitely companies that can get there; it's not that no software company works. I didn't really think very much about which ones do and don't, and that seems like a reasonable heuristic. If you're selling to a mature industry that wants you to solve a problem that's stable and they don't actually want the constant upgrades, you might be able to get there.
This was an interesting piece (that I just realized is 7 months old).
The part I latched on to: do we buy SaaS for what it will become (at least in part)? I'm not sure. In instances like chatGPT I think that's obviously true. But for your corporate accounting software? I think "what it might become" is a risk. How many pieces of software have you used in your life that eventually made changes bad enough they forced you to migrate? Alternatively, I think of games like World of Warcraft, where literally any changes to the game are hated by half the population.
Regardless, the business of software is certainly more complex than 20 years ago.
Yeah, that's fair. Though in some ways, I think that makes the dynamic even trickier, because:
1.) You sell v1 to customer A, who loves v1 and wants it to be v1 forever.
2.) Competitor keeps developing, markets change, there are new expectations, etc etc.
3.) You pitch customer B and they say, hm, we really want all these new features first.
4.) And now what? B wants v2, A wants v1, and you have to sell to both.
I think you're right, that existing customers might sometimes want a fixed product that never evolves. But new customers want the latest and greatest. And SaaS economics only work if you can sell to both.
Late to the party here but I've often thought about where software development costs show up on the balance sheet. Strictly speaking they *don't.* A manufacturing company like Caterpillar shows "inventory" and "raw materials" on their assets, but software companies don't show the SaaS equivalent – or the equivalent liabilities that you're outlining here like the required future cost of development. Although hilariously if a SaaS company was to be sold, the IP of the software is definitely an "asset" in a vague sense.
I also wonder if this holds in more non-traditional (uh, non-Silicon Valley) SaaS. A friend runs a house remodeling company and they just purchased software specifically designed for contractor project management. It's a bootstrapped SaaS company and seems profitable as hell.
What's also interesting about that is some software companies report on "remaining performance obligations" (like here https://investors.snowflake.com/overview/default.aspx), which are future contractional obligations customers have already committed to, which is essentially the asset side of that liability. Which feels like some real accounting voodoo to be able to do that.
On the "non-traditional companies," I do wonder how profitable a lot of those businesses really are. That model has become appealing to a lot of Silicon Valley types recently, and while I'm sure there are some that make a lot of money, I'd guess most are running roughly breakeven (ie, people make good salaries and all that, but there's not a lot of extra profit for owners/shareholders. But I have no idea.)
Ooh gonna have to dig into that, I hadn't heard of that before. But on first impression - real accounting voodoo haha, especially for companies that sell upmarket and need to customize software. A manufacturer having to customize production lines would definitely have more transparency on the capital investment required and the asset created.
I'd say maybe a famous example of a non-Silicon Valley business succeeding with great profit would be 37Signals (notwithstanding the investment from Jeff Bezos they took many years ago). They've been fairly public that they make millions in profit every year, and have for many years. Mailchimp was another bootstrapped exception to the rule.
Is every non-Silicon Valley that successful? No. And that's where I think – as Cedric @ Commoncog has been writing about recently – the "operational efficiency" of the company matters a lot less than the chosen niche, product strategy, macro environment, etc etc. So yeah, probably broadly agree that many aren't running deeply profitable firms because they don't have as many customers (or the resources to acquire them at a fast clip).
Yeah, 37signals is the example that a lot of people use. My guess is that’s actually somewhat repeatable, but the hard part is in the 15 year slog to get there. Like, if you wanted to build that business, a lot of people probably *could*; they’d just have to be reallllly committed to doing it, and would have to endure a lot of lean years first. And that’s sort of anathema to the entire Silicon Valley culture of moving fast and breaking things.
I used to also think that operational efficiency could have a moment because of the down market, but I’m not sure that’s true. Growth is too much a prisoner’s dilemma - if the market is down and everyone is being operationally efficient, you’re better off being the big spender. And if the market is up and everyone’s a big spender, then you’ve got lots of money to spend so spend it. The only time when there’s a real incentive to be efficient is if you’re playing a different game than everyone else.
Will these margins naturally IMPROVE post-ZIRP, with the collapse/consolidation of B2B SaaS madness?
Here's the argument: B2B software buyers have benefited from an incredible explosion of SaaS products that collectively compete away margin from each other.
Two vectors of competitive pressure:
1. Feature competition
2. Pricing depression
Perhaps it's too easy to build and market a B2B SaaS product now, and too easy to iterate/add features. And SaaS was too cheap.
In 2024, hundreds of SaaS "companies" with "products" will realize that they were only ever a feature anyway... and will need to be consolidated into larger companies, or die.
My theory points to "zero cost money" plus "software is easy to build" as the true culprits for poor margin. We're all witnessing a fundamental restructuring of the SaaS landscape underway right now. The terminal result will be fewer and more expensive products owned by bigger companies, with customers no longer able to simply switch out to the cheaper, more feature-rich upstarts.
I see natural selection happening in SaaS every day. Friends are losing their jobs and companies, and it sucks. But it might be cruel, necessary, and natural selection in a world where money costs money (again).
Bonus take: Will AI make SaaS feature development smarter, faster, more profitable? I'm betting my own company (LeftHook.com) on that belief. More/better integrations for SaaS companies should be cheaper, faster, and less of a headache, and AI will be at the center of making this reality.
Eh, I'm not sure how much competition has to do with it for me. I'm sure that matters, but I think it matters more in that competitive pressure pushes SaaS companies to keep building and having new things sell. I'm sure there's some pricing pressure, but I imagine the bigger issue is that competitors prevents companies from ever getting to that hypothetical "steady state" point where R&D spend can be relatively low, and they can reap the benefits of the high margins.
It's hardware, but my rough sense of the smartphone market is that, if Apple stops making tons of money from selling iPhones, it wouldn't happen because Samsung et al undercut them on price; it's would happen because Apple puts out a bunch of incremental improvements, people feel like iPhones aren't doing anything new and exciting, and they get bored and drift away. I'd imagine that dynamic is kind of the same in software.
What you seem to be arguing implicitly is that there has been and continues to be a big arbitrage because Wall Street bankers continues to hold a (mostly) false assumption that "steady-state condition" is attainable for SaaS companies. This belief drives the underwriting of exists for VCs.
The other takeaway is that VCs don't care if their SaaS companies turn a profit, it just has to grow revenue at a high enough clip to exit. And maybe bankers don't (truly) care whether the IPO they underwrite will turn a profit either, because if a good eventual "steady-state condition" story is told and sold, other institutional investors will buy the shares anyways.
Yeah, I think that's basically the idea, though I don't think it's necessarily an arbitrage. Like, if everyone believes SaaS businesses are worth a lot, then they're worth a lot. There's no rule that says they have to return to some "steady-state condition" based on how much profit they make for shareholders. Like, the price of gold has been a "bubble" for all of human history - people value it because they assume other people will continue to value it in the future.
I am not an accountant and am still working my way through the post, but I think two things already
1. I think you might be conflating gross margin and contribution margin. The latter is about variable costs. You seem to argue that COGS is the place for variable costs and it's actually the place for production costs. Unsure, but wanted to get that out there.
2. Agree that SaaS took us to a place where costs of the running the software are eaten by the vendor and put into COGS. Hence the glorious days of 98% GM are long behind us. In effect, SaaS vendors are in two (inextricable for now) businesses: building the software and running the software.
3. I've always believe that software companies are worth a lot not because they are profitable but because they have the potential to be at scale. Most trade-off profit in favor of acquriing new customers to grow (and that gets rewarded). But if you look at at-scale software companies, they throw off a lot of free cash flow (e.g., CRM $9B, ORCL $8.5B)
More later as I process this article more.
What are the consequences for the existing VC model? If stage 3 is recognized as myth, that makes the VC's job to sell investments of multiples of X returns that much harder.
Seems the myth lives to serve, and as long as we don't come up with a better, more lucrative story, we may yet be trapped in the fantasy.
Maybe end of ZIRP will lead to a rude awakening, but we still need a new narrative to act as cornerstone for the next generation of more sustainable (dare I say rational) companies.
Any bets on what that new cornerstone belief may be?
Yeah, I mean, I think that's the ponzi-like scheme in all of this. The VC model isn't trying to figure out how much cash a business will eventually generate; it's trying to figure out how much some future investor will eventually pay for the company's shares. There's no requirement that the "business fundamentals" need to match the share price. People could just collectively decide it's worth a lot, and it'll be worth a lot.
If there's a new cornerstone though, I think it'll be more about lowering the cost of "design" (ie, making the software). Like, the zero marginal cost of delivery is a thing and is real, so the model works great if you don't spend so much money building it. And I think we probably can make that a lot cheaper. I'd guess that a lot of software companies are pretty bloated, and their development costs are higher than they need to be, in part because of this assumption that they can be.
(Fwiw, I think that bloat isn't that engineers are overpaid or lazy or whatever; it's that companies do a lot of things that aren't valuable. Today, you can build a lot of stuff that doesn't have any impact, and kind of handwave it away. I suspect over time, that's what changes.)
Outstanding insights, Benn !! Hard and fair.
We can only truly grow once we embrace the reality instead of excusing it. Really hope more people get into the habit of digging themselves out of the bullshit holes we ourselves have been digging for such a long time ...
Thanks! And the problem is that the holes are fun and sometimes full of money...
Yepp, this is just one example of the gifts given by the materialistic world we created ourselves.
Basically, you're arguing that the stock market doesn't know how to value software companies and I think that's incorrect. The reason public SaaS companies have good gross margins (median 77% per Meritech) and only modest operating margins (13%) is, in a word: sales & marketing (34%). The shit is expensive to sell. We thought PLG would help, but it didn't (perversely, I think TT showed that PLG companies spend more on both R&D and S&M). The reality is any SaaS company could print money if decided to stop acquiring new customers (that 13% operating margin could shoot to 43%). But they get valued for growing so they make the investment to grow. And once they hit scale, big scale to your point, then they start to repeat the benefits. It's one reason why they often sell for nice prices in M&A -- because if I can leverage my S&M to sell your software, then I get real synergies.
Also, as per my other comment, I think you're conflating COGS/GM with a different concept (contribution margin) that measures (all) variable costs and only variable costs. But ultimately profitability isn't about GM. If you put R&D into COGS or OPEX it doesn't change the bottom line -- cash flow or net income. I agree that the market appears to value software companies on their potential, which to me assumes both large TAMs (true) and the ability, one day, to print money.
For fun, you might invert your perspective and see what you come up with. I learned this from a friend when we were arguing about the second-serve in tennis. He said to me, "Dave, you're arguing that the established practice is wrong. Why don't you try it the other way. Given that virtually all professional tennis players have a second serve, try to understand why." That is, do a similar post from the POV of trying to explain/defend software valuations. I'd love to see what you come up with.
(Combining my response to this comment and the other one)
I'm not an accountant either, but I think you're right about gross margins vs contribution margins? That said, I used gross margins because it was much easier to find data on that than contribution margins. Data on these things is pretty messy (or at least for me), and finding SaaS vs S&P GM and operating margins was fairly straightforward, but not so for contribution margins (or net margins, which is probably a little better to use than operating margin). I wouldn't think that distinction would make much of a difference though, and software companies would have high contribution margins relative to other businesses too.
Still, amateur accounting liberties aside, I think the blunter question remains. No matter how you do the specific math, software companies have a huge potential edge over other industries, in that the thing they can "manufacture" a high-dollar product for almost free. And yet, as TT showed in his post, the biggest and best software companies are only slightly more profitable than everyone else. That's the real oddity to me here; given that giant advantage, shouldn't the gap be bigger?
My best guess we're both to some degree right here. Part of it is that production costs (be it all variable costs or COGS, I don't really care about the specific accounting) is a lot higher - and persistently higher - than people assume. And I think you're right that software is expensive to market and sell. Though that's a very interesting point, because if S&M expenses eat up the huge gross/contribution margins of software businesses, that implies that software is a lot more expensive to sell than other products. Which I've never seen anyone actually argue explicitly, so now I wonder.
I agree with you that a lot of valuations are based on potential (and potential scale, more than margins, probably), though I'm not sure that doesn't imply to me that things are overvalued. Like, I think you can put software companies in four rough buckets:
1. Private companies, which mostly lose money
2. Small publics (eg, the company Jamin tracks, which are roughly break even (bad margins, better FCF, etc)
3. Big publics (eg, S&P 500 companies, which per TT's analysis, are probably slightly better than the average S&P 500 company)
4. The whales (eg, MSFT, Meta, Google, which print infinite money)
It seems to me that people look at the whales - "tech companies can make unlimited amounts of money!" - and roughly assume that companies in the other three buckets can either get to that scale, or can capture the same economics around margins. But neither of those things seem particularly true to me. If you get that big, sure, the combination of scale and good economics will make you insanely rich. But most companies can't get to that scale - and without that scale, you don't get the economics either. I don't think markets discount the former - the odds of getting that big is hard - but not really the latter - if you don't get that big, the economics are kind of meh.
Which I think is my answer to your last question (though I need to think more about that). It's that people see the best software companies, believe those economics are available to smaller companies, and don't fully discount how unlikely it is for a company to achieve it. (That, and tech is sort of Ponzi-ish, where valuations are to some degree set by people's expectations of other people's valuations. And once that baseline gets inflated, it's not that easy to push it down.)
Great description of the VC-funded SaaS model, really insightful.
Adding to the Veeva example, another category of software businesses with lower on-going development costs are very niche software applications. Think of those that hold cos love to buy, e.g. software for orchestra management, kindergarten management, CRM for metal workshops etc. The type of customer that wouldn't buy software in a box because they don't know how to host it. The sales argument "buy our software instead of hiring a part-time secretary" works really well there and also precludes any expectation of continuous improvement because the comp (the secretary) doesn't improve every quarter either
I think that's true, but most of the niche products I know like that run into scale problems. If you're selling to orchestras, your market isn't that big, which makes it a lot harder to cover your initial development costs. This seems like a huge problem in things like ed tech - the products are good and the markets are stale, but nobody can pay anything for them.
I think part of the poor margins has to do with the sequencing of things required to succeed - which you touch on.
So you can’t effectively hire a team to build your product and then layoff the entire team and have sales sell it - then bring back the team that built it to improve it when you really need them - then layoff the entire sales team when you are hitting a nice organic growth stride - then bring back the sales team when organic growth starts to dwindle… etc.
My point being even if everyone is “staying busy” the one crucial thing you need to do next won’t usually involve everyone at any given point in time. So therefore you end up with resources that you needed last quarter or may need next quarter that are getting paid this quarter.
I suppose this is where I could pitch fractional work and agencies - which sure that can be helpful, but I honestly think this is mostly unavoidable. You need a large number of people that are bought in when you are scaling a startup. Unavoidable side effect is a good number of people at any given time are idle on the current “growth blocker.” Oh and not to mention we don’t often actually know what the next unlock is going to be (often is just time). Which means the entire company may just be staying busy waiting for any significant growth to occur.
Yeah, so there's another way to look at all of this, which is basically that software companies undisciplined. They overhire, build a bunch of stuff that people don't really want, do a lot of busywork, and so on. I think that's probably true, at least in terms of how you'd ideally want to run a business. But I think your point is right, which is part of that is because hiring is a ratchet, and if you hire a bunch of people to build the initial thing, you're very unlikely to scale down the team after that. So, you keep building, expand more than you should, and so on.
Which, might be the real impact AI has on all of this. It's less about replacing engineers exactly, but more about making your production capacity more variable. We do the same for our AI bots that we do for cloud compute, and scale teams up and down as needed.
The AI bot comparison to cloud computing thing is really interesting. If that can turn into scaling departments up and down I think it would have a significant impact on the economics for sure.
Ha - I just thought of this. The other terrible thing about all of this is the statistics on what percentage of features people actually use in a product. It’s 20% or something like that based the last stats I read.
So - borrow money to constantly improved the product and keep adding features all of which 80% isn’t going to be used. Yikes. I think modern product analytics and product practices help with this - but still…
Yeah, if anything gets "corrected," I think it'll be that - that software companies spend a lot less money building things that they don't need to build. It feels like a lot of software development is a kind of busy work, that has all the appearances of being productive, but doesn't really mean much to anyone.
Yup - I think that’s right. So inspirational… 😃
Really interesting and thoughtful perspective
Thanks!
You're right that the software customer is buying a TRAJECTORY of technology, not the current features. As the customer's need changes, as compute capabilities change, as user preferences and budgets change, so must the solution evolve or it will die. The software business will incur costs to accommodate product evolution.
To the extent conventional accounting fails to capture it, an investor might assume that some of the costs of future innovation will be recoverable in annual price increases or through breaking out the new features and selling them for additional fees, and then add back what's left to appropriately burden the valuation.
I agree that that's how investors often think about this, but I think the same assumptions are still embedded in that. Like, I think SaaS businesses are actually more like magazines than product companies. Magazines can't release new articles and thing we can sell them for additional fees; they realize the new articles are what people are paying the existing fees for. When software companies sold software in a box, customers expected to buy what was in the box. Now, they expect to buy something that's like a magazine - a steady cadence of new releases. And you can't charge for that; that's the expectation.
Thanks for the summary! Two thoughts came into mind to look out for in those software businesses:
Has the vertical/line of business they serve matured with a low rate of change + what are the switching costs to potentially move to another provider.
Low rate of change of customer's business model + high switching costs -> signal for maturity mode.
One company software company that comes into mind that satisfies both criteria is Veeva Systems ->
GPM 72.42% & OM 21.45% -> sounds like a printing machine.
Yeah, there are definitely companies that can get there; it's not that no software company works. I didn't really think very much about which ones do and don't, and that seems like a reasonable heuristic. If you're selling to a mature industry that wants you to solve a problem that's stable and they don't actually want the constant upgrades, you might be able to get there.
This was an interesting piece (that I just realized is 7 months old).
The part I latched on to: do we buy SaaS for what it will become (at least in part)? I'm not sure. In instances like chatGPT I think that's obviously true. But for your corporate accounting software? I think "what it might become" is a risk. How many pieces of software have you used in your life that eventually made changes bad enough they forced you to migrate? Alternatively, I think of games like World of Warcraft, where literally any changes to the game are hated by half the population.
Regardless, the business of software is certainly more complex than 20 years ago.
Yeah, that's fair. Though in some ways, I think that makes the dynamic even trickier, because:
1.) You sell v1 to customer A, who loves v1 and wants it to be v1 forever.
2.) Competitor keeps developing, markets change, there are new expectations, etc etc.
3.) You pitch customer B and they say, hm, we really want all these new features first.
4.) And now what? B wants v2, A wants v1, and you have to sell to both.
I think you're right, that existing customers might sometimes want a fixed product that never evolves. But new customers want the latest and greatest. And SaaS economics only work if you can sell to both.
Late to the party here but I've often thought about where software development costs show up on the balance sheet. Strictly speaking they *don't.* A manufacturing company like Caterpillar shows "inventory" and "raw materials" on their assets, but software companies don't show the SaaS equivalent – or the equivalent liabilities that you're outlining here like the required future cost of development. Although hilariously if a SaaS company was to be sold, the IP of the software is definitely an "asset" in a vague sense.
I also wonder if this holds in more non-traditional (uh, non-Silicon Valley) SaaS. A friend runs a house remodeling company and they just purchased software specifically designed for contractor project management. It's a bootstrapped SaaS company and seems profitable as hell.
What's also interesting about that is some software companies report on "remaining performance obligations" (like here https://investors.snowflake.com/overview/default.aspx), which are future contractional obligations customers have already committed to, which is essentially the asset side of that liability. Which feels like some real accounting voodoo to be able to do that.
On the "non-traditional companies," I do wonder how profitable a lot of those businesses really are. That model has become appealing to a lot of Silicon Valley types recently, and while I'm sure there are some that make a lot of money, I'd guess most are running roughly breakeven (ie, people make good salaries and all that, but there's not a lot of extra profit for owners/shareholders. But I have no idea.)
Ooh gonna have to dig into that, I hadn't heard of that before. But on first impression - real accounting voodoo haha, especially for companies that sell upmarket and need to customize software. A manufacturer having to customize production lines would definitely have more transparency on the capital investment required and the asset created.
I'd say maybe a famous example of a non-Silicon Valley business succeeding with great profit would be 37Signals (notwithstanding the investment from Jeff Bezos they took many years ago). They've been fairly public that they make millions in profit every year, and have for many years. Mailchimp was another bootstrapped exception to the rule.
Is every non-Silicon Valley that successful? No. And that's where I think – as Cedric @ Commoncog has been writing about recently – the "operational efficiency" of the company matters a lot less than the chosen niche, product strategy, macro environment, etc etc. So yeah, probably broadly agree that many aren't running deeply profitable firms because they don't have as many customers (or the resources to acquire them at a fast clip).
Yeah, 37signals is the example that a lot of people use. My guess is that’s actually somewhat repeatable, but the hard part is in the 15 year slog to get there. Like, if you wanted to build that business, a lot of people probably *could*; they’d just have to be reallllly committed to doing it, and would have to endure a lot of lean years first. And that’s sort of anathema to the entire Silicon Valley culture of moving fast and breaking things.
I used to also think that operational efficiency could have a moment because of the down market, but I’m not sure that’s true. Growth is too much a prisoner’s dilemma - if the market is down and everyone is being operationally efficient, you’re better off being the big spender. And if the market is up and everyone’s a big spender, then you’ve got lots of money to spend so spend it. The only time when there’s a real incentive to be efficient is if you’re playing a different game than everyone else.
Will these margins naturally IMPROVE post-ZIRP, with the collapse/consolidation of B2B SaaS madness?
Here's the argument: B2B software buyers have benefited from an incredible explosion of SaaS products that collectively compete away margin from each other.
Two vectors of competitive pressure:
1. Feature competition
2. Pricing depression
Perhaps it's too easy to build and market a B2B SaaS product now, and too easy to iterate/add features. And SaaS was too cheap.
In 2024, hundreds of SaaS "companies" with "products" will realize that they were only ever a feature anyway... and will need to be consolidated into larger companies, or die.
My theory points to "zero cost money" plus "software is easy to build" as the true culprits for poor margin. We're all witnessing a fundamental restructuring of the SaaS landscape underway right now. The terminal result will be fewer and more expensive products owned by bigger companies, with customers no longer able to simply switch out to the cheaper, more feature-rich upstarts.
I see natural selection happening in SaaS every day. Friends are losing their jobs and companies, and it sucks. But it might be cruel, necessary, and natural selection in a world where money costs money (again).
Bonus take: Will AI make SaaS feature development smarter, faster, more profitable? I'm betting my own company (LeftHook.com) on that belief. More/better integrations for SaaS companies should be cheaper, faster, and less of a headache, and AI will be at the center of making this reality.
Eh, I'm not sure how much competition has to do with it for me. I'm sure that matters, but I think it matters more in that competitive pressure pushes SaaS companies to keep building and having new things sell. I'm sure there's some pricing pressure, but I imagine the bigger issue is that competitors prevents companies from ever getting to that hypothetical "steady state" point where R&D spend can be relatively low, and they can reap the benefits of the high margins.
It's hardware, but my rough sense of the smartphone market is that, if Apple stops making tons of money from selling iPhones, it wouldn't happen because Samsung et al undercut them on price; it's would happen because Apple puts out a bunch of incremental improvements, people feel like iPhones aren't doing anything new and exciting, and they get bored and drift away. I'd imagine that dynamic is kind of the same in software.
Incredible clarity and argument, Benn. Love it!
What you seem to be arguing implicitly is that there has been and continues to be a big arbitrage because Wall Street bankers continues to hold a (mostly) false assumption that "steady-state condition" is attainable for SaaS companies. This belief drives the underwriting of exists for VCs.
The other takeaway is that VCs don't care if their SaaS companies turn a profit, it just has to grow revenue at a high enough clip to exit. And maybe bankers don't (truly) care whether the IPO they underwrite will turn a profit either, because if a good eventual "steady-state condition" story is told and sold, other institutional investors will buy the shares anyways.
In other words, the bubble continues to grow.
Yeah, I think that's basically the idea, though I don't think it's necessarily an arbitrage. Like, if everyone believes SaaS businesses are worth a lot, then they're worth a lot. There's no rule that says they have to return to some "steady-state condition" based on how much profit they make for shareholders. Like, the price of gold has been a "bubble" for all of human history - people value it because they assume other people will continue to value it in the future.