A useful step in any fundamental research process is to look not just at the returns that a company is getting but at the returns its customers are getting. If you're considering two software businesses with similar growth and margin profiles, a good tiebreaker is to choose the one that plausibly gives its customers the highest return on investment. That implies pricing power, means that those customers are more likely to stick around, and is a bit better for the world, too. But you can extend this question arbitrarily: are those customers selling something valuable to the world, or is the product in question something socially harmful, like a good tool casinos can use to identify customers who have a tendency towards gambling addiction in order to optimize their marketing? That's probably great for ROI one hop down the value chain, i.e. at the level of the software buyer. And it's terrible for ROI one hop past that. On the other side of this spectrum, you might have someone like Veeva whose software helps biopharmaceutical companies develop and bring drugs to market faster. Now, the jury is out on the pharma industry as a whole, but if Veeva’s study startup product helped AstraZeneca launch their lung cancer trial 6 months faster than status quo, that means 6 more months that their lung cancer drug (say Tagrisso) is under patent earning monopoly profits (very good financial ROI). But more importantly, a patient who may have otherwise died is now likely cured. And maybe they are a single mother whose child would have otherwise grown up alone. It’s positive externalities all the way down.

This kind of question can lead to infinite regress, where you start out asking which stocks to include in your basket of low price/book small-cap banks and end up asking questions about the meaning of life. In practice, you just have to assume at some point that for the economy to function at all, the average benefit of a transaction between two economic agents is positive, and to zero in on cases where you can get useful information. Basically, as long as there isn't significant concentration somewhere in the supply chain, one hop in either direction—i.e asking "are this company's suppliers in a sustainable position, and are there customers in such a position, too?"—is about as far as you can go.

This is a pretty broad philosophical point, but it has practical implications. A lot of the collateral for consumer credit in the US consists of single-family homes, and any given homebuyer will have wildly divergent prices they'd pay for these homes. And yet, a system where the typical homebuyer starts out levered about 4:1, against an asset with indeterminate fundamental value, whose yield primarily takes the form of consumption and thus has to be imputed to even approximate price levels, actually works pretty well!

In plenty of consumer Internet businesses, executives start to feel guilty after a while about just how many human lifetimes have been spent idly using their products. Aaron Swartz reflected on this:

But you can flip this around: economic growth is a miraculous thing, which has produced enough material goods to sustain over eight billion people, many of whom are so fantastically wealthy, in historical terms, that one of the problems they face in life is figuring out what to do with their gobs of idle time. Multiply eight billion by even a tiny amount of leisure time and you get plenty of lifetimes worth of time spent on nothing in particular. And if that's just a fact about the world, you can at least commit to giving them the highest-quality leisure time they could get. Some people will be slightly addicted, some will use the service more than they'd really prefer to, and some of them will meet their future spouse, finally get clean from whatever they're addicted to, learn something useful for work, or learn something that's just intrinsically interesting.

In B2B software, one place this shows up is that some unknown fraction of SaaS purchases are either massively underused or actually never used. Companies go through the entire ritual of assessing multiple products, negotiating a contract, paying for the product, etc.—and never manage to turn it on and make it do something useful for them. Or, they buy it and find out that, for reasons they hadn't previously considered, it's slightly worse than the old way of doing things. If you've only worked at one such company, and you look at your metrics and realize that some customer has, in effect, paid five figures per hour for using your software, you'll probably panic and realize that the entire business is a house of cards. But over time, it'll become clear that this is just a fact of life; the customer who was unlikely to buy a product will probably make their purchasing decision at the peak of their potential enthusiasm, and will statistically be a little less excited once it's actually time to use it.¹ It's gotten to the point where "short companies that are switching ERP providers" is financial folk wisdom, and that approach has worked in the past, like with Lamb-Weston last year. There’s also a way in which enterprise software purchasing decisions mirror buying an insurance policy. A CIO would rather buy everything that employees could possibly need than be mocked when the platform they purchased couldn’t do something that some business line absolutely needed. This naturally has a side effect of some modules remaining chronically underutilized.²

Among companies whose product is designed to improve business outcomes, there will be a range of returns, and the way things work out is that the customers who are getting a bad deal will churn out (either because they realize they made a mistake or because they make so many other mistakes that they go bust) while the customers who are getting a return will grow, and their software spend will grow with them.³

Everyone in the publishing business has to come to terms with the fact that many of the best books will be bought-but-unread, and that the readers who really are getting to the last page routinely are much more likely to be reading low-effort genre fiction than modern masterpieces. And that's for roughly the same reason as the unused-SaaS example above—if your excitement about reading Ulysses fluctuates around "I'll probably get around to it some day," the day you're inspired enough to buy a copy is likely to be the most inspired you ever were to read it.

In some cases, the downstream risk is opaque until it really hits people hard. You'd think that fast food, snack foods, cigarettes, and alcohol were all different enough from one another, and had a sufficiently fragmented customer base, that thinking hard about the downstream supply chain question wouldn't matter much. Yes, you can get fat, or get cancer, but that's been well-known for a while and people keep buying. It turned out that these sectors were all at least partly a bet against GLP-1s. Some money was made by people who figured this out early, and some money was lost by those who didn't, but "early" did not mean reading papers about potential uses of gila monster venom, but looking at how fast sales were rising for these drugs instead.

This is just the flipside of abundance. The world produces enough stuff, and the production of that stuff creates enough purchasing power, that the optimal setup for the economy involves a fair amount of waste. There are domains where that waste is costly, and ones where it isn't—unimaginable amounts of money and talent have been invested in ensuring that you can get next- or same-day delivery for some product you may never end up using. But if the aggregates hold up, and the immediate supply chain is intact, there's a thin boundary between good investment research and asking philosophical questions that are rewarding on their own but unlikely to produce much alpha.

Figuring out how much to understand supply chains is a surprisingly common part of the Diff beat, including:

• 100 Due Diligence questions

• A look at ServiceTitan’s S-1 ($).

• Instacart, and how grocery stores cross-sell ads.

• The supply-chain split between full-stack chip companies and foundries ($).

Join the discussion!

1  This is one reason that, if you get excited about some form of self-transformation, whether it's getting fit or getting good at math, you should absolutely not make any purchases reflecting that upfront. Don't buy a gym membership; start jogging or doing pushups or some other 100%-free-of-charge kind of fitness first. If you stay enthusiastic, you'll end up signing up for the gym or buying the kettlebell soon enough, but if that feeling wanes you'll have a regular bill or a physical artifact to mock you the next time you're tempted to make the effort.

2  Steve Ballmer talked about this in a recent episode of Acquired: “I want to make sure I bought everything. I don’t want to look bad because either I paid too much or I have holes in what I bought for people. When you sell the enterprise, you have to provide peace of mind, which is like an insurance policy. Buying more than you might be using or some users are using. It’s an insurance policy.”

3  This is one reason that sales-facing software companies can have such insane growth curves. If they make the difference between a sales process that almost works and one that really works, their customers start adding more salespeople, which means buying more seats, which means faster growth. And since sales jobs have naturally high turnover—the underperforming reps get fired and some of the best performers get poached—brand recognition diffuses throughout the industry and soon enough new logos start popping up.