What Salesforce is Selling (and Buying)
Last week, I wrote about IBM, which was the center of the computing industry for decades because it mastered distribution. For most of my lifetime, "Who is the next IBM?" has meant "Who is about to be crushed by a shift in where value gets captured?" but for a long time before that, "the next IBM" was what growth investors were praying they'd find. The IBM model during its peak was:
- Have an excellent sales force.
- Ensure that most revenue is recurring, and ideally expands with usage.
- Develop a comparative advantage in finding new products to feed the voracious sales machine.
Palantir (written up in The Diff here) sounds a bit like this, but the company is still in its rapid growth stage. And Palantir’s distinctive culture also makes it hard to compare to other companies—although IBM's old culture, with its corporate songs, might come close.
Salesforce, on the other hand, fits all three. Just like the name says, it has an excellent sales team. Its products produce recurring revenue, and opportunities for expansion. And Salesforce has developed a model for finding new products and improvements on existing ones, albeit different from IBM's: Salesforce can grow by building new products, by promoting third-party integrations, and, increasingly importantly, by acquiring companies whose weaknesses are complemented by Salesforce's strengths.
It would be reductionist to call Salesforce a purely sales-driven company; their CEO heir-apparent co-created Google Maps, founded Friendfeed, and was the CTO of Facebook. But it's undeniable that Salesforce has been able to grow in part by acquiring good products, integrating them together, and ramping up distribution. As larger companies use more individual pieces of software, plugging them into each other turns into a bigger deal—both Dropbox and Box, for example, tout their third-party integrations as a key selling point. Salesforce has been able to systematize this process, which puts them in a position to acquire companies that are better at product and technology than they are at sales, and then make them, by default, good at all three.
Salesforce's founding story went from auspicious to inauspicious over the course of its first two years: the company was founded by a top Oracle salesperson in 1999, and was able to launch its first CRM product just in time for the peak of the dot-com boom; in Behind the Cloud, Benioff recalls that the development team was able to build a prototype "within a month." This timing gave the company an exceptionally rough start; by 2001, it was bleeding money, and survived partly by offering steep discounts for annual rather than monthly contracts. By 2004, Salesforce was on its feet and able to list on the NYSE.
The company did not have an easy time pitching itself to investors. In 2004, there were two familiar kinds of enterprise software companies: mature, low-growth, high-margin businesses that had been around for decades and were throwing off cash flow, and dead or dying B2B businesses that had gone public in the late 90s. (The B2B bubble crashed slightly later than the NASDAQ as a whole; for a few months, there was a narrative that prospecting for gold during a gold rush was a low expected-value endeavour, but that all the irrational dot-coms would spend some of their money on the picks and shovels of software and services, making the businesses that sold to them viable. This did not pan out.)
When Salesforce pitched itself to investors, it had to make two separate claims:
- If they were going to compare it to Microsoft and Oracle, they'd need to compare it to Microsoft and Oracle at the same age. Salesforce was actually growing faster than those companies (and is still impressive compared to the typical company at its age).
- Because of the stage of the business, they'd have to look past immediate profit-and-loss metrics, and look at how much it cost the company to land a customer and what unit profit looked like after that.
These points are obvious today, but what's non-obvious is that Salesforce played a major role in convincing investors to think about subscription software this way. The company's stock trades at $216 today, compared to a split-adjusted price around the IPO of $4; all the skeptics of SaaS math either changed their minds or got fired.
Salesforce's CFO eventually figured out the problem: most of the analysts looking at the company hadn't been around when Microsoft and Oracle went public; they had no idea what a good company at that stage looked like. They had been around for the rash of late-90s IPOs, so they knew what a cash-burning company with big promises sounded like. Salesforce could point to stickier long-term economics, but they had to deal with skeptics for a long time.1 It was a long generation problem: the boom and bust cycle created a large population that had learned the wrong lessons.
Today, investors do understand subscription models well. Salesforce may have created more aggregate market cap for the rest of the industry than it captured for itself, first by giving investors the right vocabulary and later by giving them a compelling comp.
But the Salesforce model is not just sales/marketing expense in, customer lifetime value out. That's a big part of it, sure. One of the funny but revealing lines in Marc Benioff's book, written in 2009, is: "We quickly discovered that the more salespeople we hired, the more we saw revenue increase." That is, indeed, part of the formula. The company also talks a lot about unit economics a lot (previously: mid-30s unit economic margins; soon, 40% or so). But that's a high-level abstraction that papers over what they're selling, and to whom. A company that reports 90% penetration of the Fortune 500 has limits to how much it can grow merely by landing new customers.
Not strict limits, of course; Salesforce likes to point out that while dealing with smaller businesses is a challenge because of higher churn, it has venture-like characteristics. Of their 2015 cohort of small business customers, 80% have churned off. But the survivors increased their spending by 19x, so overall revenue growth from that cohort is in the mid-30s. A company that relies on Fortune 500 companies alone is always hoping that the list won't change too much; Salesforce tries to hedge by betting on the companies that will join.
The company can also manage towards a higher retention number. Salesforce's plugin directory, AppExchange, is retention layered on top of retention: As soon as SurveyMonkey, Twilio, or Mailchimp has an integration with Salesforce, that partner is indirectly trying to keep Salesforce's churn low and usage high. They can also use training for lock-in. Salesforce goes out of its way to make this easy, by providing training for different kinds of users. (In an indirect way, Salesforce has accidentally cloned a subset of Lambda School: it aligns incentives to teach people useful skills, verify their credentials, and charge them nothing upfront if the educator can collect a portion of the economic upside. Salesforce just does it through more seats sold and higher retention.)
But still, there's a limit to how many CRM seats one company can sell. Salesforce, naturally, has a solution to that: they've gotten in the habit of launching new products, and, increasingly, of acquiring companies—Demandware ($3bn, 2016), Mulesoft ($7bn, 2018), Tableau ($16bn, 2019), Slack ($28bn, 2020). Each acquisition gives them a bit more to cross-sell and more to integrate—or, as they sometimes discover, more vendors to consolidate. (After the Slack acquisition, Slack learned that one of their big customers was using Slack as a front-end for answering customer questions that were coming in through Salesforce's Service Cloud. It's an M&A synergy version of the Piña Colada song.)
At their most recent investor day, the company's CFO talked about how no C-level executive dreams of dealing with more vendors.2 When vendor consolidation happens, it's often top-down: someone will run through a list of the hundreds of software providers that a company subscribes to, and figure out which ones are actually still in use. (Just as most people open only a handful of the apps they install unless prompted, it's possible to buy a product—with your company's money—and then forget to actually use it after a while.) But Salesforce is driving consolidation in another direction, by acquiring more adjacent products and then pitching them as one bundle. Since customer acquisition costs are so dominant in subscription software economics, this is an incredibly efficient way to use capital; one more five-, six-, or even seven-figure sale for the same number of plane tickets, hotel room-nights, or (more recently, but somewhat temporarily) Zoom calls.
It's somewhat redundant for there to be so many enterprise sales teams out there, all chasing meetings with the same handful of decisionmakers at big companies. And, once they have the meeting, it's a further inefficiency to find out that the company's pain points are better served by a different product. If the problem is managing customer relationships, a CRM solution is helpful. If the problem is being overwhelmed by data, and being bottlenecked by a limited supply of people who either know enough R or D3 to produce meaningful graphs, or who know enough Excel to know their limits, then the right product to sell is Tableau. If the company's big issue is keeping everyone up-to-date without an avalanche of email, Slack is the right offering. A salesperson's job is some mix of listening and talking, and the more solutions they can sell, the more listening they can do. Since they know less about the customer than the customer does, that tends to be a good deal for both parties.
Bret Taylor talks about Salesforce's suite as the "single source of truth." This is a big-company cliche, but for a good reason; C-level Fortune 50 executives spend more time worrying about the nature of truth than freshman philosophy majors. Accounting, which looks so precise and scientific at the level of income statements, is the convergence of a long series of assumptions, many of which are audited imperfectly and rarely stress-tested. (Steve Cakebread, Salesforce's former CFO, talks about the audit process from his first accounting job, at Hewlett-Packard, part of which involved wandering around the factory floor and looking at which bins of parts were full or empty. Hewlett-Packard reported a precise number for its inventory every quarter, but that number rested on assumptions about usage, waste, obsolescence, the frequency of purchasing mistakes, and the company's ability to keep track of all of the above.) It's exceptionally hard to know what's going on at a company that employs tens of thousands of people and earns tens of billions of dollars in revenue; a CEO can talk to direct reports, but they're all promoted based on some combination of a) doing a good job, and b) appearing to do a good job, and the mix of A and B is knowable only in retrospect. Having a systematic way to keep track of everything, and to frequently visualize it, is inordinately valuable in a large and complicated company, and that's what Salesforce is selling, in many different ways.
It's a natural human weakness to 1) overestimate the importance of what you're good at, and 2) underestimate the difficulty of things other people do. So some companies get started by people who love building useful products and writing elegant code, and who assume that future customers will hear about them, somehow, and be blown away by their feature list and immediately sign up. And other companies start with a few great salespeople who are reasonably sure that a working product can be put together in a weekend, at most. For the former case, Salesforce is a natural acquirer; it has done the hard part of getting at least one invoice paid by an enormous company; step two of relentlessly upselling and cross-selling them is also a Salesforce strength. Building new products in-house is certainly part of the Salesforce strategy—as a BATNA for deals, if nothing else—but it's much more cost-effective for them to acquire companies that are very good at everything they do except selling.
Further reading: Behind the Cloud is a good early history of Salesforce. The IPO Playbook is a very good guide to taking a company public, written by Steve Cakebread, who was Salesforce's CFO when the company IPOed.
A Word From Our Sponsors
Diff readers want to know everything worth reading in finance and technology. But some of us want to know everything that’s worth reading about everything. The best place to start is The Browser, a curated newsletter of five fascinating stories each day on every topic imaginable, from how Aztecs thought about pandemics to how novelists write from the perspective of chickens. I’m a regular reader, and The Browser has featured a few Diff pieces in the past. I encourage you to sign up here—Diff readers get 20% off your first year with the code DIFF20 at checkout.
Multilateral Vaccine Diplomacy
China, as Tyler Cowen once noted, has the financial and research resources of a rich country, but the risk tolerance of a poorer country. This, along with China's vigorous response to the initial pandemic, has given them the opportunity to distribute its vaccine according to different priorities than most places; it would be reasonable, for example, to think that the US, rather than China, would be the country sending doses to the US's geographic neighbor and close trade partner. But this is changing, in part because the US collectively with its allies has the resources of a very rich country, and any effort to vaccinate everyone in the country will lead to plenty of spare capacity that can be redirected elsewhere: America and Japan are funding 1bn doses, to be made in India and distributed by Australia ($, FT).
In related news, The Economist had a great profile of India's Serum Institute last week.
State Capacity Straussianism
Tyler Cowen argues that vaccine passports should be created, but not legally enforced. His basic argument is that the existence of a vaccine passport encourages otherwise-reluctant people to get shots, but probably couldn't be effectively implemented. This kind of policy approach—announce something that won't work, but whose existence alters behavior in a way that improves outcomes—can be effective for a short time. But every attempt to do it makes the next one less effective, because people slowly acclimate to the idea that these projects are ineffective. The effectiveness "Straussian State Capacity" is a function of perceived state capacity, which itself is based on actually getting things done.
Jordan Schneider has a great interview with Adam Tooze and Matt Klein, mostly on China and analogues to earlier history. A particularly interesting point from Adam:
So I think we see China playing a rather skilled game in the UN, which is an unweighted global body where they can assemble coalitions which are quite powerful. BRI is part of that.
That was why the UN was increasingly uncomfortable for America from the 1960s onwards. As soon as you start multiplying post-colonial States, many of them are actually aggressively anti-American in their positioning, whereas the IMF and institutions like that are shareholder weighted.
It's very difficult for an up and coming power like China to make much headway in the IMF, because they have to overcome an incumbent American veto that just can't be moved. Whereas in the U.N., the Americans are so sleepy and ideological that they don't even appear to play the committee game with real energy. In the IMF they certainly do.
This is a good way to divide up global institutions. As a rule, if those institutions weight countries' influence by some measure of size and importance, they'll tend to cater to the interests of established players; if they're unweighted, they cater to the interests of whoever is mostly willing and able to game the system, which usually means rising players. Looking at Russian and Chinese policy in Africa in the 60s, and China's policies there today, as a way to pack the UN might be extreme, but that is part of what's going on. It's sometimes quite affordable.
Chip Insourcing, Slowly
The center of gravity for the chip industry has slowly converged on Taiwan and South Korea (see my earlier writeups on TSM ($) and Samsung ($)). Other countries are trying to move it back, or spread it out, with the EU and US discussing subsidies for local production. Meanwhile, though, Dutch chip equipment manufacturer ASML is hiring more workers in Taiwan ($, Nikkei)) to get closer to its biggest sources of demand. And the Texas power outages have cost local chip companies, including Samsung, an unknown but large amount of money, and might put Samsung's future expansion plans there at risk. The short useful life and delicate equipment of chip fabs makes reliable power especially important—if it takes weeks to restart after an outage, that's weeks of depreciation in the value of billions of dollars in capital equipment.
Meanwhile, smartphone maker Oppo says that chip shortages are the new normal ($, Nikkei) due to rising demand. It's possible that shortages will continue in the future, but supply, rather than demand, seems like a bigger driver: the chip industry has gone through long periods of rising consumption, but as the capital cost of individual fabs rises and the number of companies capable of making cutting-edge foundries shrinks, production changes increasingly follow a step-function pattern.
Password Price Discrimination
For a company that is obsessed with tracking user-level data, and that has to be keenly aware of the ins and outs of data transmission because it's one of the single largest bandwidth consumers, Netflix has always been oddly unwilling to enforce its one-household-per-account rule. The reason, of course, is a combination of marketing and price discrimination: there are people who are individually willing to pay a bit less than what Netflix charges, but who wouldn't bother at full price. And there are people who convert from free subscribers to paid subscribers when their ex changes a password. Netflix is now discouraging password sharing. It's still possible to share voluntarily, albeit with some friction, but this will reduce some free-riding. And while the effects will show up in Netflix's subscriber count, it's really a measure of the company's pricing power. The stated price of Netflix was always a ceiling, and the number of paying subscribers was a floor on the number of active viewers. Now those numbers are converging with reality.
The record of companies that pioneer new accounting metrics is mixed. Some conglomerates have special definitions of profits they like to cite, like Roper's cash return on investment and Berkshire's look-through earnings. And John Malone did an altogether too-good job of selling investors on the merits of EBITDA, which makes sense for a business that has high depreciation relative to incremental capital expenditure needs, but often gets used to make businesses with ongoing capex needs look cheap. The "Novel Non-GAAP Metric" portfolio would have some spectacular winners, but plenty of messy eyeball-related losers, too. ↩
He name-drops the World Economic Forum as the venue for some of these discussions, which is a clever way to say "I assume you, too, have been to Davos." ↩