A quick blueprint for a successful company:

1. Start with a team of talented perfectionists, ideally ones who prefer non-verbal communication.

2. Work on a task that’s either impossible or so vague that it might as well be.

3. Once it’s apparent that you will never achieve your original goal, start selling the internal software you built to manage it.

This is what happened with Yammer. Squint a little, and it’s the Palantir story — tracking down Bin Laden turns out to be more feasible than ensuring that any given email address represents a trusted financial counterparty. Tobias Lütke just wants to sell snowboards. And, famousy, Stewart Butterfield did it twice, with Flickr and then Slack.[1] Flickr was a solid exit, at least at the time. And Slack is now publicly traded, under the ticker WORK.

This is a good pattern, for obvious reasons. Recruiting people to pursue an amazing consumer-facing vision is fun. Recruiting people to build digital plumbing — we’re gonna change the way middle-managers interrupt each others’ workdays! — is not. Plumbing pays well. (A literal plumber will get a far better IRR on their apprenticeship than the median high school student will on attending college.) But it’s un-glamorous. The trick is to make it the most glamorous alternative to failure.

The Company

Slack’s origin story is a microcosm for their entire model: their marketing strategy is to pitch the product as a consumer app and then sell it as enterprise software. The original business was a cooperative game called Glitch, which was meant to involve exploring and building things collaboratively, rather than combat. This is a hard project: on the Internet, you have as many lives as you need, so games are relatively more fun when they simulate extremely fatal activities such as shooting, getting shot at, blowing things up, getting blown up, etc., after which your character comes back to life and does it again.

And the iron law of Internet communities is that, however nice they are at the beginning, eventually they will attract trolls, who will take excessive effort to defeat. All the building/simulation games end up:

A) Being run in small, non-scalable communities (the nice Minecraft servers),

B) Being dominated by a single-player experience (Factorio, various simulation titles), or

C) Dominated by griefers (the other Minecraft servers, Habbo Hotel).

I’m not sure the Griefer Problem would have ruined Glitch, but it’s an open-ended, hard-to-solve one.

A nearer-term problem is that of building a simulation that’s fun, comprehensible, and not prone to "Munchkining," the situation where a character uses his Scroll of Tripling on his Scroll of Tripling twenty times until he has over a million Scrolls of Tripling, which he uses to create an infinite supply of everything. Any sufficiently complex game becomes Turing Complete, which makes it prone to Munchkining, and that makes the game boring for everyone else.

Open-ended sandbox-style games are hard. Earth has had a couple billion years to achieve a somewhat balanced system, but even so various players — cyanobacteria, H. Sap, etc. — end up munchkining the system and basically ruining it for all the other participants.

While they were building this complex, ever-expanding, combinatorially hard project, the Glitch team needed a tool to manage internal communications, so they built one, called Slack. You may have heard of it.

Glitch did eventually launch, then temporarily un-launched, then ended permanently. Slack, of course is still going.

So far, it’s working quite well. Slack’s billings were up 79% in 2018, and paying customers grew 49%. Every SaaS analyst’s favorite metric is dollar retention — take the amount your customers spent last year, compare it to how much only those customers spent this year, including the ones who stopped using your product. Slack’s dollar retention was 143% in 2018 — if they hadn’t closed a single new account, they’d still be a monster growth story.

Analysts love that number because it’s a metric that underpins a long growth story: even if closing today’s new customers is expensive, today’s small customers are tomorrow’s huge customers. Of course, their dollar retention is declining, as it does for many companies — and it benefits from the fact that Slack’s early customers are themselves fast-growing startups.

The Slack story is about compound interest on top of compound interest: they close new customers, their existing customers grow headcount, and they grow their penetration of existing customers’ businesses. Even if dollar retention slows, that’s not necessarily going to lead to multiple compression: if 100% of the people at a given company use Slack, it’s that much harder for a competitor to rip it out and replace it with something else. And this means that as Slack loses organic user growth, it starts to get more pricing power).

This is pretty straightforward to model. You can make some reasonable assumptions about the incremental cost of customer acquisition, the cadence of the decline in dollar retention, the eventual steady state of the business, etc. I’m sure a sensible person can justify today’s market price, or a price half that or 2x, without making heroically unreasonable guesses.

With a business like that, where all the components of the revenue build are pointing in the same direction (up, up up!), the main thing to worry about is competition. Chat existed before Slack; if their headquarters gets struck by a meteorite, it will go on existing without them. I’ve written before about the business of enterprise chat, and you may notice that the most valuable chat product in the world doesn’t sell itself as chat.

If you’re editing code, maybe Github should have a chat window where you can highlight a line, see who last edited it, and send them a quick message. If you’re making sure Alice’s Excel model corresponds to Bob’s Powerpoint, that’s another good use case for chat. Your Cisco phone handles synchronous 1:1 and 1:many and many:many conversations all the time; maybe a text layer on top of it would be a good thing. A CRM system also sounds like something that would be a better, stickier product with chat attached.

As it happens, Slack integrates with many of these products. They’re clearly thinking ahead. But still, it’s a dicey position to be in; Slack’s optimal move is to integrate where they can and clone where they can’t; a Slack office suite would be expensive, but not a bad idea. Or at least, not a bad BATNA to promote future integrations with large providers of office software, from Redmond to… Redmond.

Slack Time and Work Time

Slack, the product, is part of a paradoxical side effect of smartphone ubiquity: we’re never fully away from work, but we’re never fully at work, either. Andrew Taggart has written about the phenomenon of "total work," in which all human activity becomes subordinate to, and takes the form of, work. Zvi Mowshowitz has written about the near-opposite of total work: "Slack: The absence of binding constraints on behavior."

Slack, the product, and WORK, the company, embody the weird modern dialectic: work is relentlessly invading leisure, but leisure is also sneaking its way into work.

Before cell phones, you could be reasonably sure that when you left work, you were done until the next morning. You could bring some work home with you, but that was a conscious decision, and not necessarily a default. But when you were at work, you were actually at work. It would be a specific choice to show up at work with a book or a newspaper and read at your desk.

In the pre-digital days, there was less of a return on motivation in white collar work: the office was so boring that working was probably the most interesting thing you could do. That’s all over now: every one of us takes to work a small computer offering a cornucopia of socializing, news, games, and porn; many of us leave notifications on, allowing constant Harrison Bergeron-style interruptions. And, when we leave, we bring the notifications with us. Slack technically lets you turn off alerts after work hours, but my guess is that the first time you wake up to a bunch of frantic late-night messages from your boss, you switch to having alerts on during all of your waking hours.

Even within the product, Slack blurs the distinction between leisure and work. One of the most fun things humans do, which can absorb an arbitrary amount of our time, is gossip. If you’re at work — and working long hours, because you’re not fully engaged with the tasks at hand, or merely because it’s socially unacceptable to leave early — your social circle is dominated by work friends.

Because it happens in the same interface as work conversations, gossip looks just like work. In fact, it might feel just like work — one minute, you’re discussing how long it will take to get a new feature through QA, thirty minutes later you’re trading funny stories about That One QA Guy.

Not all gossip is one-to-one. Slack allows users to create private groups. This is natural: some people at the office eat lunch together, or socialize after work. But what’s not natural is that one of Slack’s features is indefinite text search. And since the most interesting gossip topic is often whoever’s the most recent person to join an organization, Slack-cliques necessarily exclude that person from ever joining. You could go back and delete all your incriminating conversations from their first week at work, but it’s easier to just never add them to the group. So organizations that use Slack heavily tend to have more literal social strata: layers rendered impermeable by text search.[2]

Michael Pollan sometimes writes about "edible food-like substances," and an important modern skill is to avoid time-consuming work-like activities. This is not at all Slack’s fault; before Slack, managers texted; before that, they emailed, BBMed, or even picked up the phone. But Slack is a gorgeous product, and on a modern device it’s reasonably fast; Slack makes things easier.

So far, white collar work has survived this blurring. But it’s putting unprecedented demands on everyone’s executive function.

In their S-1, Slack says their focus is "realizing the enormity and simplicity of Slack’s singular mission: to make people’s working lives simpler, more pleasant, and more productive." Enormity refers to badness, not scale. The word Slack is going for is "bigliness." But it’s a revealing error. Work is not necessarily simple (or it wouldn’t pay well) and it’s not necessarily pleasant (again, that’s what the money is for). Whenever you start to improve one of the fundamental traits of an institution, it’s important to ask what you’re trading off against: maybe the simplicity and pleasantness enabled by instant-messaging is more pathological than it looks.

Slack and the Law

In their S-1, Slack talked about legal risk, and it’s a big one. There are two kinds of legal risk the company faces:

1. Criminals

2. Law Enforcement

The risk from criminals is obvious: someone could hack into Slack, and read a company’s corporate documents. They might find personally identifiable information on customers, or financial documents that could give them access to a company’s bank accounts, not to mention source code, strategic plans, M&A discussion, drafts of press releases — the amount of valuable IP flowing through Slack is staggering.

Of course, they’re aware of this. And their users are aware, too; at some point in every company’s lifecycle, their legal/compliance/security team gives everyone a big "Please, do not be a colossal idiot" lecture.

Still, it’s a risk — a meaningful one, as Keybase’s CEO discovered.

The other legal risk to Slack users is from a culture clash; Slack is a written medium, but it’s closer to the spoken word. Some people write paragraphs into Slack, some of the time, but a lot of messages are basically a substitute for a phone call.

That’s a colossal problem, because you can’t trivially search the corpus of everything somebody has said on the phone, but you can extract that text from Slack, pretty easily. And since it’s verbal, nobody has the background process running in their brains: "How would this sound if read aloud to a jury?" (Oh, you included a wink emoji — does it sound better if a lawyer reads it aloud and adds, completely deadpan, "wink emoji.")

There’s a saying among lawyers in regulated industries: the "e" in "email" stands for "evidence." When you’re emailing, don’t joke about breaking laws, stiffing suppliers, lying to customers, skirting regulations — eventually, long after you’ve forgotten what you said, you’ll hear it again, and it won’t be very funny. We don’t have a good pithy saying for practicing good IM hygiene, but how about this: you can’t spell "most professionals lack the skill of never saying anything in an IM that wouldn’t sound really bad later" without Slack.

The natural response to this is to say that nobody should fear transparency: if you don’t want people to hear about it, maybe you shouldn’t say it. But consider this: have you ever assessed something, decided it’s worth the risks, done it, and then had those risks materialize? Even if you had a positive expected value from your actions, there’s an easy negative spin: "They knew this could happen — and they did it anyway." Henry Blodget got banned for life from his chosen profession, not because he was too optimistic about dot-com stocks, but because he started to feel misgivings and wrote about them in a legally discoverable medium. If he’d been less prudent, he might still be publishing equity research to this day.

This has always been true: nuance looks like hypocrisy to anyone motivated to see it that way. But now, more nuance is expressed in a bland, inflection-free medium, which can be read for years after the fact.

The searchability of text is not, on the whole, a bad thing. It is in fact highly convenient to rewind conversations that happened a month ago, or a year ago, to remember the thought process behind decisions. In fact, it lends itself to some cool analytics — you can probably date the peak of any organization’s productivity by looking at the ratio of ``` (the Slack convention for coding a block of code) to /giphy, the Slack command for posting a reaction gif.

But as smokers, cheaters, liars, loan officers, and insurance underwriters all learn to their sorrow, it’s prudent to pay attention to risks you’re accumulating, so you’re prepared when they come home to roost.

Slack is a bit like Adderall. Ask someone about it the week after they start using it, and they can’t praise it enough. Talk to them six months later, they’re back to baseline but now they have an addiction. And talk to them years after the fact, and they’re just regretful.

Thanks to Alexey Guzey and Jonathan Libov for edits.

Don’t miss the next story. Sign up for my occasional email newsletter. Or check the About/FAQ page for more.

[1] This doesn’t just describe software companies, either. One of the pioneering works of non-Euclidean geometry was called Euclid Freed From Every Flaw, the Catholic metal band Reverorum Ib Malacht was originally a Satanic metal band whose members decided to study theology to write more blasphemous lyrics, and Richard Feynman’s Nobel Prize-winning work was the result of his decision to embrace burnout and work on fun problems rather than solve anything important.

[2] I’m basing a little of this on my personal experience, and a lot more on the various spurts of Internet Media drama that arise from people at online media companies leaking damaging Slack transcripts.