…The funny thing is it’s as though they are playing a game of Sardines, all trying to crowd into the corner of the room labeled “strategy”. But the other corners are just as necessary and affect each other in interesting ways.
Tech companies function much like the Roman Republic. Your influence is more or less proportional to your equity stake, which is itself proportional to how early you arrived. The Republic assigned voting power and civic privileges proportional to one’s wealth and pedigree. The bulk of voice and power went to the gentry, somewhat less to smallholders, and so on. At the bottom were those citizens with nothing of value but their labor. They were called the “capite censi”, literally the “headcount”.
Philosopher-gurus are the exception. They sit to the side of the main social structure, but within whispering distance of the top. Half of their value comes from this otherness, enabling aristocrats to argue ideas outside of the rules of the hierarchy. A wink from the ruler is as good as a nod, and it’s tiresome to have your underlings scramble their plans, get spooked, spread rumors, or generally misunderstand when you are simply working through an idea. Talking with a guru is a holiday from intrigue.
The guru only listens and responds. He doesn’t scheme because he has nothing to scheme with. This allows him to challenge & question to a degree that would be unwise from others. Only the music master can interrupt the king’s playing. Only the fencing master can attack him with a sword. This part of the guru service is appropriately called “executive sparring”.
The other part is developing systems of thought focused on the unique problems of the age they are in. Every Chinese philosopher had hidden in his robes the one true way to unify the state. Every Greek, the way to rationalize the world. There are lots of hard problems tech execs face today: how to stockpile talent in a way that doesn’t cost too much, cornering the market on useful data flows, dealing with nation-states trying to hack them, or managing sprawling empires with relatively few people. But whether they know it or not, the fundamental problem facing today’s tech exec is the problem of governance.
So the guru says…
“Go ahead and eat the world, but learn to chew before you swallow.”
Corporations do not want to legislate, let alone enforce, more of humanity’s behavior than they need to. They have better things to do. The universe isn’t going to dent itself. But because they are outpacing traditional governments in their ability to predict & shape human behavior, they walk right into stupidly impossible situations over and over again. Software isn’t eating the world so much as larping it.
The governance problem is not about dealing with governments as such, but about accidentally adopting or creating arenas of human behavior that you then have to govern, with “externalities” that don’t remain external. Starting a ride-sharing service? Congratulations, you’re now an urban planner. Want to disrupt shipping logistics? Get ready to do your part to curb terrorism and slavery.
The problem of governance arises wherever “move fast and break things” runs right into an older saying: “you broke it, you bought it”.
Sometimes things blow up squarely on your turf. Think about what it takes to enforce a social network “real names” policy across hundreds of cultures. One account per person, one person per account, authentic names as used in real life. Zuck’s buddy Calvin can post pictures of himself holding pounds of weed but he can’t use the name he’s known by, Snoop Dogg.
The rule was relaxed a bit after the Real Names Wars of 2014.
To police the policy you rely on automation and crowdsourcing. That sets the stage for an instant ideological slapfight, with brigades of trolls disabling accounts of people they don’t like by reporting their names as fake to a dumb program that doesn’t know any better. Anti-spam tools subverted to enable harassment. Nobody is happy and everyone blames you.
And then there is fake news. From a quick glance at Facebook posts about him, it seems that ol’ Snoop is simultaneously dead, alive, sober, high, recovering from a heart attack, fleeing the country, and under arrest. Where does he find the time?
Jerking people around like that is relatively harmless. But elections are based on the notion that a single sampling of consensus at a pre-set time is a good way to settle important decisions. This turns social media into a battleground for propaganda wars with real consequences. Building a messaging platform? Execute well and your users can topple dictators… or make a reality-show host President of the United States.
Tech companies are stumbling their way into responsibility over big important chunks of the real world. Through the perfect bureaucracy of software they know more, can predict more, and can direct & influence more human behavior than any government dreamed of. They can also data-drive us right off a cliff. Companies are as like gods, so they better get good at it.
And then the priest says…
“Bullshit. We’re running a business here.”
Priests (aka viziers) are the people arrayed around the exec. If the wandering guru gets his mojo from staying in motion and questioning assumptions, the priest gets his from executing on the current vision and building a repeatable business. He doesn’t have time to humor some strategy consultant with a Poirot Complex, and is jealous of his hold on the nerd-king’s ear.
“We’re not a government,” he repeats. “We’re just making atoms move like bits. Once we’ve worked out the bugs in the espresso-brewing delivery drones we won’t even need physical cafés. We need to focus on execution.”
“We still haven’t decided to go all-in on that,” the nerd-king says.
“Ah,” says the guru. “Assume you do. What happens to liability –let alone security– when you can drop double half-caf ventis through anyone’s window? What happens to the principle of hospitality when you are no longer a host, but a guest?”
There’s a lot going on here. Let’s relabel the quadrants and remove the people. Talented individuals can and do occupy any of the four quadrants from time to time, but tend to gravitate toward one.
- Doctrine, aka “the vision thing”. Fundamental truths about a particular domain. One of Google’s earliest doctrines was that instead of ever-smarter O(N³) algorithms over small data, you should run dumb linear ones over big data. Manipulating doctrine is about screwing around with basic assumptions. You take what seems like a law of nature and deny it or flip it upside-down and see how the universe would play out differently. If that world-view describes reality better, you run with it.
- Strategy, which is not about high-falutin’ abstract ideas, but specific actions that are informed by the doctrine and context under which you operate. For example, from 2003 or so it was a very effective stratagem for programmers to learn how to write well and to found open-source projects in order to level up their influence. Today it’s still valid but less of a home run.
- Tactics, which is about abstract ideas. Or rather generalized repeatable ideas. Think of A/B testing and all of the tried and true tactics that help you do it well. Tacticians collect and catalog things that seem to work.
- Operations, or getting shit done. GSD is about hitting a target revealed by strategy developed under a doctrine, using the right tactics. Or developing new ones out of desperation, which is generally how new tactics get tried out. An operations person gets positive joy from working on exactly the right problem in exactly the wrong way.
So a priest, a guru, and a nerd-king walk into a conference room. It measures 2×2.
The priest thinks he’s strategy but actually is the custodian of tactics, curating the repeatable processes that seem to work. Execution is what he worries about but doesn’t do directly. That is left to the operational folks whom he mislabels “tacticians”.
The nerd-king thinks his area is doctrine, defining the mission. But you can’t do that all the time. His natural position is in strategy, choosing when and where to strike at a given moment. When the chips are down, the shot clock is running out, and there are no more gambling and sports metaphors to turn to, it’s the executive who makes the call about what strategy to pursue.
The strategy guru is neither strategic nor a guru. He’s a doctrine prophet, sparring with the nerd-king while juggling worlds.
And of course, the operations grump isn’t even in the room.
About Carlos Bueno
Priest = COO usually in SV tech companies 80% of the time, CMO or VP Sales 20% of the time
Ratios reverse in old economy companies.
I may have sample bias.
Thinking out loud.. Google proved that getting doctrine right means a lot. Kodak proved what happens when you get it wrong. Uber and Facebook are learning that With great strategy comes great social responsibility. Aging infrastructure is a sign Of tactical lethargy and dogmatic hatred is a sign of tactical over-emphasis and the debate over automation is essentially debate about what role humans should play in operations. Any flaws in that synopsis?
Anyway, love this site and posts like this.
Any recommendations for people/companies to follow who are at the frontier of new doctrine?
One quadrant is not “higher” than the other. Often you’ll find that a new tactic developed under the gun informs a new doctrine, which then generates new strategies.
At Facebook we stumbled into the tactic of using CPU instructions as a stable measurement of the computational cost of running code. This was a field-expedient way to get a particular job done. Then we realized that this allowed us to model our entire fleet of servers, no matter what the specs, in a simple way. This lead to the doctrine of modeling computers as “instruction machines” each with a rating (15.5 giga-inst/sec, 27.3/sec, etc) and “instruction cost” as a way to model products & features of the site.
This lead to many many new useful strategies, tactics, and even new doctrine, that altogether saved the company north of a billion dollars. For example: the HHVM compiler team realized that the single most important factor in the speed of the code their compiler spit out was the number of load/store CPU instructions. It didn’t matter how “fast” their code was if the CPU spent most of its time waiting to fetch that code from RAM before executing it. This lead them to focus on sophisticated management of the CPU cache, and not “speed” as such, which counterintuitively yielded another 2X of performance.