182 Customers, Give or Take Eight

The company has 182 customers, unless it has 174.

There's good news, though, in that 182...seems to check out. You can see the customers. There they are, in neat rows with names of companies (with contracts), with humans who speak to them. Someone walks you through the list with the quiet confidence of someone who has already won a key political debate in front of the entire country.

Then, you take that list to the other team, and the person you're working with there, well, their face gets scrunchy real quick and they start crossing off some of those neat rows. These two accounts? Departments of the same company, just billed separately. This one? They're mean and signed something that everybody agrees is a contract but has never actually paid for anything and won't answer any emails ("Do you have their phone number?"). That one? Acquired two years ago, why is it still on its own row that's not how we do things around here.

Another team is peering in at the problem. Please stay out of this, but they don't. Now there is talk of “meaningfully active customers,” which threatens to produce a new number. Three numbers? Definitely don't need that right now. They are quietly, politely removed from the investigation.

Counting is an older technology, one of our oldest, so this is getting a bit embarrassing. Could the Sumerians solve this?

Fine. 174.

Except, hmmm, 174 has problems too. You're walking your way to 174, asking what exactly counts here and there, and you're hearing some hedging and the phrase “well, technically” a few more times than you'd like. Your spider sense is going off like Green Goblin is hiding in one of those rows.

At this point you usually have to become stubborn about the metaphysics of the issue. Somewhere out there in physical reality, on the Earth, is a finite collection of organizations giving us money. Capital-b Business may be complicated in how that money makes its way from point A to point B, but you refuse to accept that customer count is quantum, sometimes 182, sometimes 174. Wave, particle, observer.

So you keep going.

Eventually you discover that almost nobody has been particularly careless. Isn't that interesting? There's no laziness or malice or anything like that. People often need words to mean different things in different contexts, words like "customer." Maybe both definitions worked perfectly well for years because nobody needed them to meet until just about...now.

Then, somebody does. A lot of BizOps work starts in that space, somewhere between not right now and yeah now please.

Customer count is hardly unique here. Headcount might be the reigning heavyweight champion of the world. Ask Finance, Recruiting, and HR how many people the company has, preferably while everyone is calm, caffeinated, and emotionally ready for whatever follows. The question is unleashed, wreaks havoc in the Zoom. One person counts paychecks that went out. Another person lists every possible type of employee: approved job reqs, accepted offers, people who start next month, people who leave next month. Another type that defies and avoids the English language like Neo and bullets.

Someone else has a slightly other number. It's feeling philosophical. Is it hot in here?

Who's wrong? Nobody necessarily. Everybody?

Companies make bargains, little and large, all day long: with reality, with angels, with devils dressed as angels. With various unidentified entities who decline to introduce themselves until much later. Some bargains aren't even obviously bargains until the bill arrives. They're practical decisions so everybody can drink their coffee and conquer the day, sip by sip.

Six months later, one of those local answers is standing in front of the whole company, the board, the outside world, who knows, and it's being asked to represent the entire company. Congratulations on the promotion! It had a job to do, it was doing that job, and nobody told it that someday it would be doing this job.

That's life, kid.

Except now emails are suddenly flying around with triple question marks above them (???) and multiple, overlapping Slack channels are spinning up. Humans working with humans working with entities.

The number isn't even invited to all the Slack channels, it's getting a little political.

Which is right about when BizOps has to parachute in.

Parachuting in sounds glamorous, but is it ever a good thing in a movie when someone is using a parachute? You don't know the terrain, and your map is old, and also you might break your legs.

Mostly you're opening old documents and asking people questions they thought had already been answered definitively. Where did xyz thing come from? Who did this?!?!

Right about here, people look at you funny. You feel like the inventor of ambiguity, and you are personally revealing your work to the world. Behold, it doesn't tie out or add up. Worse, it's not defined!

Like an organizational Pope, you seek to canonically bless one number, but before you can you have to figure out what each one was trying to be useful for in the first place. What job was this for? Which regime built it? And suddenly the map is updating in your hands, you're not falling in the air with a canvas sack anymore.

Eventually the map is...reasonably clear. You can see why 174 and 182 both survived this long. They had different jobs and weren't mortal enemies. 174 and 182 both survived a long time because the word customer had been moonlighting, freelancing, taking on odd jobs. Playing a little jazz.

Sometimes it meant a company with a contract. Over here it just meant "the entity paying money." Somewhere else, two subsidiaries were one customer (everyone involved knew they were really one company), except a key system still cheerfully maintained two customers because it had no other way. Signed-but-never-paid and acquisitions and mergers lurked in each number. And, boom, you're out of the philosophical idea of a Customer. The word just moved around a little, picked up a job at that moment in time.

Inside groups of people, this happens to ordinary nouns and verbs all the time. Customer. Headcount. Revenue. Gross Margin. Handoff. Launch. Active. Done. In-Progress. Independently, each can look solidly in place right up until somebody leans on them.

Somebody always leans on them.

So the lean happens, and now everybody is an expert lexicographer, only now money or reputation or both is at stake.

When each little corner in a team, company, institution, what have you, could use their definition of some thing for their own purposes, you could tolerate a little fuzziness. But here's where the fuzziness becomes too expensive, can no longer be tolerated: now somebody needs to make a decision that crosses those little corners.

You might be asking yourself if BizOps is just a cosmically, comically linguistic job? Word definer? What does "active" or "launched" mean? Is this person really a "headcount" yet? These words pass by in normal conversation, no problem, but then now you have to put one in some planning work, some model, some loaded discussion and you notice you're standing on a trapdoor.

The cosmic joke is that you kind of need a perfect word and there isn't one. Definitions aren't perfect, language is fluid. What would a perfect definition even mean?

There is no spoon.

What you need is something that stays solid when it's leaned on, no matter what little corner you start in. Which means you have to figure out and then know what decision is being made.

Knowing the decision helps. It's nice, but that doesn't mean the next part is pleasant. Because the actual nice thing about being a wave and a particle is that sometimes you're a wave, sometimes a particle. Until the Observer shows up and bam, wave, which has consequences.

But the Observer has to choose their fighter, which isn't neutral, and which ripples outward from that decision. The Hospital Observer defines "out-patient" as someone who didn't stay longer than 24 hours, and suddenly wheel chairs are flying out the front door at hour 23. The Transit Observer redefines "on time" and suddenly the bus tables look a lot better, while not one bus arrives more predictably than the day before.

Definitions are funny precisely because they don't always hold over time. We or entropy or whatever-unexpected-thing-that-happened slides them around. The word "awful" used to mean "full of awe." My personal favorite? The word "sanction" can somehow mean two opposing things at the same time.

So, with almost angry headers and bullet points, eventually somebody writes down how you're defining xyz thing now. Take that, entropy. Before that event ever happens, though, the definition is moving money and targets deeply held beliefs around. Which means getting everybody to agree on the verb or noun or whatever is at best, I don't know, half the work? One third? The other half or two thirds is getting them to live with what comes next: what the word now has to do.

Agreement is ceremonial to me, weirdly so, because it feels like that's the end goal because it was. Of one set of Steps. But it's Step 1 of 12 toward what matters even more: adoption. You don't have much yet even though the definition gets written down under angry-looking bold headers, all the important people have read the official announcement email or slack, maybe even replied "looks good." For one brief shining moment it feels like reality has shifted into its new form. But you've really only decided on what the new form is.

Three weeks later Team 182 is still using 182.

Which doesn't even mean they disagree! Are they resisting? Maybe, who knows, but probably not! Their report uses 182. Another report uses that report, which feeds a widely circulated chart, and soon enough we're all zooming out and looking at the company's plumbing and realizing we have a small infrastructure migration. Hmmm...

New realities come with a propagation delay.^[1]

That's what I watch for now. The weird places where the old thing is still turning up. That's where Step 2 lives, and it's where you can proceed or go back to the prior set of steps. That's where the new agreement can become ordinary.

Then the weirdos arrive. By which I mean the edge cases.

Your blood pressure spikes, but edge cases are a good sign. Your first response, a hot flash of emotion, is often something like: come on. Come on. Please why. You have spent however long getting everyone moving in the same direction, and now somebody produces the process equivalent of a six-legged dog with bee wings and asks whether your beautiful new system or process can handle it. You like bees, you like dogs. We can handle bees or dogs, but bee dogs?

Come. On.

Does an online order picked up in a store belong to e-commerce or the store? Every five or so years the fiscal calendar gets a fifty-third week; turns out this year's the year. Someone is both technically leaving your company and coming back Monday. They've moved to Asia, and now they're a contractor, which makes sense for some reason because it's Asia, okay, that's why. Somewhere, there is some thing that was born to avoid your agreed-upon categories.

Edge cases are often how people bring their real world to your shared system. Can what everybody just agreed to actually contain this bee dog? The examples may be outlandish and weird, but they're most times less an objection and more just thinking through or using the new thing for the first time.

That distinction matters. It's not indifference, which nods and says "looks good to me" and moves on. Nor is it resistance, which goes back to the old way five minutes later. The person showing up with some horrible little exception has at least tried to move into the new home. They're just not sure they can get their furniture in it. Sometimes the couch just doesn't fit through the door. Also they have a parrot, which has a very large parrot furniture thing.

It's your job to figure out if that couch, that weirdly large parrot thing, can get under the roof or if people and parrot are going to be sitting/perching outside for a while.

This is where the work gets somewhat less satisfying, either for you or those you have to work with, because you cannot promise that every piece of furniture gets inside. Do you design a whole house around a parrot thing? Are there other things the parrot might like? You explain you aren't anti-parrot per se.

But sometimes the couch or parrot thing reveals your front door is badly designed, way way too small. What's hard, usually, is explaining to them why they have to wait. The couch owner usually doesn't like that you need to see if three different people show up with basically the same couch, because that's the only real way to learn something about the door.

Okay and now the parrot person has bought more parrot stuff and would like to discuss widening every entrance in the building. Can we lift the ceiling? This needs a parrot floor.

The difficulty is figuring out what exact problem you have while the person, parrot on shoulder, is standing there looking at you.

Shared systems need to contain enough of reality that people trust them, but if they try to bend and contain every possible reality they stop being shared systems. They aren't ever really going to be a house, or an apartment building, or whatever you need.

Somewhere in all of that is a line, where the designs are mostly done. BizOps spends an unreasonable amount of time trying to find it.

There is another outcome here, but this way lies many wrecked BizOps ships. Because this outcome feels fantastic for quite a while: instead of fixing every door, and denying the weird parrot couch into the house, you just become the person who knows how to get all the weird furniture in through the door.

Send it to BizOps: no door too small.

Which, yeah, it's flattering. I guess it's also efficient and perhaps in some cases, it's the right thing. Because you know which definition bends where, who needs to be asked, which exception is real, somebody sends you the customer that is simultaneously one customer and three. This is how your desk is filled with weirdly shaped furniture, because you spent more time figuring out how to get the furniture through the door. Should you? Shouldn't you be stopping the parrot owner getting ambitious with their parrot plans?

Parts of the company don't quite fit together. That's a natural thing, but they eventually need to. This isn't only a house for parrots.

Something subtle will happen, will slowly be revealed to you, if you do this kind of work long enough. When everyone knows the human adapter(s) can handle all kinds of exceptions, are they exceptions? Two systems disagree: send it to BizOps. Two teams use different definitions: okay, send it over. Nobody quite remembers why a particular approval works the way it does, but you do, or you know who does.

For a while, this is genuinely useful and sometimes extremely useful to all parties involved. An exception got through, so that person got their parrot thing in the house, the house is still a house for now, and you solved a real problem for the mortgage holder (the company). Because most times there's a weird problem, the weird problem can have a quick answer and/or a right answer, which is the ballgame, that and/or. Do you rethink the door itself? Can you piece together a new door while all the normal furniture is still coming through? What is normal furniture?

Meanwhile, stuff has to get in the house, which accumulates evidence in favor of allowing exceptions: number of exceptions successfully resolved. People learn where to send the next one. And you learn more and more about how the place works because more of the strange stuff keeps arriving at your desk and you get to size it to the door.

Which is where the "what's best for all parties?" distinction gets harder to see. You are useful because people depend on you, and people depend on you at least partly because you've become so useful. Which is nice, for them and you. There's a line there, it moves around, and you have to keep your eye on it or at some point the human adapter will quietly merge into the architecture of the house.

The line will again and again unbury itself from evidence that's accumulated in little drifts. That evidence inevitably starts pointing in opposite directions. One kind is obvious: you are useful, please solve this problem. More people come to you, harder things make their way your way. Which feels good because the usefulness is visible and arrives addressed with your name on it.

The other kind, at a surface-level, seems less flattering. Questions that used to come directly to you no longer do; they're answered somewhere else. Somebody you taught explains the definition correctly and you're not even there to hear that explanation. But you notice you're not hearing about what you used to as much.

Nothing's gone wrong there, but it's easy to linger on the question of whether something actually has gone wrong. Usually, quite the opposite, in fact: knowledge moved to the right place. The couches are getting through the door, and you don't have to supervise or explain how flipping the couch to a 45 degree angle will do the job.

That second kind of evidence is just proof it's time to work on another part of the house.

By now our house analogy is deeply straining. No one would insure this house due to the parrots, six-legged dogs, couches stuck in doors, and several code violations.

Answering the simple question of what BizOps actually does is hard for a simple reason: the surface area moves around too much.^[2] One month you're deep in figuring out how many customers exist. Next month?

The work changes depending on what's needed. If things go well, some of what used to require you stops requiring you. Which is probably the part I like most.

You can now go find and fix the next part of the house, or decide if it's time to add an aviary.