How the Tide Is Made

Somebody once gave me a phrase. It simplified the idea that instead of solving every individual case, you build something that lets more and more of those cases solve themselves. The phrase? Raise the waterline.

I immediately loved the idea. Which probably should have set off warning klaxons. I don't even know what the opposite would be if I could somehow extend the metaphor. Don't raise the waterline, bail the water? Plug the holes in the hull? Patch some leaks? Sandbag the shore?

What I took from it was simple: good work raises the waterline.

The idea is that you don't keep solving the same slightly different, little version of a problem over and over. Three customers need slightly different things, say, approval paths in your software, so you stop building approval paths and build whatever is the most common denominator underneath all the variations of approval paths. You can build a generic and customizable approval-related software layer, or you can build four different contract generators for the customers yelling at you that that generic version isn't tailored to them.

I fully admit I have a weakness for the generic path. Give me four exceptions and about twenty minutes and I will quite happily tell you they are evidence of a system.

The problem is reality, which has no obligation to be any way other than it is. Sometimes four exceptions are four exceptions. A cough, a cough, a cough, and a cough don't necessarily add up to The Coughing Problem. One person has the flu, somebody swallowed wrong, one guy smokes two packs a day (still), and the fourth person is sitting underneath an air conditioner that appears to have been manufactured by Boeing in 1974 and hasn't been cleaned and good god don't open it up.

Yet once we've named the category, it becomes difficult to unsee. Now we have a Coughing Initiative.

Solving four similar problems four different ways is mildly embarrassing because it seems, on the surface, inelegant. You start wanting to apologize before the first complaint.

An abstract, generic solution has a nice story: we understood the Coughing pattern. We found the layer underneath it (lungs and stuff). We fixed the idea of a cough, we're pretty sure.

I can sympathize with why it usually wins. Personally, I'd rather discover and attempt to solve The Coughing Problem than hand four people four different remedies. One sounds like understanding. The other sounds like perpetual errands. But the path that sounds like understanding is possibly beset by, at least lined with hubris and vanity. Because specificity contains information too, and pushing that below the waterline can be the wrong approach. Sometimes you have to bail some water. It's the spice of life.

The person with the flu and the person underneath that cursed Boeing air conditioner may both cough, but collapsing them into the same category can actually move you farther away from helping either one. That air conditioner needs a service appointment.

Sometimes the ugly little fix is ugly because reality is ugly.

Categories like "cough" are useful because of what they ignore. We call all four people coughers because, for whatever we're trying to do, the differences between influenza, cigarettes, swallowing wrong, and the moldy air conditioner can temporarily be set aside. Which works until it doesn't. It can be a bad bargain, when one of those differences is the key variable or important part, when the act of categorization shaved off some key local weirdness.

All categories have a garbage bin somewhere off to the side, full of details we decided didn't matter. But sometimes you made a bad deal, and eventually have to go look in the trash, maybe rummage around a little.

Categories are clean and convenient to think about.

But somewhere on some hard-to-perceive spectrum, we know categorization abstracts away some of what's interesting. The lines set around a category stop being a convenient way to think and start interfering with what's recognized.

Does the category serve what it was supposed to? Should these edge cases still be happening on the periphery?

You see this almost anywhere people build systems around repeated behavior: products, policies, organizations, institutions. All of those come with boundaries, so a few exceptions are easy to dismiss. Eventually, though, the volume and obvious character of those exceptions becomes evidence that the boundary is off.

Is the border set at the right place and now you're just hitting its limits? Was the border drawn around the wrong thing?

Elegance can reach a dangerous point: a cleanly designed system can make the world outside its boundaries look quite disorganized, when that disorder should actually belong to the system. Those "edge" cases should be sitting on this side of the border.

Time zones are my favorite example of what happens when a very clean idea meets people doing stuff in the world.

The planet gives us all a very seductive starting point: twenty-four hours. Easy. Great. So, draw twenty-four nice vertical bands around the globe. Let's all go home.

Nope.

Nepal is forty-five minutes off the hour. India is thirty.

The majority of Arizona doesn't change its clocks for daylight saving time, except the Navajo Nation does. The Hopi Reservation, geographically surrounded (largely) by the Navajo Nation, doesn't.

China uses one official time zone, despite having the landmass for five.

All around the world, those 24 clean bands got bent, interrupted, subdivided and generally adapted by actual places full of actual people with daily needs.

An elegant system with no exceptions might sometimes be evidence that you've finally done it, you've found the right abstraction, but it could mean you've pushed every inconvenient fact below the waterline. Or out of your timezone.

Who knows, really? Sometimes raising the waterline means building the common thing underneath. Sometimes the boats really do have holes that need plugging.

I still like raising the waterline. I always will, given my admitted preferences or aesthetic bias or whatever you want to call it.

I'm just less confident now that every collection of similar-looking problems even has a waterline underneath it. Ultimately we're always placing bets on which differences matter, which ones don't. Sometimes they're clutter, other times they're the most important information you have.

What are you allowed to throw away? Who knows. Keep an eye on the bin.