"Single panes of glass" are horrible
A rant
This is the semi-weekly post for paid subscribers of what’s on my mind and data related thoughts.
This week, I happened to come across a tweet, the specifics aren’t important, but the tweet caused… feelings. It was essentially asking business leaders whether they’d pay for a “single pane of glass” that pulled data from a bunch of places, while also promising a slick UI, analysis tools, and other stuff.
The feelings that message evoked in me was… “I’ve been on a sinking ships like that before”.
The key term that did it for me was “single pane of glass”. I still don’t know where the origin of that term comes from beyond that it’s extremely common in tech circles. It vaguely smells military-ish, like someone pulled the concept of the glass cockpit HUD from aviation and said “we need THIS”. Or maybe they pulled the image of a sci-fi/cyberpunk elite hacker movie with magical floating screens and the power to manipulate the world’s information with their dancing fingertips.
Either way, I’ve been staffed on far too many “build a single pane of glass” projects over my 15+ years of work, and I don’t think I’ve ever succeeded. It’s ridiculously easy to say you’re going to build the one magic dashboard that will replace all other dashboards. It’s quite another matter to actually deliver in a meaningful way.
Whatever the origin the origin of the term, it’s now used as a shorthand for “user has everything piece of data they want or need to make their important decisions in this one place, instantly”. I don’t particularly object to the invocation of magic when specifying a user story since that’s where there’s usually an opportunity, but the proposed magic of single panes of glass seem fundamentally flawed.
Physical “single panes of glass” are special things
The few places I can think of that have single panes of glass are in transportation, when someone is driving a car or maybe flying a plane. There, we consistently see lots of companies and designers throw tons of effort and money into developing an interface so that the user has everything they need to do the job at hand — piloting. A car tells you about your speed, critical details about the car’s mechanics like engine temperature, RPMs, and remaining fuel, while also providing you a clear view of the road. Planes take that same idea and scale it up because they’re much more complex machines. Fighter jets take that complexity and add in “people shooting at you/you shooting at others” to even further hone the importance of good decision-making.
These specially designed interfaces differ from the magical IT interfaces that keep getting requested in a number of ways.
First, they’re designed for a fixed number of users, often just one, or a tiny crew. That person holds sole responsibility for making all the important decisions. They’re trained, either through practice or study, to know what they should do when the dashboard tells them certain things.
Then, these dashboards are extremely optimized for one particular, unchanging, task. No one would ever think of ripping out an airplane dashboard and attempting to use it for… well, anything else. Most of the interface elements were put there due to careful thought, user testing, and oftentimes the blood of past incidents. Even as things evolve over time to adapt to new technology and situations, the main problem they’re solving remains largely unchanged.
NONE of that really applies to what a typical “give me a single plane of glass to monitor my business/computer/process needs” requests that come down.
The executive who wants every bit of information at their fingertips isn’t actually making critical snap decisions. Why would anyone want anyone to make critical decisions at a moment’s notice? Steering an organization of any decent size isn’t like steering a slightly overloaded van.
The task is also constantly changing. One year we might be focusing on getting new users. The next year we’re trying to increase retention. Maybe someone else is cutting costs or trying to boost productivity. I’m pretty sure that there’s no set of metrics in the world that can speak to any and all of these simultaneously.
There’s always going to be multiple users from all over the place that are supposed to somehow use the new tool to do their work. But they literally have completely different jobs that require different things to do. Executives want to see one thing, product leads want another, engineers and researchers will want something else. That “single” pane of glass wants to become ten panes before it gets off the drawing board. Otherwise it’ll have a ton of weird knobs and settings to make it “customizable” (read: unusable).
What everyone wants to see also shifts over time as priorities and organizations change. So how can anyone do proper iterative research and work to pick metrics and make good designs? Everyone wants these projects done since yesterday, while it honestly takes months, years of work to figure out what should go into the dashboard or not.
If all this sounds like a very elaborate ticket to failure, you’d be correct. It’s a nightmare to have to start such projects. It’s a different nightmare to inherit one of these and have to maintain them. It’s yet another different kind of nightmare if a vendor sells it to your company and now you’re tasked with getting it working.
Probably the only reason many of us embark on these projects is because the people requesting them are powerful enough that we can’t say no. As in, we like to have our jobs and paychecks and if the Big Boss really wants people to dig a big hole in the yard and then fill it up again on their time, then sure?
Usually at the end, we wind up with this weird scoped-down dashboard that proves to have some small amount of utility for one group… so it’s not completely wasted effort…
But maybe one day I’ll come across a big success story.