Record first, filter later
The biggest lesson of this year: what you never record, you can never analyze. And recording has become a lot cheaper than it used to be.
If I had to keep one thing from the past year, it is this: the more data, the better. The filtering happens later.
That runs straight against what is usually advised, so let me explain why I see it that way.
The usual advice
Start small. Define your KPIs first, measure only what you need, avoid collecting data without a purpose.
That advice is well meant and it is partly right. A dashboard with forty numbers nobody looks at is worse than five that do get looked at.
But one thing gets confused in it: what you show and what you keep. For what you show, less is better. For what you keep, the opposite holds.
Why keeping is different
A number you do not record today is irrevocably gone a year from now.
That sounds obvious and it is structurally underestimated. At the moment you decide not to capture something, it looks superfluous. Twelve months later somebody asks a question you needed that data for, and then the only honest answer is that you do not know and will not know for another year.
I ran into that several times this year. Not with exotic data, with ordinary things. When a quote was refused and why. Which crew was on which job site. How long the gap was between site survey and execution.
None of that would have been hard to record. It is just that at the moment the process was set up, nobody asked the question.
What has changed
This used to be a cost benefit trade off. Recording more meant more storage, more maintenance, more work for whoever had to fill it in.
Two of those three have largely fallen away. Storage costs almost nothing now. Setting up a database is a matter of minutes. And building a small tool that captures data where it originates is days of work instead of months.
What has stayed is the third one: somebody has to fill it in. That is the real cost and it has not gone down.
The rule I use now
If something is not being recorded and it could answer a question later, we make sure it gets recorded. As close as possible to where it originates, and as cheaply as possible for whoever has to do it.
That last part is the condition the whole thing stands or falls on. Recording that makes someone's work heavier does not last. Recording that falls out as a side effect of something useful does last.
That is why we prefer to build tools that already solve something on their own. A stock app that spits out an order list. A survey app that produces a sheet. The user gets something, and meanwhile a series comes into being.
That is recording without it feeling like administration, and it is the only kind that keeps running.
Where you do draw lines
Two of them, and they matter.
You do not record personal data because you can. There the opposite applies: collect what you need and no more. That is not only legal, it is simply sensible, because data you do not have cannot leak.
And recording without structure is not recording. If it ends up in free text with no key to link it to anything else, you do not have data, you have notes. The point is not that something gets kept, it is that it can be used later.
The claim
Limit what you show, not what you keep. One number too many in a database costs you almost nothing. A number you never captured costs you the answer to the question you are going to ask a year from now.