What the Data Is Not Telling You
Every dataset is a record of what someone bothered to write down. The interesting question is what the writing down left out.
We run on data and automate everything we can measure. That is precisely why the second question in every meeting is what the data could be missing. A number is rarely wrong. It is routinely incomplete in a direction nobody checked.
Take leasing. Tour-to-lease conversion is measured, reported, and optimized. The people who never toured are not in the dataset at all. Tighten the listing until only pre-qualified applicants inquire and conversion looks superb while vacancy sits where it was — the metric improves as the business gets worse.
Or maintenance. Ticket volume is easy to count, and a falling count reads as a well-run portfolio. It also reads as residents who have stopped bothering to call. One of those shows up in this quarter’s report. The other shows up at renewal, where it costs several times more.
Abraham Wald made the canonical version of this point in 1943. Asked where to armor bombers based on where returning aircraft showed damage, he pointed out that the sample consisted entirely of planes that made it back. The armor belonged where the survivors had no holes, because the aircraft hit there were not in the data.
So the standing pair of questions is: what is this telling us, and what could it be leaving out? The second is harder in a specific way — the missing thing has no column, no chart, and no owner. Nobody is going to raise it on your behalf.
Which is why someone goes and looks. Walk the property, sit in the leasing office, take the maintenance call, read the cancellations. Not because anecdotes outrank data; they do not. Because the anecdote is usually the first evidence that a column you never built should exist.
It also changes how we define a metric. If a number can be improved without the underlying thing improving, it eventually will be — not through bad faith, but through ordinary people responding rationally to what they are measured on. The only durable fix is to define the measure so that gaming it and improving the business are the same act.
Automation compounds both the value and the risk. The more automated the reporting, the more confident and the narrower it becomes, because a dashboard is an argument about what matters — made once, by whoever built it, and then rarely revisited. We revisit it.
The data is necessary. Treating it as complete is the mistake.