What do fleet managers, service leaders, and Teach For America recruiters have in common? Well, for starters, both learn eventually that tired data lies a lot.
What exactly is “tired data”? Years ago, I spent my days in a Teach For America phone bank, trying to woo college seniors away from lucrative Goldman Sachs and Google offers, and convince them to teach in schools around the country for two years instead. Often the calls went straight to voicemail. Occasionally, someone would pick up and abruptly hang up, or worse, say something memorably unkind (Hope you’re well, Jeff from Syracuse!) When someone did stay on the line, we’d connect on the mission, the pitch, eventually our shared interests. It was total extrovert bliss and I loved that part of my job.
But after each call came the part I absolutely hated: logging all the details of the conversation in the TFA recruitment database.
Picture an MS‑DOS‑era, gray‑screen database with clunky drop downs that took ages to save each entry. Some diligent recruiters spent 10 or 15 minutes entering meticulous notes after every call. Me? As a twenty‑something, I stacked as many calls as I could - the part I enjoyed—and delayed the note entry—the part I hated—for as long as I could get away with it without getting reprimanded. I’d often back‑fill my notes on a Friday afternoon, skipping details or half‑guessing so I could race off for $2 tacos. My entries became rushed, copy‑pasted, unintentional half‑truths. And I certainly wasn’t alone.
The result? A dataset full of entries that barely or inaccurately captured the richness of the actual conversations. Tired data.
At the time, I blamed an outdated, clunky system. Looking back, that hasn't changed and I know better that when data inputs rely entirely on humans— tired, busy, shortcut‑loving mere mortals— it needs to be effortless to use, to remain accurate.
From Teaching to Trucks
I saw the same problem years later while scaling Amazon’s fleet. Poor data quality reappeared all the time, this time with a bigger price tag.
Amazon isn't one for dashboard data, but on the repair and fleet side, our fleet management companies gave us these dashboards. They looked relatively respectable. We had decent looking repair logs, timestamps, cost codes, rows of excel sheets, even some pie charts. Crack them open though, and you’d see fiction everywhere:
-
Brake RO: “Issue: brakes.” Which brake? Pads? Rotors?
-
Repeat failures: Expensive, repeat repairs treated as brand-new events because no one logged the issue right the first three times.
-
Escalating out‑of‑service rates: Issues flagged days earlier in an inspection never made their urgent way to the shop floor; now a couple vehicles were out of service, kicking up 10x more in costs as we shuffled around for tows, rentals and replacements.
-
Missed warranties: $4,000 part reimbursements across 200 vehicles never recovered because we forgot to tag the part as being under warranty in time.
-
Pie‑chart abyss: Hundreds of “XXX” codes making up the largest slice on our part pie-charts, saying nothing at all.
Ultimately, no matter how sophisticated or visually appealing our systems and charts became, the truth remained: the data we relied on originated from real people, rushing through repetitive administrative tasks under real pressure. No one was slacking off: techs were hustling, vendors were working around the clock, internal operations teams were buried. Yet, when capturing the truth depends on tired, rushed people typing tiny truths into tinier boxes, the inevitable result is fiction that drives budgets, strategy, and decisions.
We all know the real power of high‑quality data: clarity. But even the most sophisticated dashboard or perfectly structured database in a fancy system can’t fix bad inputs. You have to fix it at the source.
Bad Data’s Biggest Risk
I still daydream about that old TFA phone bank and how different things could have turned out if we'd had today's note-taking apps transcribing every word, tagging context, and surfacing action items automatically. I might still be recruiting, for one! Similarly, easy data capture would have completely transformed my fleet management days, allowing proactive decision making, that can actually save money, rather than reactive decision-making.
So, tired data lies: loudly, confidently, expensively. The fix isn’t another shiny dashboard. It’s a solution that makes honesty almost effortless. Give people an easier way to tell the truth, and the numbers will finally tell it too.