Bottom Line Studio
The dashboard nobody opened
Every dashboard refreshes on schedule, whether or not anyone looks. Before you migrate, rebuild or pay for another seat, find out which ones people actually use.
If nobody would notice it going stale, nobody needs it fresh.
Sound familiar? Your BI tool has more dashboards than your sales team has reps. Every one of them refreshes overnight. You couldn’t name who opens half of them, and nobody wants to be the one who turns one off.
Parity is the wrong finish line
On a client platform this year, we moved a sales leaderboard off a licensed, per-seat BI tool and onto a small app that read the same governed warehouse tables. The new version could change the same day someone asked. The old one needed a workbook edit and a publish cycle every time.
I almost made it a rebuild-everything project. Matching the old tool feature for feature feels safe, but it answers the wrong question. Parity tells you the new thing can replace the old one. It doesn’t tell you anyone needs either.
The rule we actually used to switch the old leaderboard off had two parts: parity and adoption. The new version had to match what mattered, and people had to be opening it every day. Only then did the old one go dark.
Staleness tells you who cares
The most useful signal turned out to be a side effect of something else. The platform sent freshness alerts per owner: when the tables feeding a dashboard went stale, that dashboard’s owner heard about it, and nobody else did.
That design forces a question every dashboard should have to answer: who gets the alert? If the honest answer is “nobody,” you’ve found a dashboard that refreshes every night for an audience of zero. If a dashboard could go stale for a week without anyone noticing, it isn’t informing a decision. It’s decoration with a compute bill.
Nobody turns them off, so they pile up
Dashboards pile up for ordinary reasons. Someone asked a question once, an analyst built a view to answer it, and the view outlived the question. The person who asked moved on. The analyst doesn’t want to delete something a VP might open next quarter. The tool bills per seat, not per dashboard, so nothing on the invoice says which ones matter.
The cost never shows up as one big line. It shows up as refresh jobs competing with the ones that matter, definitions that drift because three dashboards compute the same number three ways, and new hires who can’t tell which pipeline view is the real one. Every extra dashboard is one more place the number can disagree with itself.
Sort every dashboard three ways
You don’t need a committee. You need usage logs, an owner list and an afternoon.
| Keep | Merge | Retire | |
|---|---|---|---|
| Opened in the last 30 days | Yes, by the people it’s for | Yes, but the same numbers live elsewhere | No, or only by its builder |
| Named owner | Yes, and they’d notice it stale | Yes, on the other dashboard | Nobody claims it |
| Changes a decision | A meeting or a forecast uses it | Duplicates one that does | Nobody can name one |
| What you do | Fix it, alert its owner, keep it fresh | Point people at the survivor, then archive | Archive it, keep the query for 90 days |
Retire it quietly, keep the query
Retiring a dashboard feels risky because someone, somewhere, might need it. So make retirement reversible:
- Pull the usage log first. Who opened what in the last 30 and 90 days. Count people, not page loads; the builder checking their own work doesn’t count.
- Put a name on every survivor. If nobody will take the alert, that’s your answer.
- Ask what decision it changes. “It’s nice to have” is a merge or a retire.
- Archive, don’t delete. Keep the query and the definition for 90 days. If nobody asks, it’s gone for good.
- Stop paying for the empty seats. Licenses and refresh schedules are the line items that make this worth an afternoon.
The dashboards that survive get better, too. Fewer of them means each one has an owner who hears when it breaks, and a number that matches the meeting it’s used in.
How many of your dashboards would anyone notice going stale for a week?