2,650 Reports and Nobody Knew Which Ones Mattered

Stop reading the names. Check the last run dates.

2,650 Salesforce reports, one question that sorted them, and a hidden folder that wasn't hidden.

2,650 Reports and Nobody Knew Which Ones Mattered

Every Salesforce org I've ever seen has the same folder somewhere. It's called something like "Old Reports" or "Bob's Stuff" or, in my case, a dozen folders I'd renamed to start with "ZZZZ" so they'd sink to the bottom of the list. Out of sight, out of mind. Except they weren't out of sight.

At the start of this year our org had about 2,650 shared reports, 58 dashboards and 185 folders. People built another 124 reports along the way. Today our users see 434 reports, 16 dashboards and 21 top-level report folders. The orphaned report folder called "Unfiled Public Reports" went from 121 reports to zero. We also cut 26 custom report types nobody had ever built a report on.

434
Reports left, down from 2,650
84%
Fewer reports for users
16
Dashboards, down from 58
0
Unfiled reports, down from 121

Nobody has asked for one back.

The question that matters: "Has anyone run this?"

You can't clean up reports by looking at their names. "Q3 Pipeline FINAL v2" might be the CFO's favorite report, or it might be from 2017. The only honest signal is whether someone actually runs it.

Salesforce gives you one good field for this: Report.LastRunDate. It's org-wide. It updates when anyone runs the report, when a dashboard refreshes it, or when a subscription fires. That made it our source of truth.

Some fields look useful and aren't:

  • LastViewedDate only tells you about you, the person running the query. Useless for "has anyone used this."
  • Dashboard refresh dates came back blank for every dashboard in our org. Instead, we used a simple rule: a dashboard can't have been refreshed more recently than the oldest report on it, because a refresh re-runs every component.
  • "Who ran it" is unknowable without Event Monitoring, which we don't have. So we never claimed to know who. We only claimed to know when.

The criteria

This is the part I'd steal if I were you. We set the rules first, then let the data sort itself.

Last runWhat we did
Over 1 year ago (or never)Back up, then delete
6–12 months agoMove to a hidden "holding" folder, review in 3 months
Under 6 monthsLeave it alone

And the exceptions, which mattered as much as the rules:

  • Pricing folders: hands off. Price lists get run once a year at renewal time. A year-old run date doesn't mean dead.
  • Private reports: hands off, unless one was actively blocking something.
  • Reports less than 6 months old: kept, even if never run. Someone just built them for a reason.
  • Anything still in use somewhere: kept. Dashboards, custom links, Lightning pages, flows and our integration platform all got checked before every delete.
  • Dashboards on a home page: always checked first. One dashboard that hadn't refreshed in months is embedded on our main home page, so it stays until we redesign that page.
  • Custom report types: deleted only if zero reports used them, including private reports. A report type with one private report on it is still in use.

The "holding folder" trick

The middle category is where most cleanups stall. Six months unused isn't dead enough to delete, but it's clutter.

So we didn't delete. We moved 112 reports and 3 dashboards into subfolders under a hidden ZZZZ folder, named after where they came from ("ZZZZ Sales — Opportunities", etc.).

  • Hidden means no folder sharing. Only three admins can see them, after one more fix I'll get to below.
  • Moving keeps the record ID. If someone shouts "where's my report?", we move it back, and every link and bookmark still works.
  • Review date: February. Anything nobody asked for gets deleted then.

It's the Salesforce version of putting things in a box in the garage with a date on it.

How it actually got done

The approach was the same one from my earlier junk drawer post: VS Code, the Salesforce CLI, and Claude doing the analysis and the heavy lifting, with me making every keep-or-kill call.

The workflow for every batch looked like this:

  1. Query. Pull every report with its folder, last run date, created date and report type.
  2. Classify. Sort them against the criteria above. Claude produced the lists and I reviewed them, and more than once I moved something from "delete" to "keep."
  3. Back up. Retrieve every report and dashboard as metadata files before touching anything. Every deleted item has a file I can redeploy.
  4. Dry run. Salesforce can do a "test deploy" of a deletion. It tells you everything that would fail without changing anything.
  5. Delete or move. Run it for real.
  6. Verify. Re-query and confirm the counts and IDs.

In the first big pass this summer we cleared out 2,012 long-dead reports and 27 dashboards. In November, in one day, we did the rest: 223 more reports deleted, 20 dashboards deleted, 51 folders removed, 112 reports and 3 dashboards parked in the holding area, all 121 unfiled reports sorted into proper folders, and 33 folder-sharing entries removed because they didn't do anything.

What surprised me

"Hidden" folders weren't hidden. I'd removed all sharing from the ZZZZ folders months ago, and people could still see them. It turned out a couple of profiles had permissions ("View All Reports"-type settings) that bypass folder sharing entirely.

Folder sharing only works if nobody has the master key.

Invisible dashboards block deletes. Several times Salesforce refused to delete a report because "a dashboard references it." No dashboard I could see referenced it. The culprit was always a private dashboard, usually owned by someone who'd left the company years ago. Salesforce Support pointed us at this the first time. The second time we knew how to find them ourselves.

Some reports were too broken to move. Three reports had filter logic pointing to a filter that no longer existed. Salesforce happily displays them, but it won't let you save, move or restore them. We deleted them.

The AI got numbers wrong. So did I. At one point Claude said there were "about 40" reports in a group. There were 25. When I first added up the totals for this post, I left out the 124 reports people built this year. Claude's mistake wasn't dangerous, because every delete went through a list I could see and a dry run. Mine got caught when the totals didn't add up. That's the same lesson as last time: a wrong answer looks exactly like a right one on screen. The process has to catch it, not the confidence.

Where we landed

Start of yearVisible to users today
Shared reports~2,650434
Dashboards5816
Folders18537
Unfiled Public Reports1210
Custom report types6034

That's about 84% fewer reports for users to wade through, and every one that's left has been run in the last six months or has a reason to exist.

If you're staring at your own pile, start with one question: "When was this last run?" Let the AI do the sorting. You make the call.