Dashboard Bloat: A Practical Audit & Archive Framework
If your BI workspace feels like a thrift store—duplicates, abandoned experiments, and "final_v7" tabs—adoption will stall. Here’s the pruning framework I use each quarter to restore clarity without political fallout.
Symptoms of Bloat
- Stakeholders screenshotting numbers “just in case.”
- Conflicting definitions of the same metric across decks.
- New hires asking “Which dashboard is the source of truth?”
- Slow load times and failing refreshes.
The SCORE Model
I triage each candidate dashboard with a simple label. Goal: process 50–100 assets in a day, not start an archaeology expedition.
1. Inventory Pull
Export metadata (owner, last viewed, refresh status). In Power BI / Tableau / Looker this is scriptable. Sort by last_viewed ascending—biggest wins live there.
2. Rapid Classification
I skim structure, filters, refresh cadence. I don’t validate numbers yet. Just tag.
3. Overlap Pass
Group Candidate dashboards covering similar KPIs. Pick a survivor. Archive the rest, link to the survivor in description so dependency anxiety fades.
4. Standardize Survivors
- Add owner + data source disclaimer.
- Pin glossary or metric panel up top.
- Remove novelty visuals that add no signal.
5. Communicate Archive
I post a short changelog in Slack/Teams: what was archived, what replaced it, and how to request restoration (rare). Transparency prevents the “you deleted my thing” drama.
Sample Metadata Query (Looker / Postgres)
SELECT title, last_viewed_at, user_id, space_id
FROM looker_content
WHERE type = 'dashboard'
ORDER BY last_viewed_at ASC
LIMIT 200;
Guardrails
- Never delete within 30 days—archive first.
- Tag experimental dashboards at creation (
[EXP]prefix). - Enforce owner + refresh SLA fields in a lightweight template.
Dashboards age like produce, not wine. Pretending otherwise creates shadow spreadsheets.
ROI Tracking
After a cleanup cycle I log: dashboards archived, duplicates merged, refresh failures reduced, average load time delta. That tangible win buys the next cycle.
Takeaways
- Audit fast with a coarse model.
- Archive, don’t nuke.
- Surface owners + glossary.
- Prune overlaps ruthlessly.
- Celebrate performance wins.
Run this twice a year and your BI layer stops feeling like a landfill.