How sprawl happens
Nobody decides to have too many dashboards. They accumulate through reasonable acts: a manager wants one chart changed and the analyst copies the dashboard rather than risk breaking the original; a new campaign gets its own view that nobody deletes when it ends; an agency builds its version; a departing colleague leaves theirs behind; a stakeholder saves a filtered copy to their own folder. Each copy carries its own calculated fields, its own date defaults and its own connector credentials, and each drifts from the others as soon as any one of them is edited.
The cost arrives slowly and then all at once. Two people bring different numbers to the same meeting and spend it arguing about which is right. A connector's owner leaves and every copy using their credential goes blank. A metric definition changes and only the dashboards someone remembers get updated. A board deck quotes a figure from a dashboard that was superseded months earlier. The underlying problem is never the tool; Looker Studio, Power BI and Tableau sprawl identically. The problem is that creation has no cost and abandonment has no signal. Governance supplies both.
Sprawl in dashboards is the same failure as sprawl in a tag container, and the container governance guide describes a parallel routine for Tag Manager; teams that run one will recognise the other. The measurement governance playbook schedules both routines alongside the tracking plan, the weekly data quality check and the yearly implementation audit.
Ownership and certification
Every dashboard has exactly one named owner, who is a person and not a team. The owner answers for the numbers, approves changes, decides who sees it and is the one asked at review time whether it still earns its place. When the owner leaves, ownership transfers before their last day or the dashboard is retired; an unowned dashboard is treated as a defect in the inventory.
Certification is a visible status that tells a viewer how much to trust what they see. Three tiers are enough:
| Tier | Meaning | Who can create one | Where metrics come from |
|---|---|---|---|
| Certified | The agreed source of truth for its subject; quoted in reviews and decks | The data owner, after review | The metrics layer only; no dashboard-side formulas for defined metrics |
| Team | Maintained by a team for its own operations; trusted within the team | A named owner in the team | The metrics layer where it exists; documented exceptions |
| Personal or draft | Exploration, prototypes, one-off analysis; never quoted | Anyone | Anything; no guarantees |
The tier is written into the dashboard's name and its header, so that nobody mistakes a draft for the record. Promotion from team to certified is a review: the owner shows that every metric reads from the metrics layer, that the totals reconcile to the sources, that the layout follows the dashboard design guide well enough for its audience, and that an entry exists in the inventory. One certified dashboard per subject is the rule; a second certified dashboard on the same subject is a definitions argument waiting to happen. For a dashboard about to reach stakeholders for the first time, the dashboard launch QA checklist applies these rules as a one-off gate: sources, reconciled totals, definitions, access, an owner and an inventory entry. Because certified figures are quoted in decisions, rates built on small counts should show those counts or carry a low-volume flag, as reading small numbers in marketing reports describes.
Tables tested on every load make the reconciliation part of that review a standing fact rather than a one-off check, and the tutorial on writing data tests for marketing tables shows how to build them.
The inventory
The inventory is a single list of every dashboard the organisation treats as real. It can be a sheet, a table in the warehouse or a feature of the BI tool; the medium matters less than the columns and the discipline of keeping it current.
| Column | What it records | Why it matters |
|---|---|---|
| Name and link | The dashboard's name under the convention, and its URL | Finding it; detecting duplicates |
| Owner | A named person | Someone to ask; someone to review |
| Tier | Certified, team, personal | How much to trust it |
| Audience and purpose | Who it is for and the decision or routine it supports | The test at review: is the decision still made? |
| Subject and KPIA metric tied directly to an objective. list | The metrics it shows, by their layer keys | Finding every dashboard a definition change touches |
| Data sources | The tables, connectors and credentials it reads | Impact of a source change; credential ownership |
| Refresh and cadence | How often data refreshes; how often it is meant to be read | Spotting dashboards read less often than they refresh |
| Access | Who can view and who can edit | Audit; sharing rules |
| Last reviewed | The date of the last governance review and its outcome | The retirement clock |
| Status | Active, under review, deprecated with a sunset date, retired | What the viewer should expect |
The inventory earns its keep the first time a metric definition changes: filter by the metric key and every affected dashboard is listed, with its owner, in a minute. Without it, the change reaches the dashboards someone happens to remember, and the rest go on showing the old number with the same confidence as before. The metric definition change checklist makes that filter one of the checks ticked before a new definition takes effect. A dashboard that was written up in a dashboard specification before it was built arrives with most of these columns already decided, which makes adding it to the inventory a matter of copying rather than reconstructing.
Naming, folders and access
Names carry the information a viewer needs before opening anything: the tier, the subject, the audience and, where it matters, the market or brand. A convention such as tier, then subject, then audience, in a fixed order with fixed separators, makes a folder listing readable and makes duplicates obvious. Dates and version numbers belong in the inventory, not the name, because a dated name is out of date the day after it is made. Names describing the chart type or the data source describe the wrong thing.
Folders mirror the tiers and then the subject. A certified folder that only the data owner can write to, team folders that teams own, and personal folders that are understood to be scratch space. A dashboard's location is then a second signal of its status, and a certified dashboard in a personal folder is as wrong as a personal dashboard in the certified one.
Access follows a few rules. View access is granted to the audience in the inventory, by group rather than by person where the tool allows it, so that joiners and leavers are handled once. Edit access to certified dashboards is restricted to the owner and a deputy. Data source credentials belong to a service account or a role, never to a person who might leave. External sharing, public links and embedding have their own rules, and the sharing and embedding guide sets them out for Looker Studio, including what a viewer of a shared report can see of the underlying source. Access is reviewed at the same time as everything else.
Change management
Certified dashboards change through a process; everything else changes freely, which is the point of having tiers. The process is light but real:
- Request. A change is written down with its reason, by anyone, to the owner.
- Assess. The owner checks the inventory for what the change touches, including whether it needs a metric definition change, which has its own versioning rules.
- Build in a copy. The change is made in a draft copy, named as a draft, not in the certified dashboard.
- Review. A second person compares old and new, reconciles any number that moved, and checks the layout.
- Release and announce. The certified dashboard is updated, the change log in the inventory is written, and the audience is told what changed and why, in the dashboard's header for a period and in whatever channel the team uses.
- Retire the draft. The copy is deleted, not kept just in case.
A change log on each certified dashboard, kept in the inventory or in a notes page within the dashboard, is what makes last quarter's figure explicable this quarter. The Looker Studio section covers the mechanics of report versions and copies in that tool.
Usage review and retirement
Dashboards die quietly, and the inventory's job is to notice. Most BI tools report views per dashboard and per viewer; where the tool does not, the owner asks the audience. The review asks, for each dashboard, three questions: is it still read, is the decision it supports still made, and does it still reconcile to the certified source. A no to any of them starts retirement.
Retirement is a process rather than a deletion, because someone always has a bookmark. The dashboard is marked deprecated in its name and header, with a sunset date and a link to its replacement if there is one; the inventory status changes; the owner tells the audience; and on the sunset date it is moved to an archive folder that only the data owner can see, where it stays for a defined period before deletion. Viewers who object in the deprecation window are the usage data the tool did not capture, and the owner decides whether the objection changes the decision.
A retailer's first governance review
A multi-brand retailer's first inventory finds 143 Looker Studio reports across the marketing group. Usage data shows 61 had no views in the previous 90 days, 38 were copies of four underlying dashboards with minor filter changes, and 11 read from a connector owned by an employee who had left, so their charts had been blank for weeks. The team certifies five dashboards, one per subject, assigns owners and moves them to a locked folder; marks 74 as deprecated with a 30-day sunset; and converts the four most-copied dashboards into certified ones with viewer-side filter controls so that the copies are unnecessary. Three months later the inventory lists 29 active reports, every certified dashboard reads from the metrics layer, and the weekly trading meeting opens the same URL every time.
The quarterly routine
Governance that depends on remembering does not happen, so the review runs on a calendar. Once a quarter, the data owner and each dashboard owner work through the inventory:
- Reconcile the inventory to the tool. List every report in the tool and match it to the inventory; anything unlisted is either added with an owner or deprecated.
- Confirm owners. Every dashboard has a current, present owner; leavers' dashboards are reassigned or retired.
- Read the usage. Views per dashboard and per audience over the quarter; flag anything below the threshold the team has set, and anything whose audience has changed.
- Check certification. Every certified dashboard still reads only from the metrics layer and still reconciles to the sources at the total level; any that computes its own version of a defined metric loses certification until fixed.
- Review access and credentials. Viewer groups match the inventory; edit rights are limited; no source credential belongs to a person.
- Apply pending definition changes. Any metric version change from the quarter is confirmed as applied to every dashboard the inventory lists for it.
- Retire. Start the sunset for everything flagged; delete whatever passed its sunset last quarter.
- Record the review. Last-reviewed dates updated; the count of active, deprecated and retired dashboards noted so that the trend is visible.
The routine is a morning's work for a team that has done it before and a week for one that has not, which is itself the argument for starting. The reporting and BI overview covers the cadence and audience design that decides which dashboards should exist in the first place, and the BI hub places governance among the other layers of the stack. The second most important KPIA metric tied directly to an objective. of a BI programme, after whether its numbers are trusted, is how few dashboards it took to earn that trust.
Dashboard sprawl is one of several reasons a monthly report becomes slow to produce; the diagnosis of a monthly marketing report that takes days places it among the others.
Common mistakes
- Governing by deletion. Deleting without a sunset teaches people to keep private copies, which is the sprawl you were trying to end.
- Owners that are teams. A team owns nothing; a person does.
- Two certified dashboards on one subject. The definitions argument has simply moved.
- Dates and versions in names. Out of date the next day; put them in the inventory.
- Credentials on people. The leaver's blank dashboards are the most common governance incident.
- Fixing certified dashboards in place. Build in a draft, review, release.
- Skipping the quarter because nothing changed. Something always has; the review is how you find out.
- Reviewing dashboards without reviewing definitions. A certified dashboard reading a stale definition is confidently wrong.
- Keeping the inventory somewhere nobody looks. It belongs where the dashboards are found, linked from the certified folder, or it rots as fast as the dashboards it was meant to track.
