First, find how far the failure reaches
Looker Studio has three layers, as the data sources and connectors guide explains: a connector fetches rows, a data source holds the credentials, fields and freshness, and a report draws charts. The reach of a failure shows which layer it belongs to.
| Question | If yes | If no |
|---|---|---|
| Does every chart on the same data source fail, across reports? | The source or the connector: credentials or quota | The chart and what it reads: a changed field, scope, filters, a blend |
| Does the failure come and go with time of day or the number of viewers? | Quota or load | A setting that is always wrong |
| Does it fail for some viewers and not for the owner? | Report access or viewer's credentials | Something every viewer shares |
| Did it start on a date that matches a change: a leaver, a GA4 edit, a report edit? | Start with that change | Work through the table below |
Open the report as the affected viewer, or a test account with the same access, because the owner's view can hide the problem. Error messages and connector limits are documented in the Looker Studio help centre and change over time, so this page does not quote them. If the chart shows numbers that are simply different from GA4, the sibling page on Looker Studio not matching GA4 applies instead, and a warehouse table that disagrees with the interface belongs to the page on a BigQuery export that does not match GA4. Other measurement symptoms are routed from the analytics troubleshooting collection.
Diagnosis table
Diagnosis
| Symptom | Check | Likely cause | Fix |
|---|---|---|---|
| Every chart on one or more sources fails, across reports, from one date | Each source's credential mode and the credential owner's account status | Data source credentials belonging to someone who left or lost access | Reconnect the sources with a shared account the business controls |
| Charts fail at busy times and work later or after a wait | Viewer numbers, charts per page and the source's freshness setting | Connector quota exhausted at busy times | Let the cache work, cut queries, extract stable data, move heavy use to a warehouse |
| Charts that use one field fail; others on the same source work | The field list in the data source against the fields the product now offers | A field renamed, removed or retyped at the source | Refresh the source's fields and remap the charts to the current field |
| A new chart fails as soon as a dimension or metric is added | The scope of each field the chart combines | Dimensions and metrics that cannot be combined | Use fields of the same scope, or split the chart |
| A chart shows no data, without an error | The chart's filters, the page and report controls' defaults, and the date range | A filter, control default or date range that excludes every row | Correct the filter or default; widen the range |
| A blended chart is empty or its metrics are blank | Join keys, key types and join types, with a temporary full outer join | A blend whose join keys never match | Normalise the key, align types and grain, choose the join deliberately |
| Errors for some viewers, nothing wrong for the owner | Report sharing, link settings, embed settings and each source's credential mode | The viewer lacking access to the report or, under viewer's credentials, to the data | Grant the right access, or change the credential mode deliberately |
Fix, verify and prevent each cause
Credentials that belonged to someone who left
Each data source fetches data with either owner's credentials or viewer's credentials. Under owner's credentials, the data source guide notes that the data stops flowing when the owner leaves and their account is suspended, and the same happens when the owner simply loses access to the property or account. Every chart on those sources fails together, often across several reports at once. The dashboard governance guide describes the same failure from the ownership side.
Fix: reconnect each affected source using an account the business controls rather than a person, transfer ownership of the sources and reports to it, and check which other sources the departed account owned.
Verify: every chart on the affected sources loads for a test viewer, and the source settings show the shared account as the credential holder.
Prevent: record every source with its credential holder in the dashboard inventory, and add "transfer report and source ownership" to the leaver's checklist so it happens before the last day.
A connector quota used up at busy times
Connectors that call product APIs, such as the Google Analytics and Google Ads connectors, are bound by those APIs' documented request limits. The performance and caching guide and the data source guide explain that a busy report with many charts, many viewers and a short freshness window can exhaust them, and that charts then fail with an error instead of showing stale data.
Fix: set a Data FreshnessWhether a table, feed or report holds data as recent as it should; a freshness check fails when the latest date is older than the source's expected arrival. window that matches how often the data actually changes, default date controls to a shared closed period so viewers' queries can be served from the cache, cut charts and data-driven controls per page, and move stable aggregated data to an extract. When a report carries heavy daily use, move it to a warehouse table, which answers queries without spending the API quota.
Verify: the report loads at the busiest time of the next working week.
Prevent: review busy reports whenever their audience grows, because quota is consumed per query and every new viewer adds queries.
A field that changed at the source
A data source stores the fields as they were when it was built or last refreshed, so a change in the product can leave the source and its charts out of step. In GA4Google's event-based analytics platform., a custom dimension registered at the wrong scope cannot be edited, as the events and parameters guide and common mistakes page explain; it is archived and registered again, often under a new name, and the inherited property clean-up lists remapping the dashboard fields that read it as part of that change. A field retyped in the source editor can also stop a chart working, for example a date returned as text cannot drive a date control until its type is set.
Fix: refresh the data source's fields, map each failing chart to the current field, check any calculated field that refers to the old one, and set types at the source.
Verify: every chart that used the old field loads, with values for dates after the new field was registered.
Prevent: before archiving or renaming anything in GA4, search the data sources that read it, and log the change with its date.
Dimensions and metrics that cannot be combined
A GA4 number only makes sense when the dimension and the metric describe the same thing. The reports and dimensions guide explains the scopes, user, session, event and item, and notes that the interface hides incompatible choices and explorations flag them. A Looker Studio chart lets an editor put any fields together, so a combination the property cannot answer reaches the connector and the chart shows an error or nothing.
Fix: replace the field that does not fit with one of the metric's own scope, or split the question into two charts.
Verify: the chart loads and its total matches a scorecard of the same metric for the same range.
Prevent: name fields with their scope in the data source so the next editor can see which ones belong together.
A filter, default or date range that leaves no rows
A chart can work perfectly and still have nothing to show: a filter on a value that was renamed, a control default selecting a market that no longer exists, a report-level filter reaching a chart it was never meant for, or a stale fixed date range. The dashboard design guide covers how filters, control scope and defaults should be set. Two GA4 behaviours make a narrow cut emptier still: data thresholdsThe GA4 mechanism that withholds report rows that could identify individuals, usually small segments over short periods, when signals or modelling are used. can withhold small rows, and values beyond a dimension's CardinalityThe number of distinct values a dimension or field holds; high-cardinality dimensions, such as page paths with IDs, can exceed report row limits. limit are folded into an "(other)" row, so a filter on one of them finds nothing.
Fix: remove the chart's filters one at a time until rows appear, correct the one that excludes everything, and reset defaults to the period and values the page is about.
Verify: with every control at its default, each chart shows the rows a viewer expects.
Prevent: label any chart that ignores a page control, and check filters whenever campaign names, markets or channel labels change.
A blend whose keys never match
A blend joins tables inside the report on shared fields. The blending tutorial explains why keys fail to match: case, spacing or spelling differences, a number on one side and text on the other, or a date against a date-and-time. With an inner join, every unmatched row disappears, and if none match the chart is empty; with an outer join, the joined side's metrics are blank. A mismatch of grainWhat one row of a table represents, such as one order, one order line, one session or one customer per day; queries and metric definitions must match it. causes the opposite fault, duplicates, which the sibling page on numbers that do not match covers.
Fix: switch the join to full outer temporarily to list unmatched rows on each side, create a normalised key at the source, align types and dates, and choose the spine and join type as the tutorial sets out.
Verify: the spine's blended totals equal its own totals, and the unmatched rows are explained.
Prevent: move any blend used on more than one page into one upstream table with a tested key, and record the key's convention.
A viewer without access
Sharing a report is two decisions: who may open it, and whose data fills it. The sharing and embedding guide explains both. A viewer not invited to a restricted report cannot open it; an embedded report carries its sharing settings into the frame, so visitors who are not signed in as an invited account see an access error there; and under viewer's credentials, anyone without their own access to the property, account or table sees an error in place of the chart.
Fix: grant report access to the right people or group, and either give the viewer access to the data or change the source's credential mode deliberately, aware that owner's credentials show the owner's view to everyone who can open the report.
Verify: a test account with the viewer's access sees every chart.
Prevent: test every report with a viewer-level account before sharing, and review access when the audience changes.
A prevention routine: owners, credentials and a launch check for every report
Most of these failures are ownership failures that surface as errors: a report runs on someone's credentials, reads fields someone else may change and is opened by people nobody tested as.
| Habit | What it prevents | Where it lives |
|---|---|---|
| Every source's credentials held by an account the business controls | Blank reports when a person leaves | Data source settings; leaver checklist |
| An inventory of reports and sources with a present owner | Reports nobody can repair | Dashboard governance |
| A launch check before anyone else sees a report | Access errors, empty defaults, unchecked blends | Dashboard launch QA checklist |
| A search of data sources before any GA4 definition is archived or renamed | Charts reading fields that no longer exist | Measurement change log |
| A review of busy reports when their audience grows | Quota errors at peak times | Report owner's review |
The dashboard launch QA checklist already asks that data source credentials belong to a shared owner, that blends and fields are checked, and that permissions suit each role; run it for every new or rebuilt report, and again when a report changes hands. Keep calculated fields at the source rather than in charts, so a changed field is repaired once. Log GA4 changes that affect reports, such as archived definitions, with the reports that read each one.
When a report does fail, find the cause behind the cause. The five whys turns "the report went blank" into "the sources ran on one person's credentials because no shared account existed when the report was built", which is the finding that stops it happening on the next report.
A cleaning company's Monday report goes blank (illustrative)
Jaxvelda Cleaning restores its Monday report
Jaxvelda Cleaning, an invented commercial cleaning company, opens its Monday report to find every chart on 4 of its 6
pages showing an error. The analyst who built it left 2 weeks earlier and the account was suspended on Friday. All 3
data sources behind the failing pages ran on owner's credentials in the analyst's name, so the reach check points
straight at credentials: every chart on those sources fails for everyone. The marketing manager reconnects them with a shared reporting account that takes ownership of the sources and the
report. One page still fails: its charts read a GA4 custom dimension, contract_type, that was archived in September and registered again
at the right scope as contract_kind. The fields are refreshed and the 5 charts that read it are remapped, with a
note that values begin on the date of the new registration. The report is added to the dashboard inventory with the
manager as owner, and the leaver checklist gains a line for report and source ownership.
Common mistakes
- Debugging as the owner. The owner has every permission and sees a report that works; test as the viewer.
- Fixing one chart when the source is broken. If every chart on a source fails, the fix is in the source.
- Refreshing a quota error. Each refresh is more queries; reduce the load instead.
- Archiving a GA4 definition without searching for its readers. The charts that used it stop working.
- Assuming an empty chart is an error. A filter or default that leaves no rows looks the same.
- Switching to owner's credentials to stop an access error. It works, and shows the owner's data to every viewer.
- Leaving reports in a leaver's name. Transfer ownership before the last day, not after the report goes blank.
