DebugView in GA4: what it proves and what it does not
A purchase event ships on a Friday afternoon. Someone opens DebugView, watches purchase appear with the correct order value, and closes the ticket. On Monday the client sends a screenshot: revenue in GA4 is a third below the shop platform. Both people were right. The event did reach Google from that browser. Nothing in DebugView said the report would count it the same way.
DebugView answers one narrow question well: is this browser, right now, sending the events I think it is sending, with the parameters I configured? It is a debugging panel with a 30-minute memory and a device selector. It is not a preview of your reports.
Three questions that look like one
When someone says tracking is broken, they usually mean one of three separate things, and each has its own failure mode.
Did the browser send it? DebugView shows this, second by second, for a device you control.
Did it arrive with the values the report needs? A parameter can reach Google and still never appear in a report, because a parameter is not a dimension until someone registers it. DebugView will show order_tier: "gold" on every purchase while the report has no such column.
Did the report count it? This one depends on processing intervals, data filters, data thresholds and attribution, none of which DebugView knows about.
Most tickets described as broken tracking are the third question wearing the clothes of the first. The event was fine. The reporting layer was still working through yesterday.
Turning debug mode on
DebugView stays empty until debug mode is enabled for the device you are watching. There are two routes, and they do different things.
For your own device, use Google Tag Assistant at tagassistant.google.com or preview mode in Tag Manager. Both add a parameter to the page address that switches debug mode on for that browser session, and they leave no lasting setting behind.
For every visitor, set debug_mode to true in the configuration. With the Google tag that is a parameter on the config command, or on a single event if you only want one covered. In Tag Manager, set the debug_mode field on the Google tag, or on the Google Analytics event tag when the scope should be that tag alone.
The trap is in the off switch. Google states that debug mode is disabled by removing the debug_mode parameter, and that setting it to false does not disable it. A property that had the field left in during a test in March is still collecting debug traffic in November, and nobody sees it because the traffic looks like ordinary sessions.
What the panel is actually showing
DebugView opens from Admin, under Data display. The layout is three columns, and each one answers a different question.
The Seconds stream in the middle lists events logged in the last 60 seconds, newest at the top, each with a timestamp taken from the device that logged it. Click an event and you get its parameters. That click is where most real diagnosis happens, because a wrong value is invisible in a list of names.
The Minutes stream on the left holds up to 30 circles, one per minute, each with the number of events received during that minute. Click a circle and the Seconds stream fills with that minute instead. Thirty minutes is the entire memory of the tool, so a session you wanted to inspect an hour later is gone.
The right column carries Top Events for the 30-minute window and the current user properties for the selected device. If a user property changes mid-session, the newest value is what shows.
The device selector sits in the upper left, labelled DEBUG DEVICE. On a team where three people enable debug mode on the same afternoon, each picks their own device and stops stepping through someone else's events.
One limit worth knowing before you debug a channel problem: Google says DebugView, like the Realtime report, does limited attribution analysis to stay responsive, and recommends the Acquisition reports for accurate attribution. Campaign mapping tested in DebugView can agree with itself and disagree with the report.
When the panel stays empty
An empty DebugView sends people hunting through container tags. Three causes account for most of it.
Consent. Google's own note on the page: events are not visible in debug mode if you have implemented privacy controls on the client side, or if you use consent mode and the user has not consented to Analytics cookies. On a UK site with a CMP that denies analytics_storage by default, a clean browser profile produces exactly this result. Accept the consent banner, or test with a profile where you already have, and the events appear.
The wrong device. Debug mode was enabled on a phone, and the selector is pinned to a laptop that has not sent anything since breakfast.
The tag did not fire. Debug mode was set on one event tag, and that tag's trigger never matched the page you are on. DebugView shows what reached Google, not what you expected to reach it.
Your tests are traffic
Every event generated in debug mode is collected like any other event. Left alone, a week of tag work becomes a week of sessions from one desk, skewing conversion rates, device splits and engagement time in exactly the reports you are about to show a client.
Google's answer is a filter type built for this, not an IP rule. In Admin, under Data collection and modification, open Data filters, click Create Filter, choose Developer Traffic, and set it to Exclude. Once the filter is active, Analytics removes data collected from users while debug mode is enabled.
Details that decide whether it works:
- Name. Unique among the property's data filters, starting with a letter, using unicode letters, numbers, spaces and the
_character, up to 40 characters. - State. Testing, Active or Inactive. Testing assigns matching events to the
Test data filter namedimension with the filter's name as the value, so you can check the rule before trusting it. A free-form exploration with Test data filter name and Event name as rows, event count as the value, and a filter containing your filter name shows what would be removed. Google notes a filter can take 24 to 36 hours to apply. - Role. Editor at property level, for creating, editing and deleting.
- Permanence. An active exclude filter is final. Matching data is never processed and never appears in Analytics or BigQuery, and there is no undo that brings it back.
- Budget. A property holds at most 10 data filters. Internal traffic, developer traffic, hostname and custom condition filters all come out of the same ten, which is why test filters get left in Testing mode for months.
From arrived to counted
DebugView runs on realtime data, which Google describes as typically a few minutes behind and limited to a small set of dimensions. Reports do not.
Standard intraday data typically takes 2 to 6 hours. Daily data takes around 12 hours. Google also states that processing can take 24 to 48 hours, and that numbers in reports may change during that window.
The prior-day timeline, in the property's own time zone, is roughly this: collection stops at midnight, daily data is ready in Explore around 11:30 in the morning, daily BigQuery events around noon, and daily data ready in Reports around 3:30 in the afternoon. Those are typical times, not guarantees.
Two more behaviours explain a lot of late-afternoon panic. Attribution credit for key events can change for up to 12 days after the event is recorded, as Google's modelling improves. And while a property is still on intraday data, event-scoped traffic source dimensions can have temporary gaps, paid and organic last click is used by default, and stricter cardinality limits push more rows into (other).
So an 11am report and the same report at 4pm are not two views of one number. They are two processing states. Comparing them and calling the difference a tracking bug is a mistake that costs a day.
Realtime versus DebugView
They look similar and answer different questions. The Realtime report shows active users over the last 5 and 30 minutes, and its dimensions are user-scoped, meaning they describe how you first acquired the user rather than what the last click did. Its cards cap at 700 rows, so high-cardinality dimensions collapse into (other) and card totals can drift from each other. Google describes the service as best effort, with no SLO.
DebugView is event-level, parameter-level and device-level, for a browser you control. Realtime tells you whether anything is arriving. DebugView tells you whether this one event is correct.
A verification loop that holds
Six steps, in this order, and the Friday ticket does not reopen on Monday.
- Enable debug mode on your own device through Tag Assistant or preview, not for every visitor.
- Check the event name against your agreed taxonomy, then open the event and read the value of each parameter, not its name alone.
- Confirm every parameter the report needs has a registered event-scoped custom dimension. An unregistered parameter is visible here and missing there.
- Confirm the developer traffic filter is Active before a long test session, so your own clicks stay out of the numbers.
- Wait for the processing interval that matches your question, then compare reports. Daily data, or the next morning, for anything involving revenue or attribution.
- Remove the
debug_modeparameter. Do not set it to false.
None of this is exotic. It is the difference between testing that an event left the browser and testing that a business number is right. DebugView is the only view that shows you a parameter value as it arrives, which is why it gets trusted further than it can carry. The rest of the loop costs an afternoon once, and then it is a habit.
Not sure whether an event is missing or just not counted yet? Book a free tracking audit and we will check the event, the registered dimensions and the reporting layer separately, so you know which one is actually broken.