Reporting surfaces
| Surface | What it is for | Granularity | Delivery | Limits to know |
|---|---|---|---|---|
| Instant reporting | Ad-hoc pivots and charts inside the console | Aggregated by chosen dimensions | On screen, export to file | A documented maximum date range and row count |
| Offline reports | Defined report types run on a schedule and shared | Aggregated, with more dimensions and report types than instant | File download, email, scheduled runs | Report types differ in the dimensions and metrics they allow |
| Looker Studio connector | Standard dashboards without a pipeline | Aggregated | Live connection | Connector fields and history follow what the connector exposes |
| BigQuery transfer | Reports landed in a warehouse on a schedule | Aggregated report rows | Scheduled tables in BigQuery | Set up through the BigQuery Data Transfer Service |
| Data Transfer files | Event-level logs of impressions, clicks and conversions | Per event | Files in a Google Cloud Storage bucket | Enabled by Google on request; identifiers subject to privacy rules |
| Ads Data Hub | Privacy-safe queries over event-level data | Aggregated outputs from event-level inputs | BigQuery results | Aggregation thresholds enforced on every query |
| Bid Manager API | Creating and fetching offline reports programmatically | Same as offline reports | Files via API | Quotas and report expiry documented by Google |
Instant reporting is the right first stop for "what happened yesterday" and for exploring a problem across dimensions. Offline reports are the right home for anything that has to be repeated, shared with a client or reconciled against an invoice. The exports exist for joins with data DV360 does not hold.
Dimensions and metrics that matter
Dimensions follow the hierarchy: partner, advertiser, campaign, insertion order, line item, and for YouTube line items the ad group and ad, each with a name and an ID. Below those sit creative, creative size, exchange, inventory source, device type, environment, browser, country, region and city, audience list, and time at day or finer. Because the names are editable and the IDs are not, reports that will be joined to anything outside DV360 should carry the ID columns.
Metrics divide into delivery, cost, quality and outcome. Delivery is impressions, billable impressions, clicks and video events. Cost is where care is needed: DV360 reports media cost, which is what exchanges charged; partner costs and data fees as their own lines; and a revenue figure that represents what the advertiser is billed including fees. The Help Centre's metric definitions page is the reference for exactly what each cost metric includes, and the choice of cost metric changes every derived rate, including cost per thousand impressions (CPMCost per 1,000 impressions.) and cost per conversion. Quality covers Active View viewability metrics and verification data where a vendor is integrated. Outcome covers Floodlight conversions, split by post-click and post-view, with their lookback windows set on the Floodlight configuration.
The measurement plan should name which cost metric, which conversion window and which report type each KPIA metric tied directly to an objective. is read from, so that two people pulling the same number get the same number. Most reporting disputes inside a team are two valid metrics with different definitions.
Reach and frequency reports
Reach reporting in DV360 is a separate report type rather than a pair of columns, because reach cannot be summed across rows the way impressions can. The report returns, for a chosen level and date range, the modelled number of people or households reached and the average frequency, plus a frequency distribution that shows how many were reached once, a few times or many times. Google documents how reach is modelled across cookies, device identifiers and environments where neither exists, and the method has changed along with the report's naming; older documentation refers to unique reach reporting, and the current report type and its metrics are listed on the Help Centre's reach reporting page, which is the place to confirm names before building a template.
Two rules follow. Never add reach figures from two reports, two levels or two date ranges, because the people overlap; run one report at the level the question is asked at. And read the frequency distribution rather than the average: an average frequency that looks healthy can hide a long tail of people served many times, which is the figure the frequency cap was meant to control. For the cross-platform problem, reach and frequency across DV360, Google Ads, Meta and television, the reach and frequency across platforms guide takes over; this page stays with what DV360 alone can report.
Viewability and brand safety reports
Active View is Google's viewability measurement and it is built into DV360 reporting. The metrics separate impressions that were measurable, those that met the viewability standard, and the resulting share, along with average time in view and, for video, audibility and completion. Report them per creative size, per exchange and per inventory source, because viewability is a property of the placement far more than of the creative, and a line item's average hides which sources are pulling it down. Pre-bid viewability targeting on the line item uses the same measurement to filter requests before bidding, so a report that shows poor viewability on a line item with a strict threshold usually indicates the threshold is predicted rather than guaranteed. When those columns show a rate below the plan's target, the low viewability diagnosis page uses the same breakdown to separate an unmeasurable share from a placement, creative or bidding cause.
Brand-safety reporting comes from the verification report type, under whatever name the Help Centre's report types page currently gives it, and it shows how served impressions were classified against digital content labels, sensitive categories and, where a vendor is integrated, the vendor's own categories. It is the report to run when a client asks where ads appeared, alongside a site and app report at URL level; the tutorial on responding to a brand safety incident shows where these reports fit in tracing a placement complaint. The standards behind these metrics, including what the Media Rating Council accredits and how vendors differ, are covered in the programmatic measurement section; the DV360 part is knowing that these are separate report types with their own dimension sets.
Scheduling and delivery
Offline reports can be run once or on a daily, weekly or monthly schedule, with a date range that is fixed or relative to the run date, and delivered as a file to download and by email to named recipients. The Help Centre documents the file formats, how long a generated report stays available before it expires, and the limits on the number of scheduled reports and rows, which are the figures to check before promising a client a report at a particular size. A scheduled report with a relative date range is the backbone of most weekly client reporting: one template per report type, run the morning after the period closes, landed in a shared folder.
Two habits keep schedules honest. Name reports for their consumer and their cadence rather than for their contents, so that a report somebody stopped reading can be retired. And keep a change log for each template, because adding a dimension changes the row grain and silently breaks whatever was summing the file downstream. The Campaign Manager 360 reporting guide describes the same discipline on the ad-server side, where the figures should reconcile with DV360's within the documented tolerances.
Getting data out: BigQuery and dashboards
For a dashboard, the shortest route is the native Display & Video 360 connector in Looker Studio, which reads reporting data without an export. The Looker Studio data sources guide covers how connectors behave, including caching, field limits and the moment a connector stops being enough. That moment usually comes when DV360 data has to be joined with the ad server, the web analytics property or the finance system, or when history has to be kept beyond what the connector can see.
At that point the data goes into a warehouse. Google lists Display & Video 360 among the sources of the BigQuery Data Transfer Service, which lands scheduled report data in BigQuery tables without a pipeline to maintain; the Help Centre and the BigQuery documentation describe the setup and what the transfer contains. For event-level data, Data Transfer files deliver impression, click and activity logs to a Google Cloud Storage bucket on a schedule, from where they can be loaded into BigQuery; Google enables the feature on request and documents the file schemas and the privacy restrictions on the identifiers they contain. Ads Data Hub answers user-level questions without exposing user-level rows by running queries inside Google's environment and returning aggregates above a threshold. The habits for working with any of these, partitioned tables, cost control and daily table conventions, are the same ones the GA4 BigQuery export basics guide teaches, and a team that has done one will find the other familiar. The programmatic campaign management playbook puts these reports to work in a monthly reconciliation with the ad server and verification, and in the analysis that closes each flight.
A weekly reporting pack for a retail client
An agency runs five DV360 insertion orders for a homeware retailer with a monthly budget of £85,000 and builds the client's weekly pack from three scheduled offline reports. A standard report at line item grain with impressions, clicks, media cost, data fees, revenue and post-click purchases runs every Monday for the previous seven days; a reach report at campaign level runs on the first of each month; and a verification report at inventory source grain runs weekly. The reports land in a shared drive and a BigQuery transfer loads the same standard report nightly into a table that a Looker Studio dashboard reads. In week 3 the dashboard shows revenue of £21,400 against media cost of £17,900, the gap being £2,100 in data fees on two third-party segments and the platform fee, and the reach report shows 1.4 million people at an average frequency of 4.2 with 11 per cent of them above 10 exposures, which prompts a tighter campaign-level cap.
Common mistakes
- Mixing cost metrics. Media cost, total media cost and revenue answer different questions; pick one per rate and write it down.
- Adding reach across rows. People overlap; run one reach report at the level the question is asked.
- Reading the average frequency alone. The distribution shows the tail the cap was meant to prevent.
- Reporting viewability per line item. It is a property of the placement; report it per exchange, source and size.
- Scheduled reports nobody reads. Name them for the consumer and retire the ones that stop being opened.
- Changing a template's grain without telling downstream. Added dimensions break every sum built on the old file.
- Comparing DV360 to the ad server without tolerances. Small gaps are normal and documented; investigate only the ones that grow.
The DV360 hub places reporting beside the structure, audience, inventory, bidding and creative decisions it measures; the structure guide in particular decides whether these reports read like the plan.
