Google Ads GeoX data audit: prepare geographic experiments without corrupting the inputs

A Google Ads GeoX workflow should begin with an input audit, not with a conclusion about campaign impact. Google documents that automated or large-scale GeoX data preparation requires Google Ads API cost and spend data to be combined with raw, unfiltered and unattributed conversion or revenue records from an internal CRM, point-of-sale platform or sales database. That distinction matters: platform-attributed conversions are not the outcome input Google recommends for this workflow.

CreatikLab’s operational interpretation is that the experiment can only be as defensible as the join between spend, time, geography and the business outcome. A click or form submission may be useful operationally, but it should not automatically be called a qualified lead. Before implementation, the advertiser should document the outcome definition, geographic resolution, reporting day, currency handling and data owner. This article concerns GeoX geographic experimentation; it does not use GEO to mean generative engine optimization. Google’s documentation does not promise lift, lead volume or a return on ad spend, so neither should an implementation partner.

Google identifies GoogleAdsService and its Search or SearchStream methods for retrieving campaign and geographic cost records. For a typical GeoX extraction, the documented fields include campaign.id, segments.date, an appropriate geographic segment such as segments.geo_target_city, and metrics.cost_micros. The geographic level can instead be a region or country when that matches the experiment design.

The reporting view must also match the question. Google describes geographic_view as the route for the user’s physical location or location of interest and notes that it is normally used for GeoX. By contrast, location_view reports performance for locations specifically targeted by the campaign. Geographic fields return GeoTargetConstant resource names rather than convenient labels such as a city name or country abbreviation. Those resources can be resolved through geo_target_constant or Google’s downloadable geographic target data.

Two transformations are explicit. API cost is returned in micros and must be divided by 1,000,000 for a standard currency value. Targeting changes require valid GeoTargetConstant IDs; raw ISO or Nielsen codes cannot simply be submitted in their place. These are data-contract requirements, not optional reporting preferences.

The following matrix is a CreatikLab control framework, not a Google product claim. Each row links an observable symptom to evidence, an action and an accountable owner.

• Spend does not reconcile — Evidence: API totals differ from the controlled account extract — Action: inspect campaign scope, dates and cost transformation — Owner: paid-media analyst.
• Geographies appear as resource strings — Evidence: unresolved GeoTargetConstant values — Action: create and version a lookup table — Owner: data engineer.
• Postal areas lose leading zeros — Evidence: stored values are shorter than the source records — Action: enforce geographic IDs as text before import — Owner: analytics engineer.
• CRM outcomes cannot be assigned to the test unit — Evidence: records lack the required city, region, postal grouping or other agreed key — Action: repair collection or redesign the geographic resolution — Owner: CRM owner.
• Qualified leads disagree with sales records — Evidence: marketing and sales use different status definitions — Action: approve one outcome dictionary and freeze it for the analysis window — Owner: revenue operations.
• One treatment cell contains combined spend — Evidence: no separate cost series exists for each cell — Action: split extraction and reconciliation by treatment group — Owner: experiment lead.

Start with a one-page contract shared by paid media, analytics, CRM and finance. It should name the account and campaign scope, timezone, daily date field, currency, geographic unit, mapping-table version, outcome source and permitted corrections. Google specifies a daily time series for conversion data and requires each outcome to map to the exact geographic unit used by the experiment. The documented outcome values must be absolute and non-negative, such as gross revenue or a total conversion count.

If the CRM records net revenue with negative refunds, Google instructs advertisers to use gross figures for the test and apply the refund ratio later. CreatikLab would therefore preserve gross sales and refunds as separate governed fields rather than silently replacing one with the other. For lead generation, create similarly explicit fields for submitted lead, validated lead and sales-qualified lead, but select only the approved experiment outcome. Google does not define CreatikLab’s qualification stages; those are business rules that need named owners and reproducible evidence.

Do not approve the pipeline because a file was produced. Approve it when every control below has inspectable evidence and a responsible owner.

1. Scope — Save the included customer, campaigns and analysis period — Confirm that the API query matches the approved scope — Owner: account lead.
2. Query — Archive the selected fields and reporting view — Review whether geographic_view or location_view answers the intended question — Owner: Google Ads specialist.
3. Cost — Retain the raw metrics.cost_micros field and transformed value — Test division by 1,000,000 and reconcile totals — Owner: data analyst.
4. Geography — Store the resource name and resolved canonical label — Validate every mapping against GeoTargetConstant data — Owner: data engineer.
5. Dates — Preserve YYYY-MM-DD values through export and upload — Prevent spreadsheet software from silently reformatting them — Owner: analytics engineer.
6. Leading zeros — Test representative postal identifiers — Force text or an appropriate fixed-length format where required — Owner: data steward.
7. Outcomes — Export raw, unfiltered, unattributed CRM, POS or internal sales values — Reconcile them to the system of record — Owner: CRM or finance owner.
8. Cells — Produce a separate spend series for every treatment group in a multi-cell study — Reject combined files — Owner: experiment lead.
9. Approval — Record exceptions, remediation and sign-off — Do not release the dataset while a material join or reconciliation issue remains open — Owner: measurement lead.

CreatikLab recommends three measurement layers. The first is pipeline integrity: query success, complete daily coverage, mapping coverage, duplicate rate, unresolved geographies and reconciliation differences. These metrics show whether the dataset can be trusted; they do not indicate campaign success. The second is the experiment outcome, selected before analysis from the first-party system. For a lead-generation business, that might be a validated or sales-qualified lead if the CRM applies the definition consistently and the record can be assigned to the tested geography.

The third layer is commercial interpretation after the experiment: gross revenue, accepted opportunities, refunds and other downstream evidence available from the business system. Keep these fields distinct rather than collapsing them into a single platform conversion. Document the status rules, timestamp used for the daily series, late-arriving records and permitted restatements. Google’s page explains the required input structure but does not specify a universal qualified-lead definition, analysis threshold, budget, duration or expected performance. Those decisions require an experiment design appropriate to the advertiser’s data and commercial cycle.

Scenario A is ready for implementation: daily spend reconciles, geographic resource names resolve, first-party outcomes map to the same units, values follow the approved gross or count definition, and every treatment cell is separate. The decision rule is to proceed to design review while preserving the audit trail.

Scenario B needs remediation: spend is complete, but CRM geography is missing for a material share of outcomes. The decision rule is not to infer locations from campaign targeting. Repair the first-party capture, choose a defensible coarser geographic unit or pause the experiment plan.

Scenario C should be rejected: the only outcome available is attributed Google Ads conversion data, dates were reformatted without a trace, or multiple cells share one spend series. These conditions conflict with the documented input approach or remove necessary controls. Rejection is not evidence that GeoX cannot work; it means the current dataset cannot support a responsible implementation. CreatikLab would log the blocking evidence, remediation owner and retest condition.

• Do not assume targeted location and user location answer the same question; the reporting view must follow the experiment design.
• Do not treat a GeoTargetConstant resource name as a readable or submission-ready geographic code without resolving it correctly.
• Do not assume spreadsheet software preserves dates or leading zeros. Test the final uploaded representation.
• Do not substitute platform-attributed conversions for the first-party outcome structure Google describes for GeoX.
• Do not mix net revenue containing negative refunds into a field expected to hold gross, non-negative values.
• Do not combine treatment-cell spend when separate series are required.
• Do not call every form fill a qualified lead. Qualification is a governed business definition, not an automatic Google Ads property.
• Do not infer availability, price, eligibility, sample requirements, rollout scope or likely performance. Google’s documentation cited here does not specify those points.

Automation can retrieve and transform records, but human accountability remains necessary for scope approval, geographic meaning, outcome definition and exception handling. A technically valid query cannot decide whether a sales status is commercially meaningful.

The primary next step is a Google Ads account audit focused on GeoX readiness. The concrete deliverable should include the extraction specification, view selection, geographic lookup design, date and currency transformations, CRM outcome dictionary, reconciliation report, unresolved-data register, responsibility map and a proceed, remediate or stop recommendation. It should not promise lift, ROAS or lead volume.

If the audit exposes a design choice rather than a data defect, use senior Google Ads consulting to decide the geographic unit, first-party outcome and governance model. Review the Google Ads Expert route when comparing the level of strategic ownership available. To continue the diagnosis conversationally, tell Lia which accounts, geographic units, CRM fields and outcome definitions are currently available. That context is more useful than a generic request to run an experiment.

What data does Google describe for a GeoX experiment?

Google describes Google Ads cost data segmented by date and geography, combined with raw, unfiltered and unattributed conversion or revenue data from an internal CRM, point-of-sale system or sales database.

Should attributed Google Ads conversions be used as the experiment outcome?

Google’s GeoX guidance says the conversion outcome should come from an internal or third-party system rather than using attributed conversions from the Google Ads API. The business must still define which first-party outcome represents a qualified lead.

What is the difference between geographic_view and location_view?

Google identifies geographic_view for reporting the user’s physical location or location of interest and describes it as the usual choice for GeoX. location_view reports performance for specific locations targeted by a campaign.

Why do cost values need transformation?

The API returns cost through metrics.cost_micros. Google instructs users to divide that value by 1,000,000 when a standard currency value is required for the upload file.

Can ISO country codes or raw Nielsen codes be submitted for targeting changes?

No. Google says targeting changes require the appropriate GeoTargetConstant identifier. Raw ISO country codes and raw Nielsen codes are not accepted substitutes for that identifier.

What does a CreatikLab GeoX readiness audit deliver?

The audit maps extraction queries, geographic identifiers, date and currency transformations, CRM outcome definitions, reconciliation controls, owners and acceptance tests. It identifies whether the data is ready for an experiment without promising a particular result.