Google Ads GeoX data audit: reconcile API spend with CRM outcomes

A GeoX workflow is not ready merely because a geographic report can be exported. For automated or large-scale GeoX data pulls, Google’s documentation says cost and spend must be pulled through the Google Ads API. More generally, the same guidance says advertisers are expected to use the API for cost metrics. Conversion or revenue outcomes should come from raw internal systems such as a CRM, point of sale or sales database and should be unfiltered and unattributed rather than copied from Google Ads reporting.

The immediate implementation task is therefore reconciliation, not optimisation. Confirm that campaign cost, calendar dates, geographic units and business outcomes describe the same experimental cells. If one side uses cities while the other uses postal-code groups, or if one file contains attributed conversions while the other expects raw outcomes, analysis can be internally consistent yet answer the wrong question.

• Use GoogleAdsService with Search or SearchStream for the API cost extraction.
• Select campaign ID, date, the required geographic segment and cost in micros.
• Join that spend to daily, geographically aligned, absolute non-negative business outcomes.
• Keep separate spend records for every cell in a multi-cell study.
• Reject the dataset until date, geography, value and ownership checks all pass.

Google identifies geographic_view as the reporting resource for the place where a user was located or in which the user showed interest. The documentation says this is generally the default view for GeoX. It identifies location_view as the alternative when the requirement is performance based on locations that the campaign specifically targeted. These are different location concepts and must not be silently substituted.

The documented field pattern contains campaign.id, segments.date, a geographic segment such as city, region or country, and metrics.cost_micros. Google’s example uses a recent-date filter, but that example does not define the correct evaluation window for every experiment. The experiment plan should own the actual period.

Geographic segment fields return resource names that point to GeoTargetConstant records rather than ordinary place names or short codes. Names can be resolved by querying geo_target_constant or by using Google’s static geographic-target file. Cost values are expressed in micros, so Google instructs users to divide them by 1,000,000 for the normal currency unit.

CreatikLab uses the following matrix as an operational acceptance layer. It does not replace Google’s product documentation or a statistical design. Its purpose is to expose mismatches before they become persuasive charts.

• Cost completeness — Evidence: API totals by campaign and day. Failure signal: unexplained gaps or duplicate rows. Action: rerun and reconcile the extraction. Owner: media data analyst.
• Geographic identity — Evidence: GeoTargetConstant mapping and the experiment cell map. Failure signal: unresolved resource names or mixed geographic levels. Action: resolve identifiers and rebuild the join table. Owner: analytics engineer.
• Location meaning — Evidence: recorded choice of geographic_view or location_view. Failure signal: the selected view does not match the approved design. Action: stop analysis and correct the extraction. Owner: experiment lead.
• Outcome integrity — Evidence: CRM, POS or sales export specification. Failure signal: outcomes are attributed, filtered, negative or unavailable by day. Action: regenerate raw absolute values. Owner: revenue operations.
• Temporal alignment — Evidence: date dictionary and timezone convention. Failure signal: shifted days or spreadsheet-reformatted dates. Action: normalise and document the calendar. Owner: data engineering.
• Cell separation — Evidence: spend and outcome files by study cell. Failure signal: cells are pooled before analysis. Action: restore separate records. Owner: measurement lead.

Google requires the external outcome data to be organised as a daily time series, attached to the exact geographic units used by the experiment and expressed as absolute non-negative values. For lead generation, CreatikLab recommends defining each business status before export: submitted enquiry, sales-accepted lead, qualified opportunity or another internally governed milestone. That taxonomy is a company decision, not a Google capability.

A qualification field should have a named owner, a stable definition and a traceable timestamp. The audit should test whether sales teams apply it consistently across experimental geographies. It should also preserve raw counts instead of replacing them with a platform-attributed total.

When an internal system records net revenue with negative refunds, Google directs advertisers to use gross figures for the experiment and apply the refund ratio later. The official guidance does not define the appropriate refund window, lead-scoring model or qualification threshold. Those choices should be documented separately and approved by finance or revenue operations.

Start with an explicit cell dictionary: experimental cell, geographic level, Google resource name, canonical place name and the matching business-system field. Do not infer a country from a campaign name or convert a city into a sales territory without a reviewed mapping rule. Geographic precision should be identical on both sides of the join.

Postal codes need special handling because leading zeroes can disappear in spreadsheet software. Google also warns that dates can be reformatted automatically and recommends preserving the required YYYY-MM-DD representation. Treat both fields as controlled schema elements, not presentation formatting.

For targeting changes, Google requires the exact GeoTargetConstant identifier. ISO country codes and legacy Nielsen-style codes are not substitutes. Keep targeting operations separate from reporting transformations so an analyst cannot accidentally write a display label back into a campaign configuration. The documentation does not specify who may make those changes; assign approval through your own account-access policy.

The measurement file should include a data dictionary, source system, extraction timestamp, campaign identifier, day, geographic identifier, study cell, converted cost and the selected raw outcome. Add validation fields that show whether geography resolved, whether the row belongs to one cell and whether an outcome was available. These additions are CreatikLab controls, not extra Google Ads product requirements.

Track four acceptance indicators without inventing a universal benchmark: extraction completeness, geographic mapping coverage, spend reconciliation variance and outcome availability by day. Report the numerator, denominator and exceptions for each indicator. A green label without the underlying counts is not sufficient evidence.

Our decision rule is strict: do not interpret experimental impact while a material row is unmapped, duplicated, assigned to multiple cells or based on an incompatible outcome definition. The owner may accept a known exception only when its scope, reason and expected effect are recorded. Google’s documentation does not provide a performance guarantee, statistical significance threshold or expected lift.

1. Experiment lead: record whether the design requires user location or interest, or explicitly targeted locations; retain the approved rationale.
2. Google Ads developer: run Search or SearchStream against the selected view and retain the GAQL, account context and extraction log.
3. Media analyst: verify campaign IDs, daily rows and separate spend for each experimental cell.
4. Analytics engineer: resolve geographic resource names through GeoTargetConstant data and preserve the mapping table.
5. Data engineer: convert cost micros once, record the transformation and prevent a second conversion downstream.
6. Revenue operations: export raw, unfiltered and unattributed daily outcomes at the matching geographic level.
7. Finance or sales operations: approve the outcome definition and the treatment of refunds or rejected leads.
8. QA owner: test date formatting, leading zeroes, duplicates, missing rows, negative values and cross-cell contamination.
9. Measurement lead: issue a signed readiness decision plus an exception register before analysis begins.

The checklist deliberately separates production from approval. A developer can prove that a query ran successfully, but cannot alone certify that a sales outcome represents qualified demand. Likewise, a CRM owner can define qualification but cannot confirm that the Google reporting view matches the experimental location concept.

Do not assume that geographic_view means physical presence only; Google describes it as location or demonstrated interest. Do not assume location_view answers the same question, because it concerns targeted locations. Do not treat a canonical place name as a valid targeting identifier, and do not send an ISO or Nielsen code where a GeoTargetConstant is required.

Do not assume platform-attributed conversions can replace the requested raw business outcomes. Do not net refunds into negative experimental values when Google calls for absolute non-negative totals. Do not merge multi-cell spend before delivery, and do not trust a spreadsheet preview as proof that postal codes or dates survived.

Google’s page does not specify GeoX pricing, account eligibility, rollout geography, minimum spend, minimum sample size, statistical method or expected performance. It also does not validate your lead-quality taxonomy. Those questions require separate product confirmation, experimental design and business governance.

Ask each provider to show the proposed extraction schema, the distinction between geographic_view and location_view, the GeoTargetConstant resolution method and the CRM outcome contract. A credible scope should identify who owns campaign data, geographic mappings, qualification status, financial treatment and final acceptance. A dashboard alone is not an audit deliverable.

For lead-generation work, require a concrete definition of a qualified lead and an exception report showing missing, rejected or geographically unmapped outcomes. Compare providers on inspectable artefacts: saved query logic, transformation log, reconciliation table, cell-level files, data dictionary, QA results and approval record. Do not compare them on promised lift, ROAS or lead volume because neither this workflow nor Google’s documentation guarantees those results.

Also ask how corrections are rerun. A reliable process should preserve versioned inputs and make it possible to identify which rows changed, why they changed and who approved the revision.

Begin with a Google Ads account audit covering the GeoX extraction, location-view choice, geographic mapping, cost conversion, CRM outcome contract, cell separation and readiness decision. The concrete deliverable should be a reconciled extraction specification, validation results and exception register, not a performance promise.

If the audit reveals a design or governance decision, use senior Google Ads consulting as the next step. The Google Ads Expert route shows how a senior specialist can support implementation and review.

To continue the diagnosis with context, tell Lia which geographic units, campaigns, business outcomes and internal systems are involved. Include whether the current export uses user geography or targeted geography, and describe any known date, refund or qualification issue.

Which Google Ads data should be pulled for GeoX?

Google documents campaign, date, geographic segment and cost as the core API inputs. Cost is returned in micros and must be converted into the normal currency unit before preparing the upload or analysis file.

Should attributed Google Ads conversions be used as the outcome?

Google instructs advertisers to combine API cost with raw, unfiltered and unattributed conversion or revenue data from an internal CRM, point of sale or sales database. The business system therefore needs its own auditable outcome definition.

What is the difference between geographic_view and location_view?

Google describes geographic_view as reporting where a user was located or showed interest, which is generally the GeoX default. location_view instead reports performance for geographic locations explicitly targeted by a campaign.

Can ISO country codes or Nielsen codes be used for targeting changes?

No. Google requires the relevant GeoTargetConstant identifier when changing geographic targeting. Raw ISO or Nielsen codes are not accepted substitutes for that operation.

Why can spreadsheet processing damage the file?

Spreadsheet software can reinterpret dates and remove leading zeroes from postal codes. The audit should preserve dates in YYYY-MM-DD form and keep postal codes as fixed text-like values.

What should a GeoX data audit deliver?

A useful audit should provide an extraction specification, field and geography mapping, spend reconciliation, CRM outcome contract, exception register, ownership plan and a documented decision on whether the dataset is ready for analysis.