The Melrose Arch Meeting Where Three Dashboards Disagreed (Hypothetical Scenario)
Picture a marketing director at a Johannesburg consumer brand opening a Monday finance review. The Ads UI shows R412,000 of attributed revenue for the month. GA4 shows R368,000. The CRM shows R341,000. Nobody is hacking the account. All three numbers are technically honest, and none of them can be reconciled quickly because the exports behind each dashboard were pulled on different days, from different surfaces, with different definitions of a conversion. The phrase "google account data export tool" implies a single switch that dumps everything into one tidy file. In practice, a Google account is a federation of surfaces, each with its own export mechanism, its own owner, and its own idea of what a session or a conversion is.
The Five Export Surfaces, Side by Side
Before comparing vendors, map what you are actually exporting. South African teams typically touch five surfaces:
- Google Takeout - a personal-account archive covering Gmail, Drive, Calendar and more. Useful for data-subject requests and offboarding, not for marketing reporting.
- Google Ads - campaign, cost and conversion data pulled via the Ads API or scheduled reports. Versioned and retired on a rolling schedule.
- GA4 - event-level data, best exported to BigQuery rather than CSV, because sampling and thresholds distort large pulls.
- Search Console - query and page performance, available through the bulk export to BigQuery or the API for smaller volumes.
- Workspace - admin-level exports of org data, governed by admin roles and audit logs rather than marketing definitions.
Each surface has a different format, refresh cadence and access model. A tool that handles Ads well may not touch Search Console bulk export at all. That is why the buying question is not "which tool exports Google data" but "which surfaces does it cover, and how does it reconcile them".
Why the Numbers Diverge: Definitions, Timing and Consent
The Melrose Arch gap is rarely a bug. Ads reports at click time with the longest attribution window. GA4 applies consent-mode modelling and session-based deduplication, so a user who converts across two sessions may be counted once or modelled rather than observed. The CRM reflects refunds and returns that were never back-fed to Google. Add time-zone mismatches - a property set to UTC versus a team reporting in SAST - and a month boundary alone can shift a day of revenue. Documenting these causes is the difference between a reconciled number and a monthly argument.
Vetting a Third-Party Export Tool Against POPIA
Under POPIA, exporting personal data to a third-party tool makes that vendor an operator, and section 21 requires a written contract with security safeguards. Ask four questions before signing: which OAuth scopes does the tool request (read-only versus write access); where does the data land geographically; does it retain raw event data and for how long; and can it answer a data-subject access or deletion request within a reasonable window? A tool that cannot answer these in writing is a compliance risk, not a shortcut.
What to Compare on the Shortlist
Score candidates on surface coverage (all five or a documented subset), export destination (cloud storage versus local download), scheduling and retry behaviour, credential handling, and total annual cost in ZAR including FX and VAT. A cheaper monthly headline can lose on the annual figure once connector fees and BigQuery storage are added.
The Export Playbook, Stage by Stage: From Baseline Inventory to Monthly Reconciliation
Run this as a system, not a heroic pull.
- Inventory. List every Google surface holding revenue-relevant data, who owns each property, and its time-zone setting.
- Baseline. Save one approved export of each surface on the same day and label it clearly.
- Warehouse. Land raw files in cloud storage before transformation - export to a cloud destination rather than downloading multi-GB archives over an unstable link, and schedule off-peak to survive load-shedding.
- Reconcile monthly. Agree one revenue number per month across Ads, GA4 and the CRM, with a documented reason for every gap.
- Govern. Record who holds API credentials and review version currency each quarter.
Flag: Google Ads API versions retire on a rolling schedule. Check the current version before building - never hard-code a version you found in an old tutorial.
A Worked Reconciliation Table: Why R412,000 and R341,000 Describe the Same Month (Hypothetical)
| Source | Revenue | Likely cause of gap |
|---|---|---|
| Google Ads | R412,000 | Reported at click time, longest attribution window |
| GA4 | R368,000 | Consent-mode modelled conversions, session-based dedup |
| CRM | R341,000 | Refunds and returns not back-fed to Google |
None of these is "wrong". The system is only working once everyone agrees which figure governs the P&L and why the others differ.
What an Export Query Actually Looks Like: An Illustrative GA4 Data API Call
This is an example, not a production script:
POST https://analyticsdata.googleapis.com/v1beta/properties/PROPERTY_ID:runReport
{
"dateRanges": [{"startDate": "2025-02-01", "endDate": "2025-02-28"}],
"dimensions": [{"name": "sessionDefaultChannelGroup"}],
"metrics": [{"name": "purchaseRevenue"}]
}Note the explicit date range and the single dimension-metric pair. Production pulls add pagination, currency handling and a landing step into the warehouse. The point is that every export is a query with assumptions baked in - the date range, the property, the metric definition - and those assumptions must be recorded alongside the file.
Cost and Effort Reality in ZAR: Ranges, FX and VAT Caveats
BigQuery and most connectors bill in USD, so a ZAR budget carries FX exposure every month. South African VAT may apply to imported digital services at 15% - confirm with your accountant; treat it as a cost line, not settled law. BigQuery storage for GA4 event-level data grows with traffic, so estimate before you enable it. Connector fees are usually subscription-based; request the annual figure, not the monthly headline.
Tip: our reporting approach starts from reconciliation, not dashboards - worth reading before you scope tooling.
Outcomes and How to Measure Whether the Export System Is Working
Measure four things: variance between surfaces each month (should be explained, not eliminated); time to produce the monthly pack; whether anyone re-runs an export mid-month to "fix" a number; and how fast a data subject request can be answered. For a wider view of warehouse-backed reporting in practice, see the case studies.
