Ecommerce & Retail · Growth problem
Marketing attribution is unclear
Every paid platform, GA4, and CRM system typically claims credit for more revenue than the business actually generated, because each is built to report favourably within its own attribution logic. The aim isn't to build a perfect, unified model of every touchpoint — that's rarely achievable or necessary at this scale — it's to build a measurement approach that's good enough to make the next set of spend decisions with confidence.
Symptoms
What this usually looks like
- Summing reported revenue across Meta, Google and email exceeds total actual revenue by a wide margin
- Different stakeholders in the business quote different numbers for the same channel's performance
- GA4 and platform-reported figures disagree significantly and no one is quite sure why
- Budget decisions are made based on whichever platform's dashboard was reviewed most recently
- Post-purchase survey data ('how did you hear about us') doesn't match platform attribution
Diagnostic questions
What we would test first
- Reconcile summed platform-reported revenue against actual finance-confirmed revenue for a defined period
- Compare GA4 channel attribution against platform-reported figures and identify the size and pattern of discrepancies
- Pull an assisted-conversion or multi-touch view alongside last-click for the top 5 channels
- Cross-check post-purchase survey attribution data against platform-reported channel mix
- Agree a single blended reporting definition and source of truth with all relevant stakeholders
- Identify which decisions currently rely on unreconciled platform data, and prioritise fixing measurement for those first
Root causes
Why it happens
- 01
Overlapping attribution windows across platforms
Meta, Google and email each use their own attribution windows and models, meaning the same purchase can be claimed as a conversion by more than one channel simultaneously — the sum of platform-reported revenue will therefore almost always exceed actual total revenue.
- 02
GA4 configuration and sampling limitations
GA4's default settings, data-driven attribution model, and sampling behaviour at higher traffic volumes can produce figures that don't reconcile cleanly with either platform-reported data or finance-confirmed revenue, without any of the three necessarily being 'wrong'.
- 03
No agreed source of truth for blended reporting
Where each function (paid media, SEO, CRM) reports its own numbers independently without a shared reconciliation process, the business ends up with several competing versions of performance rather than one agreed view leadership can act on.
- 04
Over-reliance on last-click without an assisted-conversion view
Last-click attribution systematically under-credits channels that play an earlier, assisting role in the journey — typically content, SEO and upper-funnel paid — which distorts channel-level investment decisions if used as the only lens.
- 05
Post-purchase survey data collected but not reconciled against platform data
Survey-based attribution ('how did you hear about us') offers a genuinely useful, platform-independent signal, but is rarely combined systematically with platform and GA4 data to sanity-check the overall picture.
Evidence
The numbers we would look at
These are the metrics that make the constraint visible, and the cuts that stop them being reassuring by accident.
| Metric | What it tells you | How we read it |
|---|---|---|
| Blended vs platform-summed revenue reconciliation | Quantifies the scale of double-counting across channels, which is the first step to a usable view.Track the gap over time; a growing gap suggests worsening overlap or attribution drift. | Track the gap over time; a growing gap suggests worsening overlap or attribution drift. |
| New vs returning customer revenue split (finance-confirmed) | A relatively attribution-independent way to check overall channel health.Reconcile against order management/finance system, not marketing platforms. | Reconcile against order management/finance system, not marketing platforms. |
| Post-purchase survey attribution vs platform-reported attribution | An independent cross-check on where customers say they came from.Compare distribution by channel, not just headline top channel. | Compare distribution by channel, not just headline top channel. |
| Assisted vs last-click conversion share by channel | Shows which channels are under- or over-credited by a last-click-only view.Use to inform investment weighting, not as a replacement metric. | Use to inform investment weighting, not as a replacement metric. |
| Contribution margin by channel (post-reconciliation) | The commercially meaningful figure once double-counted revenue has been addressed.This is the number that should inform budget allocation, not platform ROAS. | This is the number that should inform budget allocation, not platform ROAS. |
Measurement traps
What can mislead you
- Looks fineGA4 shows different numbers to our ad platforms, so GA4 must be wrong
- GA4 and platform dashboards use different attribution models and data collection methods by design; disagreement between them is expected and doesn't indicate either is faulty — it indicates neither should be treated as the sole source of truth.
- Looks fineSumming all channel-reported revenue gives us total marketing-driven revenue
- This figure will almost always overstate actual revenue because of overlapping attribution windows across channels — it should never be compared directly to total company revenue.
- Looks fineLast-click data shows this channel drives little revenue, so we should cut it
- A channel with low last-click credit may still be playing a significant assisting role earlier in the customer journey; cutting it without checking an assisted-conversion or incrementality view risks removing a genuinely productive part of the funnel.
Outcome
What better looks like
Not a promised number. A clearer basis for the next investment decision.
- The business has one agreed, reconciled view of channel performance that all functions use, rather than several competing versions
- Budget decisions are made against contribution margin and a blended/assisted view, not the most favourable platform dashboard
- Discrepancies between GA4, platform and finance data are understood and explained, not treated as a mystery to be ignored
- Measurement is treated as decision-useful rather than an attempt at perfect attribution, which is rarely achievable or necessary
Where a Growth Diagnostic would start
A three to four week senior review across demand, discovery, acquisition, conversion, retention, measurement and capability — sequenced so this problem is either confirmed as the constraint or ruled out early. Read alongside the ecommerce & retail model page for how we frame the wider system.
Related problems
Ecommerce & Retail
Paid media dependency
Growth only happens while spend keeps rising. A look at how much of that spend is creating demand versus capturing it.
Ecommerce & Retail
Rising customer acquisition costs
The cost of winning a new customer keeps climbing, even where individual platforms report healthy performance.
Ecommerce & Retail
SEO traffic isn't driving revenue
Organic visibility looks strong but it isn't translating into sales. A look at intent, landing pages and product availability.
Questions about this problem
- Do we need a full marketing mix model to fix this?
- Usually not as a first step. Most of the immediate confusion can be resolved with reconciliation work across existing platform, GA4 and finance data; MMM or formal incrementality programmes tend to be worthwhile only once spend and complexity justify the investment.
- Who needs to be involved from our side?
- Typically whoever owns analytics/GA4, the paid media function or agency, and someone from finance who can confirm actual revenue figures — the reconciliation work depends on having all three perspectives in the room.
- Will this tell us exactly what each channel is worth?
- It will give a considerably more reliable, decision-useful view than platform dashboards alone, generally with a reconciled blended figure and an assisted-conversion read — genuinely precise per-channel attribution isn't fully achievable with typical data, and claiming otherwise would overstate what's possible.