Meta, Google, TikTok and Snapchat can together claim €800,000 in revenue while Shopify has collected only €400,000. Nobody created phantom revenue: several platforms claimed the same orders.
This paradox exposes the central weakness in advertising management. Attribution is not accounting, let alone proof of causality. Yet six-figure budgets shift on last click or ROAS calculated by the platform selling the inventory.
Put Meta Ads, GA4 and your CRM in their proper roles, then measure what investment actually added to the business.
An ad can contribute to a sale without causing it, or cause one without receiving the last click. That distinction is the challenge.
Attribution does not uncover truth: it distributes credit
Marketing attribution assigns a conversion—purchase, lead, subscription or appointment—to one or more touchpoints. The model allocates credit according to its rules, time window and observed data.
| Concept | Question asked | Result |
|---|---|---|
| Attribution | Which touchpoints preceded the conversion? | A distribution of credit |
| Causality | Did advertising cause the conversion? | A cause-and-effect relationship |
| Incrementality | What would not have happened without advertising? | An estimated additional gain |
Confusing these levels rewards the best credit collector, not necessarily the best growth creator.
One sale, six influences and one reported winner
A prospect sees your product on a subway poster, then an Instagram post, opens your newsletter, compares on Google, watches a YouTube unboxing and buys after a Facebook retargeting ad.
Who deserves the sale? The poster built mental availability. Instagram reactivated the brand. Email maintained interest. Google captured explicit intent. YouTube reduced perceived risk. Meta closed the sequence.
Several touchpoints is the reasonable answer, but no tool sees the whole journey. The CRM knows the purchase; GA4 sees some digital paths; platforms see their interactions and the transmitted event. Word of mouth, device changes or store visits can disappear.
Choosing a model is only part of the issue. Understand each referee’s field of view.
Meta, GA4 and the CRM answer different questions
Meta measures what follows a known interaction
Meta attribution settings distinguish:
- click-through : conversion within 1 or 7 days of a link click
- view-through : conversion within one day of a view without a click
- and engage-through : certain ad interactions followed by conversion within one day.
These settings affect reporting and optimisation. Restricting every campaign to 1-day click for caution can exclude legitimate conversions when buying takes several days of consideration.
GA4 reconstructs cross-channel paths
GA4 assigns credit to observed touchpoints before a key event. Attribution Paths shows which channels initiate, assist or close conversions, plus the number of days and touchpoints.
Blind spots remain: inconsistent UTMs, consent, cross-device activity, phone, WhatsApp and offline sales can break the chain.
The back office determines revenue, not influence
Shopify, ERP or CRM is the source of truth for orders, refunds, margins and signed sales. It answers ‘how much did we sell?’, not ‘which channel caused it?’.
Business systems set the total. Attribution tools offer explanations. Tests then investigate causality.
1-day click, 7-day click and views: a window is not a verdict
The 1-day click favours proximity in time. It suits quick actions, simple offers or conservative analysis. It undervalues products compared over days, subscriptions, B2B and high-ticket purchases.
The 7-day click better reflects a longer decision cycle, but also lets Meta claim conversions to which other channels contributed in the meantime.
The 1-day view establishes a weaker link: someone exposed without clicking might buy anyway. But removing all views denies the memory created by images and video. Isolate this share, examine reach, engagement and retargeting, then test incrementality.
Compare Attribution Settings in Ads Manager shows reported conversions by window. ROAS that collapses without views is not automatically false; it rests on a weaker assumption.
This weakness peaks with last click.
Last click rewards whoever closes demand
Last click gives all credit to the last identified channel. Simple and readable, it explains closing. It is dangerous when it alone drives budgets.
Retargeting, promotional email and branded search look powerful because they intercept strong intent. Video, creators, social prospecting, organic content and outdoor ads build demand without final credit.
It is like crediting the last sunray for a flower while forgetting water, nutrients and months of care.
Cutting upstream channels may look sensible at first: retargeting still converts existing demand. Then audiences dry up, branded search declines and the closing channel deteriorates.
First click makes the opposite mistake. Data-driven models allocate credit more subtly but remain statistical constructions. No algorithm recovers signals it never saw.
Leave model battles behind and build an evidence framework.
A robust method: triangulate four levels of measurement
| Level | Question | Tools | Limitation |
|---|---|---|---|
| Accounting | How much did we really earn? | Back office, ERP, CRM, margin, MER | Does not explain the channel |
| Attribution | Which touchpoints came first? | Meta, Google Ads, GA4, UTMs | Credit ≠ causality |
| Incrementality | What would happen without advertising? | Conversion Lift, holdouts, geo-tests | Requires volume and a protocol |
| MMM | How does the mix influence sales? | Econometrics, Robyn | Requires history and expertise |
1. Establish economic reality first
Use net orders, signed sales, cancellations, refunds and margin. Calculate blended CAC and MER—total revenue / total marketing spend—by period and market. These show whether the business progresses, not why.
2. Use attribution to understand journeys
Standardise UTMs (source, medium, campaign, content), compare Meta windows and read GA4 paths. Do not force tools to agree to the cent: explain gaps and identify decisions that withstand model changes.
3. Reserve causality for experiments
Ask: ‘Would this sale exist without the ad?’ Meta Conversion Lift, holdouts and geo-tests compare exposed and control groups.
Where available, Meta offers incremental attribution optimising for conversions predicted to be additional. A useful model, not the equivalent of a correctly sized controlled test.
4. Use MMM for strategic allocation
Marketing Mix Modeling analyses aggregate series and may include media spend, promotions, seasonality and external factors. It needs enough history and variation. With poor data, sophistication yields an illusion of precision.
The Alyads protocol before moving a euro
- Designate the financial source of truth and use net revenue.
- Align currencies, time zones, taxes, returns and cancellations.
- Audit Meta Pixel, Conversions API, events and deduplication.
- Standardise UTMs and campaign naming.
- Document every platform’s attribution windows.
- Break Meta results down into click, view and engagement.
- Analyse journey duration and touchpoints in GA4.
- Reconcile claimed and actual revenue to measure overlap.
- Identify initiators, assistants and closers without freezing their roles.
- Test first the hypothesis likely to shift the most budget.
Never cut a channel just because another steals its last click. Monitor total sales, branded demand, new customers and exposed cohorts.
France and Israel: attribution breaks faster across borders
A France–Israel strategy combines languages, currencies, domains, buying cycles and conversion methods. Someone may discover an offer in French on Instagram, continue in Hebrew on Google and close on WhatsApp. Without CRM feedback, the sale is invisible to the channel that created demand.
Separate France and Israel in CAC, MER and conversion-time analysis. Align currencies and time zones, keep shared UTM naming, but compare each market to its own history. A combined average can hide a profitable country and another absorbing the budget.
At Alyads, attribution is a decision discipline, not a ROAS contest between dashboards. The more languages, devices and offline channels a journey crosses, the less power last click deserves.
FAQ — Marketing attribution and Meta Ads
What is Meta Ads attribution?
Meta assigns conversions to ads according to the selected model, interaction and window. It supports reporting and optimisation but alone does not prove the ad caused the sale.
Why do Meta, GA4 and Shopify show different revenue?
Shopify counts orders. Meta and GA4 allocate credit according to their visibility and models. Multiple platforms may claim the same sale or miss touchpoints.
Can you trust Meta Ads ROAS?
Yes, for comparison and optimisation in a stable framework with reliable tracking. Not as standalone proof of profitability or incrementality. Compare it with net revenue, margin and tests.
Which attribution window should you choose?
Match the decision cycle and optimised action. Compare 1-day click, 7-day click, view and engagement before concluding; none suits every business.
How do you measure incrementality?
Compare exposed and control groups with Conversion Lift, holdouts or geo-tests. Ensure sufficient volume and limit contamination between groups.
Conclusion: stop seeking one winner
Attribution cannot produce perfect truth about a fragmented human journey. It informs decisions with explicit uncertainty.
Establish actual revenue in the back office, use Meta and GA4 to read journeys, then test causality when the budget justifies it. Move from dashboard battles to evidence-led acquisition.
Investing in Meta, Google, TikTok or Snapchat in France, Israel or both? Alyads reconciles ad data with actual revenue and builds multichannel scaling that protects profitability. Contact Alyads for a marketing attribution audit.
