Which attribution model should I choose in Morocco in 2026?

There is no single "best" model, and relying on just one is the main mistake. In practice, systematically compare last-click and GA4's data-driven side by side: their gap reveals the discovery channels last-click underestimates. Use data-driven for trade-offs between channels within one platform, but never steer your overall budget on a platform number. At the global level, track blended ROAS (MER = total revenue ÷ total ad spend), which is anchored in your accounting and cannot be inflated. Finally, for big budget decisions, validate with an incrementality test. The right model is this combination — not a binary choice between last-click and data-driven.

Why do Meta, Google and GA4 show different numbers?

Because each platform uses a different attribution window and sees only its own data. Meta counts by default over 7 days after click and 1 day after view; Google Ads over 30 days after click; GA4 applies a cross-channel data-driven model with its own session; TikTok counts differently again. The same sale can therefore be claimed by several platforms at once. When you add up the ROAS displayed, you often reach 130 to 160% of real back-office sales — structural double, even triple, counting. Every platform has an incentive to claim maximum credit. The only solution is to reconcile against a single source of truth (back-office or CRM) and steer by blended ROAS.

How do I measure the incrementality of a campaign?

Incrementality measures the sales genuinely generated by a campaign — those that would not have happened without it — not just the attribution credit. Two methods are accessible in Morocco in 2026. The geo-test exposes a zone (for example Casablanca-Rabat) to the campaign while keeping a comparable region (Marrakech-Agadir) as an unexposed control, then compares sales over 2 to 4 weeks: the gap measures the real lift, independent of any pixel. The holdout deliberately excludes 5 to 10% of the audience and compares the conversion rate of the two groups; Meta offers conversion lift tests that automate this. The rule: before doubling or cutting a significant channel, validate with one of these tests rather than a platform ROAS.

Does server-side tracking solve the attribution problem?

Only partly, and it's important to be precise about what it does. Server-side tracking (GTM Server-Side + Meta CAPI + GA4 Measurement Protocol) recovers the 25 to 45% of conversions lost to ad blockers and restrictive browsers like Safari. It therefore improves the reliability of the signal sent to each platform, which makes every attribution model more accurate within its own silo. However, it does not solve cross-platform double counting: even with a perfect signal, Meta and Google will each keep claiming the same sale according to their respective windows. To correct that double counting, only blended ROAS and incrementality tests work. Server-side is the foundation that feeds reliable attribution, but it replaces neither MER nor geo-tests.

How do I decide with data without drowning in reports?

By ranking metrics instead of multiplying them. Three levels are enough. Level 1: MER (blended ROAS), tracked weekly, is the arbiter — if it holds or rises, the overall advertising machine is profitable, and it is the only number to present to a leader. Level 2: multi-touch attribution (last-click vs data-driven in GA4) serves only for trade-offs between channels, not for global steering. Level 3: incrementality (geo-test or holdout) is reserved for big decisions — doubling a budget, launching or cutting a channel. All of it rests on clean server-side tracking reconciled with the back-office as the single source of truth. This discipline — one role per metric, one source of truth — avoids paralysis by an excess of contradictory reports.

Back to blog

/ 56 · attribution-multi-touch-marocAnalyticsPublished on10 min

ATTRIBUTION · ANALYTICS · MULTI-TOUCH · MOROCCO 2026

Marketing Attribution in Morocco 2026: last-click, data-driven or position-based— how to decide with data

In 2026, no single attribution model tells the truth on its own. Last-click over-credits the final channel and erases the entire top of the funnel; GA4's data-driven model smooths this better but stays locked inside a single silo; position-based splits credit by arbitrary rules. In Morocco, where 25 to 45% of conversions are poorly tracked because of ad blockers and restrictive browsers, relying on a single platform number leads to cutting the wrong campaigns. The real decision is made by cross-referencing multi-touch attribution, blended ROAS (MER), and incrementality tests — geo or holdout — that measure what advertising genuinely delivers.

By Marouane MohachtouFounder & CEO

25–45%mis-attributed conversionsad blockers + restrictive browsers

130–160%double countingsum of platform ROAS vs real sales

MERblended ROAStotal revenue ÷ total ad spend

2–4 wksgeo-test / holdoutto measure incrementality

Marketing attribution: what the word really covers

Marketing attribution is the rule that decides which channel gets credit for a conversion when several touchpoints contributed to the journey. A typical Moroccan customer sees an Instagram story, clicks a Google ad three days later while searching for your brand, returns a week after via a newsletter, then buys. Four touchpoints, one sale. Who gets the credit? The answer depends entirely on the model chosen — and that choice radically changes budget decisions. The problem is not theoretical. Without an explicit model, each platform applies its own by default and claims maximum credit. Meta counts the sale, Google Ads counts it too, GA4 files it under a different channel, and your e-commerce back-office sees only one. You end up with four contradictory numbers for a single reality. In Morocco, the complexity is compounded by two factors. First, a significant share of traffic runs through browsers and devices that block client-side tracking: Safari on iPhone, ad blockers, WhatsApp as an untrackable sales channel. Second, a large share of conversions is finalized offline — a phone call, a WhatsApp message, cash on delivery. These conversions often have no digital touchpoint recorded at the moment of purchase. Understanding attribution, then, is not about choosing the "right" model — it is about understanding the biases of each model so you never make a blind decision.

Attribution = the rule that splits credit for a conversion across multiple touchpoints
Each platform applies its own default model and over-credits its own campaigns
In Morocco: Safari/ad-blocker traffic + offline conversions (WhatsApp, cash on delivery) distort tracking
The goal is not the perfect model but knowing the biases so you decide without being fooled

Last-click vs data-driven vs position-based: the biases of each model

Three model families dominate in 2026, and each tells a different story from the same data. Last-click. The historical default: 100% of credit goes to the last channel clicked before purchase. Simple, but deeply misleading. It systematically over-credits bottom-of-funnel channels — Meta retargeting, brand search on Google Ads, email — by 20 to 40%, and completely erases the role of discovery campaigns (Meta prospecting, TikTok, YouTube) that initiated the journey. Deciding with last-click means cutting the campaigns that fill the funnel because they don't "convert" on the last click. Data-driven. The default in GA4 and Google Ads. An algorithm analyzes converting and non-converting paths to distribute credit according to each touchpoint's real statistical contribution. It is the most technically honest, but it has one fatal limit: it only sees data from its own silo. GA4's data-driven model ignores whatever Meta and TikTok don't share with it, and since third-party cookies disappeared, those blind spots have widened. Position-based (U-shaped). 40% of credit to the first touch, 40% to the last, 20% spread across the middle. Useful for recognizing both discovery and conversion, but the weightings remain arbitrary — why 40/20/40 and not 30/40/30? No data justifies those numbers; it's a convention. The practical takeaway: never steer a budget on a single model. Compare at least last-click and data-driven side by side. If a channel looks profitable in data-driven but weak in last-click, it is a discovery channel you must not cut.

Last-click: over-credits bottom-of-funnel (retargeting, brand) by 20–40%, erases prospecting
Data-driven (GA4): the most statistically honest but blind outside its own silo
Position-based (40/20/40): recognizes discovery + conversion, but arbitrary weightings
Rule: compare last-click and data-driven side by side, never steer on a single number

The cross-channel problem: Meta vs GA4 vs platforms don't count the same

The costliest attribution trap in 2026 is not the choice of model — it is believing you can add up numbers from several platforms. You cannot. Each tool lives in its own world and counts the same sale its own way. Meta counts view-through and click-through on a default window of 7 days after click and 1 day after view. Google Ads counts click-through over 30 days and applies its own data-driven model. GA4 uses a cross-channel data-driven model but on a different window and with its own session definition. TikTok counts differently again. The concrete result: a single real sale can be claimed simultaneously by Meta, Google, and TikTok. When you add up the ROAS displayed in each platform, you often reach 130 to 160% of the real sales recorded in your back-office. This double, triple, sometimes quadruple counting is not a configuration error — it is structural. Every platform has an incentive to claim maximum credit to justify its budget. None has visibility over the others. Two operational consequences. First, never build a dashboard that sums conversions across platforms: the total is fictional. Second, always reconcile platform numbers against a single source of truth — your back-office or your CRM — which counts each sale only once. Server-side tracking (GTM Server-Side + Meta CAPI) improves the reliability of the signal sent to each platform, but it does not solve cross-platform double counting: only blended ROAS does that.

Different attribution windows: Meta 7d click/1d view, Google 30d click, GA4 different again
The same sale claimed by Meta + Google + TikTok simultaneously
Sum of platform ROAS = 130–160% of real back-office sales
Server-side makes per-platform signal reliable, but does not fix double counting across platforms

Blended ROAS and MER: the number platforms cannot inflate

Against cross-channel double counting, one metric resists manipulation: blended ROAS, also called MER (Marketing Efficiency Ratio). Its definition is brutally simple: total revenue ÷ total ad spend over the same period. Nothing to do with attribution — you ask no platform who converted. You look at what enters the till against what goes out on advertising. A concrete example. A Moroccan e-commerce store does 400,000 MAD in revenue over a month and spends 100,000 MAD on advertising (Meta + Google + TikTok combined). The MER is 4.0. It doesn't matter that Meta shows a ROAS of 6, Google of 4, and TikTok of 3 — the sum of those numbers is a mirage. The MER of 4.0 is anchored in accounting reality. It is the number to track week after week to know whether your overall advertising machine is profitable. MER has two strengths. It is unmanipulable: no platform can inflate it because it comes from your books, not their pixel. And it captures the halo effect — when an untrackable TikTok campaign drives sales that surface as "direct" or "organic," last-click misses them but MER sees them. Its limit: MER is global, it doesn't tell you which channel to pull. So you use it as a high-level guardrail (is total spend profitable?), while per-platform multi-touch attribution guides the fine trade-offs between channels. The right Moroccan reflex in 2026: steer the overall budget by MER, and internal allocations by attribution — never the reverse.

MER = total revenue ÷ total ad spend over one period — independent of any attribution
Unmanipulable: it comes from accounting, not a platform's pixel
Captures the halo effect last-click misses ("direct"/"organic" sales driven by ads)
Global guardrail: steer total budget by MER, channel-by-channel trade-offs by attribution

Incrementality: geo-tests and holdouts to measure true lift

Attribution answers "who got the credit?". Incrementality answers the only question that truly matters: "how many sales did this campaign actually generate that would not have happened without it?". A retargeting campaign often shows a ROAS of 8, but if those buyers were going to buy anyway, its real contribution is close to zero. Only an incrementality test settles it. Two methods accessible in Morocco in 2026. The geo-test. You split the country into zones: Casablanca-Rabat receives the campaign, a comparable region (say Marrakech-Agadir) serves as a control with no exposure. You compare sales between treated and control zones over 2 to 4 weeks. The gap measures the real incremental lift, independent of any pixel. It is the cleanest method to test a new channel or a budget increase. The holdout. A fraction of the audience — 5 to 10% — is deliberately excluded from the campaign. You compare the conversion rate of the exposed group to the held-out group. Meta offers built-in conversion lift tests that automate this principle for large budgets. The decision rule: do not change a budget on the faith of a platform ROAS. Before doubling a channel, run a geo-test for 2 to 4 weeks. A channel showing a ROAS of 6 in last-click but 1.5 in incrementality must not be scaled — you'd be paying for sales that were already coming. Conversely, a prospecting channel with a weak apparent ROAS but strong incremental lift deserves more budget.

Incrementality measures sales that would NOT have happened without the campaign — not just credit
Geo-test: treated zone (Casa-Rabat) vs control zone (Marrakech-Agadir), gap = real lift, over 2–4 wks
Holdout: 5–10% of audience excluded as a control group (Meta Conversion Lift automates it)
A ROAS of 6 in last-click but 1.5 incremental must never be scaled

Deciding with data without drowning: the Webotic attribution stack

The temptation in 2026 is to multiply reports to the point of paralysis. The right approach is the opposite: a clear hierarchy of a few metrics, each with a precise role, to decide fast and well. Level 1 — MER, weekly. This is the arbiter. Total revenue ÷ total ad spend. If MER holds or rises, the overall advertising machine is profitable, no matter what each platform claims. If it drops, you hunt the leak before any other trade-off. It is the only number to present to a leader. Level 2 — multi-touch attribution, for channel trade-offs. You compare last-click and data-driven side by side in GA4. The gap between the two reveals discovery channels (strong in data-driven, weak in last-click) you must not cut. This is where you decide to move 10,000 MAD from retargeting to prospecting, or the reverse. Level 3 — incrementality, for big decisions. Before doubling a budget, launching a new channel, or cutting a significant line, you run a geo-test or holdout for 2 to 4 weeks. It's slower, but it's the only proof that holds up. And the technical foundation: server-side tracking (GTM Server-Side + Meta CAPI + GA4 Measurement Protocol) that recovers the 25 to 45% of lost signal and feeds each level with clean data. Reconciled with the back-office as the single source of truth. This stack — three levels of decision, one server-side foundation — lets you steer a serious Moroccan budget without drowning in contradictory reports.

Level 1: weekly MER = the arbiter, the only number to present to a leader
Level 2: multi-touch attribution (last-click vs data-driven) to arbitrate between channels
Level 3: geo-test / holdout before any big budget decision
Foundation: server-side (GTM SS + CAPI) reconciled to the back-office as single source of truth

FAQ

Which attribution model should I choose in Morocco in 2026?: There is no single "best" model, and relying on just one is the main mistake. In practice, systematically compare last-click and GA4's data-driven side by side: their gap reveals the discovery channels last-click underestimates. Use data-driven for trade-offs between channels within one platform, but never steer your overall budget on a platform number. At the global level, track blended ROAS (MER = total revenue ÷ total ad spend), which is anchored in your accounting and cannot be inflated. Finally, for big budget decisions, validate with an incrementality test. The right model is this combination — not a binary choice between last-click and data-driven.
Why do Meta, Google and GA4 show different numbers?: Because each platform uses a different attribution window and sees only its own data. Meta counts by default over 7 days after click and 1 day after view; Google Ads over 30 days after click; GA4 applies a cross-channel data-driven model with its own session; TikTok counts differently again. The same sale can therefore be claimed by several platforms at once. When you add up the ROAS displayed, you often reach 130 to 160% of real back-office sales — structural double, even triple, counting. Every platform has an incentive to claim maximum credit. The only solution is to reconcile against a single source of truth (back-office or CRM) and steer by blended ROAS.
What is blended ROAS (MER) and why track it?: Blended ROAS, or MER (Marketing Efficiency Ratio), is calculated very simply: total revenue ÷ total ad spend over the same period, all channels combined. Its strength is that it depends on no attribution: you ask no platform who converted, you only look at what enters the till against what goes out on advertising. It is therefore unmanipulable — no platform can inflate it because it comes from your accounting. It also captures the halo effect: sales driven by an untrackable campaign (TikTok, awareness) that surface as "direct" and that last-click misses. Its limit: it is global and does not tell you which channel to activate. So you use it as a high-level guardrail, while attribution guides the fine trade-offs between channels.
How do I measure the incrementality of a campaign?: Incrementality measures the sales genuinely generated by a campaign — those that would not have happened without it — not just the attribution credit. Two methods are accessible in Morocco in 2026. The geo-test exposes a zone (for example Casablanca-Rabat) to the campaign while keeping a comparable region (Marrakech-Agadir) as an unexposed control, then compares sales over 2 to 4 weeks: the gap measures the real lift, independent of any pixel. The holdout deliberately excludes 5 to 10% of the audience and compares the conversion rate of the two groups; Meta offers conversion lift tests that automate this. The rule: before doubling or cutting a significant channel, validate with one of these tests rather than a platform ROAS.
Does server-side tracking solve the attribution problem?: Only partly, and it's important to be precise about what it does. Server-side tracking (GTM Server-Side + Meta CAPI + GA4 Measurement Protocol) recovers the 25 to 45% of conversions lost to ad blockers and restrictive browsers like Safari. It therefore improves the reliability of the signal sent to each platform, which makes every attribution model more accurate within its own silo. However, it does not solve cross-platform double counting: even with a perfect signal, Meta and Google will each keep claiming the same sale according to their respective windows. To correct that double counting, only blended ROAS and incrementality tests work. Server-side is the foundation that feeds reliable attribution, but it replaces neither MER nor geo-tests.
How do I decide with data without drowning in reports?: By ranking metrics instead of multiplying them. Three levels are enough. Level 1: MER (blended ROAS), tracked weekly, is the arbiter — if it holds or rises, the overall advertising machine is profitable, and it is the only number to present to a leader. Level 2: multi-touch attribution (last-click vs data-driven in GA4) serves only for trade-offs between channels, not for global steering. Level 3: incrementality (geo-test or holdout) is reserved for big decisions — doubling a budget, launching or cutting a channel. All of it rests on clean server-side tracking reconciled with the back-office as the single source of truth. This discipline — one role per metric, one source of truth — avoids paralysis by an excess of contradictory reports.