How we connected Google Ads, Meta, LinkedIn and GA4 through a single agent, and what became visible once the data stopped living in separate places.
One of the biggest obstacles in analysing marketing campaigns is that the information is scattered.
Every advertising platform reports on its own performance, and in many cases applies attribution models that tend to favour the platform itself. Analytics tools, meanwhile, can offer a quite different reading of those same results. Getting a coherent, global picture of how campaigns are performing becomes genuinely difficult.
In this project the disagreement was not subtle. Over the same 91 days, Google Ads reported 2,287 conversions while GA4 recorded 1,075 key events across the entire site — the platform claimed more than twice the total measurable outcomes of every channel combined.
AI agents let us approach the problem differently. Through one or more MCPs we connect the agent straight to the data sources, reach the information in natural language, and analyse it the same way.
Here we used the Coupler.io MCP, which centralises data from our different marketing sources into a single platform.
With the data connected, we can bring in specialised Skills — sets of instructions the agent reads and applies consistently.
A data visualisation Skill builds a detailed, unified view of how the campaigns are performing, independently of which platform the data came from. It also imposes rules that are easy to skip by hand: among them, that colour is never judged by eye but measured. Applied here, it rejected our first palette as too close to distinguish reliably, and we replaced it.
Spend comes from Google Ads, Meta and LinkedIn. Sessions, key events and revenue come from GA4, measured the same way for every channel — which is what lets them sit on the same rows.
| Channel | Spend | Sessions | Key events | Rate | Revenue | € / session |
|---|---|---|---|---|---|---|
| Google Ads · paid | €15,415 | 13,513 | 614 | 4.54% | €27,772 | €2.06 |
| Meta · paid | €14,596 | 8,367 | 292 | 3.49% | €100 | €0.01 |
| LinkedIn · paid | 8.665 USD | 13,762 | 19 | 0.14% | €0 | €0.00 |
| Direct | — | 10,880 | 269 | 2.47% | €72,570 | €6.67 |
| Organic search | — | 4,903 | 238 | 4.85% | €17,998 | €3.67 |
| Referral | — | 1,825 | 75 | 4.11% | €10,741 | €5.89 |
| Organic social | — | 1,491 | 83 | 5.57% | €1,391 | €0.93 |
From there, a Marketing Analytics Skill takes the analysis considerably deeper: crossing metrics between platforms, detecting patterns, identifying anomalies and surfacing insights that fragmentation previously made very difficult — or simply impossible — to reach.
The three platforms bought a similar number of clicks — between 19,898 and 21,691 — on budgets of the same order. From there they resemble each other in nothing. Following the same funnel on all three, each breaks at a different point, and one of them does not break until the very end.
of LinkedIn's clicks get recorded as a session — the best of the three platforms, ahead of Google Ads at 65.0%.
of those sessions convert. Google Ads sits at 4.54% and Meta at 3.49% — between 25 and 32 times higher.
spent on LinkedIn over the period, for 19 key events and €0 of attributed revenue.
The goal is not only to understand what happened. On the same analysis, the agent proposes concrete actions to optimise the investment, correct inefficiencies and improve campaign performance.
We move from analysing each platform in isolation to an intelligent layer that connects the data, reads it as a whole and turns it into decisions. And the first finding was in none of the platforms — it was in the one that was missing.