For years, we have been promising that artificial intelligence would transform digital marketing. Now I can say, with real data in hand, that this transformation is already happening. And it is happening here, on our platform.
In this article, I share what we have built at Eulerian together with Anthropic: three integrations that, together, radically change the speed and depth with which our clients can turn data into decisions.
The problem we have always had
Our clients have extraordinarily rich data in Eulerian: visitor behavior, conversions, channels, campaigns, publishers, multi-touch attribution, MMM models… The historical problem was not the lack of data. It was the time, resources, and effort required to turn that data into decisions.
A full campaign analysis with year-over-year comparison, channel breakdown, conversion funnel, and actionable recommendations could take days of work between extraction, spreadsheet processing, and report writing. Many times, that analysis arrived too late, or was simply never done.
With the integration of Eulerian’s MCP with Claude, a complete campaign report (including analysis of all channels, publisher breakdown, SEM campaign detail, and 9 prioritized recommendations) is generated in less than 15 minutes, without exporting a single piece of data manually.
That is what changes with the three integrations described below.
1. MCP INTEGRATION: CLAUDE AS A NATIVE EULERIAN ANALYST
The MCP (Model Context Protocol) is the technology that allows Claude to connect directly to Eulerian APIs and operate on them autonomously. It is not an import/export connector. It is a real-time bidirectional connection where Claude can query, interpret, and act on platform data through natural language.
What can Claude do through MCP?
|
Batch reporting queries Claude runs queries against the Eulerian API, selecting dimensions, metrics, and timeframes autonomously based on the business question |
Time comparisons Analyzes multiple periods simultaneously (campaign vs. pre-campaign vs. previous year) in a single conversation, without prior setup |
|
Channel and publisher breakdown Extracts granular performance for each channel, each display publisher, and each SEM campaign with their efficiency metrics |
Interpretation and recommendations Goes beyond numbers: identifies anomalies, explains likely causes, and proposes concrete, prioritized actions |
|
Report generation Produces full PDF reports with charts, tables, and executive narrative, ready to present to a client |
Catalog exploration Automatically discovers available dimensions and metrics for each client, adapting analysis to the specific configuration |
A concrete example: campaign analysis
In practice, the workflow is as follows. The analyst describes business questions in natural language. Claude, via MCP, does the rest autonomously:
- Step 1: Query the segments and metrics available for the client (in a real case, we found 154 metrics available in the rt#insummary endpoint).
- Step 2: Run global performance queries for the compared periods: active campaign, pre-campaign, and the equivalent campaign from the previous year.
- Step 3: Extract breakdowns by channel, by publisher, and by campaign, along with their respective efficiency metrics.
- Step 4: Interpret the results, identify anomalies, and formulate prioritized recommendations with quantified estimated impact.
- Result: A complete report delivered in less than 15 minutes.
The type of insight that generates the most value is not always the largest number. It can be a remarketing campaign with exceptional ROAS that no one had identified because it was buried among thousands of lines of data. Claude finds it, quantifies it, and recommends scaling it. That kind of insight can be worth tens of thousands of euros in the next campaign planning cycle.
2. MMM PLUGIN: AI-DRIVEN INVESTMENT OPTIMIZATION
The second integration is the MMM (Marketing Mix Modeling) plugin connected to Claude through the Robyn model integrated in Eulerian. This integration brings predictive analytics to a level of accessibility that was previously unthinkable.
Traditional MMM required data scientists, weeks of calibration, and technical documentation that few stakeholders could understand. With Claude’s integration, the model speaks the language of business.
What can Claude do with Eulerian’s MMM?
Based on the saturation and adstock parameters of the already calibrated Robyn model in Eulerian, Claude generates investment scenario comparisons and answers strategic questions such as:
- What happens if I cut the budget by 30%? Does ROAS increase or decrease?
- How much should I spend if I could spend as much as I want?
- What is the saturation point of each channel? Where am I wasting money?
- What is the optimal channel mix to maximize profit or maximize efficiency?
- How does the optimal strategy change if I plan over 4 weeks versus a full quarter?
The scenarios that the model can compare
A typical MMM analysis with Claude compares four to five scenarios over the same time horizon, answering concrete business questions:
|
Scenario |
Question it answers |
When to use it |
|
Baseline (historical budget) |
How should I redistribute my current investment across channels? |
Always as a starting point |
|
Cut (-20% to -30%) |
If I have less budget, how do I minimize the impact? |
Budget constraints |
|
Expansion (+20% to +30%) |
Is it worth investing more? How much do I actually gain? |
Additional budget requests |
|
Unconstrained (optimal) |
How much should I ideally spend and how? |
Strategic planning |
|
Quarterly plan |
How do I distribute investment over the quarter? |
Medium-term budgeting |
Channel-level recommendations generated by the model
Beyond global scenarios, MMM allows Claude to generate specific recommendations per channel based on their real saturation parameters. For example:
- Identify which channel has the most investment headroom before saturation (high EC50 = more room to grow).
- Detect which channel is already operating on the saturation plateau, where additional spend leads to direct losses.
- Recommend an “adstock building” strategy in week 1 for channels with low memory.
- Quantify the marginal efficiency of each additional euro spent per channel.
(*) EC50 is the point where an advertising channel has “consumed” half of its response capacity. A high EC50 means the channel can absorb a large amount of budget before becoming saturated (it still has room to grow). A low EC50 means it saturates with little spend, so investing more is essentially wasteful.
Headroom = the additional room for profitable investment before reaching the channel’s saturation ceiling.
The most common finding in the MMM analyses we have conducted is that clients’ historical budgets are usually reasonably well calibrated. What is valuable is not only confirming this, but also knowing exactly in which channels (and to what extent) there is room for improvement, with concrete numbers.
3. THE ARTIFACTS MODULE: AI-POWERED DYNAMIC DASHBOARDING
The third dimension of this integration is perhaps the one that will most surprise our clients in the coming months: the use of Claude artifacts to generate interactive dashboards and real-time analytics applications.
Claude artifacts are full web applications (HTML, React, visualizations with D3 or Chart.js) that Claude can generate in seconds using Eulerian data. They are not screenshots or static reports. They are interactive interfaces that users can explore, filter, and customize directly within the conversation.
Use cases now available
|
Live campaign dashboard Claude generates an interactive dashboard with the KPIs of the active campaign, which can be updated with new data simply by asking Claude to refresh it |
MMM scenario comparator |
|
Saturation curve analysis |
Recommendation tracker |
|
Basket recovery simulator |
Persistent reports |
The difference with a traditional BI tool is profound: here there is no need to configure a single widget. Claude interprets the data, selects the most appropriate visualization, and builds it. If the client wants something different, they simply request it in natural language and Claude adjusts it instantly.
WHAT DOES THIS MEAN FOR EACH PROFILE?
|
Role |
What changes |
|
Marketing teams |
Less time in Excel, more time on strategy. Business questions are answered in minutes, not days. |
|
Marketing directors |
Ready-to-present executive reports, with the right level of detail, automatically generated for each relevant meeting. |
|
Analytics teams |
Claude handles data extraction and formatting. They can focus on questions that truly require human judgment. |
|
Media agencies and partners |
Optimization discussions move from opinions to data. Investment arguments are grounded in model parameters, not intuition. |
|
Executive leadership |
Visibility into the real ROI of each channel and campaign, with historical comparisons and forecasts, in an executive format and without requiring technical expertise. |
THE ROAD AHEAD
We are in the early stages of something that is going to scale significantly. The integrations described here are already working and already delivering real value, but the roadmap is ambitious.
If you are an Eulerian client and want to see these capabilities applied to your real data, now is the time to start the conversation. The analyses described in this article are not lab demos. They are the new standard of what your platform can do.
Note
The analyses referenced in this article were generated by Claude AI (Anthropic) connected to Eulerian Analytics via MCP integration and the MMM plugin. The capabilities described are available to Eulerian clients.
It is important to clarify that this integration does not replace a fully customized MMM developed from scratch. A bespoke MMM project allows for the inclusion of external variables not available within Eulerian (such as business-specific seasonality, pricing, competitor activity, macroeconomic factors, and offline data) and produces a model calibrated specifically to each client’s reality.
The integration with Claude enhances and makes the existing MMM model within the platform more accessible and actionable. However, when business complexity requires it, a bespoke MMM remains the most comprehensive and precise option.
Author : Roberto Macedo
