Media Mix Modelling looks at what you spent on each paid channel and what happened to your outcome over the same window — then tells you which channels really pulled their weight and where to put your next pound. This guide shows you how to run it on your own data.
Before you start
You'll need your analytics connected so the agent can see spend history and an outcome metric (typically 6–12 months for a stable read).
A confirmed model run takes about 2–5 minutes.
How to run a Media Mix Modelling analysis
Step 1: Open the agent and pick your client
From the Agent Library, find Media Mix Modelling and click Run Agent. Select the client whose data you want to model, then click Run Agent again to start the chat.
Step 2: Choose a pathway
The agent offers three ways to work:
Discovery only — a quick readiness check, no model fitted. Best when you want to scope a future run or sense-check the data.
Standard run — the default. Discovery, pre-flight summary, a fast model fit, and recommendations.
Deep run — same flow, more thorough fit. Best for bigger budget decisions.
Pick Standard run and submit.
Step 3: Give it a date range
Tell the agent the window to model — for example "Last 6 months", "Q4 2025", or "1 Jan 2025 to 30 Sep 2025". Relative ranges are resolved automatically into specific dates.
Step 4: Review the pre-flight summary
The agent runs a readiness check on your data and shows you what's usable: which outcomes you can model, which paid channels have consistent spend coverage, and any data caveats worth knowing before you commit. Take a moment to read this — it's honest about what the model can and can't do for your window.
Step 5: Confirm the exact run
The agent walks you through a short wizard to lock in the specifics: which outcome to model, which paid channels to include, and a final go-ahead. This second confirmation matters — saying "ok" to the date range alone is not enough to fit the model.
Step 6: Read the results
About 2–5 minutes later, the agent comes back with:
A channel contribution table and pie chart — how much of your outcome each channel pulled.
A budget shift recommendation — a clear, business-language read on where to reweight for next period.
A recommended next-period budget table and bar chart with absolute figures by channel.
A commercial read framing the headline finding (e.g. "Google is the dependable scale driver; Facebook looks like the efficiency opportunity").
Honest data caveats — what the model measured, what it didn't, and how confident to be in the recommendation.
Channel contribution to your outcome:
Recommended next-period budget by channel:
💡 Tip: The model measures what already happened — it doesn't forecast the future. Treat the budget recommendation as a controlled reallocation hypothesis to test, not a guaranteed outcome.
What's next?
If your client's data is too thin for a multi-channel read, the agent will tell you plainly and suggest a broader window or a different outcome. You can also rerun the analysis with a different date range, swap the outcome (sessions → conversions → revenue), or change the channel scope to test alternative scenarios — just say so in the chat.