An upgrade to Google's open-source marketing mix model adds AI agents that spot data errors before they skew your budget decisions.
Google has upgraded Meridian, its open-source marketing mix model, with agentic AI and new upper-funnel measurement capabilities, according to Marketing Dive (Sept. 14, 2025). The focus of the upgrade: help marketers audit data quality and resolve errors that would otherwise poison their attribution math.
A marketing mix model is only as good as the data it reads. If your channels are misreporting spend or conversion counts are duplicated or incomplete, the model will tell you to move budget from a real winner to a real loser. You end up chasing ghosts.
Meridian's new agentic layer works as a data watchdog. Instead of you manually spot-checking thousands of rows, the agent flags anomalies and errors before they enter the model. That means fewer hours auditing and more confidence in the allocations you make.
The upgrade also adds upper-funnel measurement, so you can now model awareness and consideration campaigns alongside your conversion-focused spend. That matters because a lot of your revenue actually came from brand work you did months ago, not just last-click paid search.
Since Meridian is open-source, you own your data and your model. The upgrade just makes both safer to rely on.
How WebKing runs this
We've integrated this version of Meridian into our attribution workflows for brands running multi-channel campaigns. The agentic layer now flags suspect data points that we used to hunt down by hand, cutting the time we spend validating before we can trust a reallocation recommendation.
It means the tool can now spot and flag data quality problems on its own, instead of waiting for you to manually audit thousands of rows. The agent acts like a detective, catching inconsistencies that would otherwise poison your attribution math.
Your mix model tells you how much revenue each marketing channel actually drove. If your data is wrong, your model tells you to move budget from a winner to a loser, costing you real revenue.
Meridian is open-source, so there's no licensing fee. You host it yourself or work with a provider who does, so costs depend on your data volume and how much infrastructure you need.
Only if your data had hidden errors before. The upgrade catches problems faster so your model gives you cleaner signals about which channels earn their spend.
Sources
The Lab is original analysis by WebKing. We summarize and interpret developments from the sources above for industrial, commercial, and small business owners. Figures are reported as published by their sources.
More from the desk
Photo: Iyan RyanA/B testing on Meta Ads works when you isolate a single change and wait for statistical proof. Here's the workflow that moves the needle on conversion rate and return on ad spend.
Photo: Anna DemianenkoWhen Google answers questions directly on the results page, your bid strategy can't stay the same. Industrial and B2B owners need to rethink which keywords are worth fighting for.
Photo: Nghia NguyenWordStream's 2026 benchmark shows Facebook advertisers are paying less per click while engagement climbs. Here's what's driving the shift and how it affects your budget.
Free audit
We read your search, ads and AI visibility, then tell you which of these moves actually applies to your business. Free, 30 minutes, no pitch deck.
