Google added a no-code Scenario Planner to Meridian in February 2026, removing the Python barrier that kept Marketing Mix Modelling out of reach for most SMB WooCommerce stores. That’s the good news. The bad news: most WooCommerce stores still can’t run a credible Meridian model — because the data Meridian needs, two to three years of clean channel-level weekly revenue, is exactly what fragmented pixel stacks fail to produce.

The Python barrier just fell. The data-plumbing barrier is now the entire bottleneck. Here’s what changed, what Meridian actually needs, and why most WooCommerce stores will discover their pixel stack disqualifies them on day one.

What Changed in February 2026#

Meridian is Google’s open-source Marketing Mix Model framework, generally available since February 2025 and now at version 1.6.1. Until early 2026, running it required Python, TensorFlow Probability, and a working knowledge of Bayesian inference — three skills almost no SMB marketer has.

Scenario Planner changed that part. It’s a no-code Looker Studio interface that sits on top of an existing Meridian model, lets a marketer reallocate budget across channels, and shows projected outcomes — without writing a line of Python (Stella MMM agency guide, 2026). Want to see what happens if you shift 20% of Performance Max spend to YouTube? Drag a slider. See the credible interval. Move on.

For the first time, the model-running layer is approachable for a non-technical marketer. That’s a real shift. As the Stella agency put it: Meridian is “one of the best open-source MMM frameworks available. It is also not plug and play.”

Translation: the modelling is now easy. The inputs are still the hard part.

Why MMM Matters More Than Platform Attribution#

Marketing Mix Modelling measures the incremental contribution of each channel to revenue using aggregated, channel-level data — no user identifiers required. It’s privacy-safe by design, immune to cookie deprecation, and works at the level platform attribution can’t reach: what would have happened if you didn’t run those ads at all.

The adoption numbers explain the urgency. 60% of U.S. advertisers currently use Marketing Mix Models, and 58% of non-users are considering adopting one (Kantar measurement study cited by Google, 2025). And the upside is measurable: C-level leaders who placed high importance on MMM were over 2x more likely to exceed revenue goals by 10% or more (Deloitte, cited by Google, 2024).

But the killer stat is this one: median incremental ROI for Performance Max is 4.64x and for non-brand Search is 5.21x — systematically 2-5x lower than the ROAS platforms self-report (Cassandra Google Ads Benchmarks 2026, analysis of 253 MMMs covering $383M in spend).

That number should keep you up at night. Your Google Ads dashboard tells you Performance Max returned 12x. MMMs of comparable spend say the real incremental number is closer to 4-5x. Google’s new Performance Max channel reporting tells you where your budget went — but not which WooCommerce products it bought, and even that platform-level view still inherits the same self-reporting bias. The gap between platform attribution and incremental ROI isn’t rounding error. It’s the difference between “double down” and “rebalance the budget.”

What Meridian Actually Needs From Your Data#

This is where the rubber meets the road. Per Google’s own documentation, Meridian requires at least two to three years of weekly data and supports fully Bayesian models with 50+ geos via TensorFlow Probability (Google Meridian documentation, 2026).

Specifically, Meridian needs:

• Weekly aggregates. Spend and revenue rolled up to the week, consistently, for at least 104 weeks.
• Channel-level granularity. Google Search, Performance Max, YouTube, Meta, TikTok, organic, email, and direct, each as a separate input. Mixed buckets like “paid social” don’t cut it.
• Consistent definitions. What counted as “Meta ROAS revenue” in March 2024 has to match what counts in March 2026. Currency, attribution window, refund handling — all stable across the entire history.
• Geo splits. Meridian’s geo-level hierarchical model uses regional spend and revenue to pull more information out of the same dataset, often tightening credible intervals on ROI. Without geo splits, you fall back to national-level estimates with looser confidence bands.
• Renewal and subscription revenue handled correctly, not just the first-order pixel ping.

If you read that list and felt your stomach drop, you’re not alone.

Why Most WooCommerce Stores Fail at the Data Step#

73% of GA4 implementations have silent misconfigurations causing 30-40% data loss (SR Analytics, 2025). That’s the same data-quality problem that disqualifies most WooCommerce stores from feeding a credible Meridian model. You can’t run a Bayesian model on a dataset that drops a third of its events and stamps another third with inconsistent attribution.

The specific WooCommerce pain points stack up fast:

• Fragmented pixel stack. Meta pixel fires from one plugin, Google tag from another, GA4 from a third. Each has different consent handling, different deduplication logic, different definitions of “purchase.”
• Missing renewal events. WooCommerce Subscriptions and most recurring billing plugins don’t fire purchase events on renewals by default — so your MMM sees all your revenue as acquisition. Smart Bidding can’t see your subscription renewals, and neither can Meridian if you feed it the same data.
• Currency inconsistencies. Multi-currency stores often store revenue in shop currency, not customer currency — or worse, mix the two across plugins over different time periods.
• Leaky checkout signals. Ad blockers, Safari’s 7-day cookie limit, and consent rejections all subtract from your client-side numbers — but only sometimes, and not consistently across channels.

This is what Stella means when they warn: “Meridian is not a shortcut for messy data. It needs consistent history and enough variation in spend to separate what worked from what just happened to coincide with a good quarter.”

An MMM trained on fragmented pixel data still produces a confident-looking dashboard. It is also wrong, in ways that are hard to detect and easy to act on. The model output looks the same whether your inputs are clean or noisy — and that’s what makes bad data dangerous.

What Clean BigQuery Data Actually Looks Like#

The data Meridian needs has a specific shape: server-side captured, channel-tagged, currency-normalized, deduplicated, weekly-aggregated. That shape doesn’t come out of a pixel stack. It comes out of a server-side event pipeline that writes to a warehouse.

BigQuery is the natural destination — Meridian was built by Google and its documentation ships with BigQuery integration paths. The same architecture that fixes platform-level signal loss produces MMM-ready data: capture events server-side, stamp them with channel and source consistently, hash PII, deduplicate at the event layer, stream to BigQuery.

From there, weekly aggregation is a SQL query. Geo splits are a GROUP BY. Channel-level revenue is a join. The data Meridian needs is one transformation away — but only if the underlying event stream is clean to begin with.

This is the structural reason MMM SaaS tools don’t support WooCommerce: they expect a Shopify-shaped data feed and a managed warehouse. Meridian doesn’t care what e-commerce platform you’re on — it cares about the columns. Build the columns, and Meridian works.

How to Actually Feed Meridian From WooCommerce#

This is the data plumbing problem Seresa exists to solve. Transmute Engine™ is a first-party Node.js server that runs on your subdomain (e.g., data.yourstore.com) and streams clean, server-side, channel-tagged event data straight to BigQuery — the natural feed for Meridian. The inPIPE WordPress plugin captures WooCommerce purchase, renewal, and refund events; Transmute Engine validates them, enhances with server-side data, hashes PII, and routes them simultaneously to BigQuery, GA4, Facebook CAPI, Google Ads, and the rest of your stack.

You’re not buying an MMM from Seresa. You’re building the data layer that makes Meridian — or any MMM — actually credible on a WooCommerce stack.

Key Takeaways#

• Scenario Planner shipped in February 2026. The Python barrier to running Meridian MMM is gone for the budget-modelling layer.
• Meridian still needs 2-3 years of weekly, channel-level, currency-consistent data with geo splits. Most WooCommerce pixel stacks can’t produce this.
• Platform-reported ROAS is systematically 2-5x higher than incremental ROI (Cassandra Benchmarks 2026, analysis of 253 MMMs). Closing that gap is the MMM upside.
• 73% of GA4 implementations lose 30-40% of data to silent misconfiguration (SR Analytics, 2025) — disqualifying inputs for a credible Meridian model.
• A server-side first-party pipeline writing to BigQuery produces the data shape Meridian needs. The bottleneck has moved from Python to plumbing.

Frequently Asked Questions#

Build the data layer before you buy the model. Start at seresa.io.