MMM Is the Only Way to Measure AI Mode’s WooCommerce Impact

What is marketing mix modelling?

Marketing mix modelling is a statistical approach that measures each channel’s contribution to revenue by analysing spend and outcome data over time — it does not depend on individual click paths (Salience, 2026). Where attribution tries to trace a single user’s journey from click to purchase, MMM steps back and asks a different question: when we spent more on channel X, did revenue go up?

The maths behind MMM is regression analysis. Feed in weekly ad spend per channel alongside weekly WooCommerce revenue, control for seasonality and promotions, and the model quantifies each channel’s marginal contribution. No cookies. No device graphs. No consent dependencies. The inputs are aggregate business data you already have — your ad platform spend reports and your WooCommerce order export.

This matters now because the measurement infrastructure that click attribution depends on is disappearing. Safari ITP, iOS ATT, ad blockers, consent rejection, and now AI Mode’s zero-click discovery pattern have collectively hollowed out the click path. MMM sidesteps the entire problem by never relying on individual tracking in the first place.

Marketing mix modelling measures each channel’s contribution to revenue using aggregate spend and outcome data — it does not depend on individual click paths (Salience, 2026).

Why can’t click attribution measure AI Mode?

AI Mode sessions drive consideration that converts days later via branded search or direct visits — there is no click path connecting the AI Mode session to the purchase (Elevarus, 2026). This is the structural break. Attribution models, including Google’s own data-driven attribution, need a chain of touchpoints tied to a single user. AI Mode doesn’t produce that chain.

Here’s what actually happens. A potential customer asks AI Mode a question. AI Mode synthesises an answer from multiple sources, possibly mentioning your product. The customer doesn’t click through — they got their answer. Two days later, they search your brand name, land on your site, and buy. GA4 attributes that conversion to branded search. The AI Mode session that planted the seed? Invisible. No referral. No click. No attribution event.

Google deprecated rules-based attribution models (first-click, linear, time-decay, position-based) in 2023 (Elevarus, 2026). Data-driven attribution remains, but it still requires observable touchpoints. When the touchpoint doesn’t exist — because AI Mode answered the question without generating a click — data-driven attribution has nothing to model. The conversion appears to come from nowhere, or from whatever channel the customer used to return.

AI Mode sessions drive consideration that converts days later via branded search — there is no click path connecting the session to the purchase (Elevarus, 2026).

How much search traffic goes through AI Mode?

47.5% of desktop click-through rate now flows through AI Mode results (Discovered Labs, 2026). That’s not a fringe feature. That’s nearly half of desktop search interaction running through a channel that click attribution cannot measure.

The scale is what makes this urgent. If AI Mode accounted for 2% of search interactions, you could afford to ignore the measurement gap. At 47.5% of desktop CTR, you’re potentially blind to the influence behind half your search-driven revenue. Meanwhile, ChatGPT sends 190x less referral traffic than Google (Bosco, 2026) — illustrating how AI interfaces consume information without generating the clicks that attribution systems depend on.

For WooCommerce stores, this creates a specific problem: your Google Ads campaigns may be driving AI Mode visibility that generates conversions, but those conversions show up attributed to branded search, direct, or organic. You can’t optimise spend on a channel you can’t see.

Related: Google Meridian Is Free: How Small WooCommerce Stores Can Finally Use Marketing Mix Modelling

What data does MMM need from a WooCommerce store?

Weekly or daily spend per channel, WooCommerce revenue, conversion counts, and external factors like seasonality — all at an aggregate level, not per-user (Salience, 2026). The data requirements are simpler than most store owners expect, and most of it is already sitting in the ad platforms and WooCommerce’s order history.

The minimum dataset for a useful MMM model is: weekly ad spend broken out by channel (Google Ads, Meta, email, affiliate), weekly WooCommerce revenue (total and ideally by new vs returning customers), and a calendar of promotions, sales events, and seasonal peaks. External variables — weather, competitor activity, market events — improve accuracy but aren’t required to start.

The critical requirement is history. MMM needs enough variation in spend to detect the signal. If you’ve spent the same amount on Google Ads every week for a year, the model can’t isolate Google’s contribution from the baseline. Twelve months of data with natural spend fluctuations is the practical minimum. Stores that have paused and restarted channels actually have better data for MMM — those on/off patterns create the variation the model needs.

Related: How to Run a WooCommerce Marketing Attribution Audit (And Find Out Which Channels Actually Work)

Is MMM only for large enterprises?

Open-source tools like Google’s Meridian and Meta’s Robyn have made MMM accessible to mid-size WooCommerce stores — the main requirement is 12+ months of consistent spend data (Elevarus, 2026). The enterprise-only era of MMM ended when Google and Meta released their models as open-source projects.

Meridian runs in Python and accepts CSV inputs. Robyn runs in R with a similar interface. Neither requires a data science team — a store owner with basic spreadsheet skills can format the input data, and both tools include documentation aimed at non-statisticians. The models take hours to run, not weeks. The output is a set of channel contribution curves showing the diminishing returns on each pound or dollar of spend.

The catch isn’t the tool — it’s the data. Stores spending less than £2,000/month across fewer than three channels may not generate enough variation for the model to detect meaningful signals. But any WooCommerce store running Google Ads, Meta Ads, and email with a combined monthly spend above that threshold has enough data to produce actionable results.

Requirement	Minimum	Ideal
Spend history	12 months	24+ months
Channels tracked	3	5+
Monthly ad spend	£2,000+	£10,000+
Data granularity	Weekly	Daily
Technical skill	Spreadsheet + CLI basics	Python/R familiarity

What is the difference between attribution and contribution?

Attribution argues over which channel gets credit for a fixed pool of conversions; contribution measures the incremental revenue each channel adds — the second question is more useful (Salience, 2026). This distinction is why MMM works where attribution fails for AI Mode.

Attribution is a zero-sum game. If a customer touched Google Ads, then organic search, then converted, attribution divides one conversion across those two channels. The total never changes — you’re just shuffling credit. Contribution asks a different question: if we turned off Google Ads entirely, how much revenue would we lose? The answer might be more or less than attribution suggests, because channels interact. Google Ads might be driving branded awareness that makes organic search more effective.

For AI Mode specifically, attribution literally cannot assign credit because there’s no touchpoint to assign. Contribution can still measure AI Mode’s impact by correlating changes in AI Mode visibility with changes in total revenue, controlling for other factors. If your AI Mode citations increase and revenue rises beyond what your ad spend changes explain, MMM attributes the residual to AI-influenced demand.

Related: Meta Just Split Click-Through and Engage-Through Attribution — and the 1-Day Engage Window Is Eating Your WooCommerce Remarketing Conversions

Does branded search volume measure AI Mode influence?

A rising branded search trend alongside flat generic organic traffic is the clearest proxy for AI-influenced demand, and it is trackable in Search Console and GA4 (Elevarus, 2026). This is the measurement shortcut while you’re building toward full MMM.

The logic is straightforward. If AI Mode mentions your brand in answers, some fraction of those users will later search your brand name directly. If your branded search volume is climbing but your generic organic traffic is flat or declining, something outside traditional SEO is driving awareness. AI Mode is the prime suspect — especially if the trend accelerated after Google I/O 2026.

Set up a Search Console segment for branded queries (your brand name and close variations). Track weekly branded impressions and clicks alongside total organic traffic. In GA4, create an exploration comparing branded organic sessions to total sessions over time. A divergence — branded rising while total organic stays flat — is your signal that AI-influenced demand exists and is growing.

This metric feeds directly into MMM as an input variable. When branded search volume is included as a channel-adjacent signal, the model can separate AI-influenced demand from paid and organic channels, giving you a cleaner read on each channel’s true contribution.

Related: Google Meridian’s No-Code Scenario Planner Just Killed the Python Barrier for WooCommerce MMM

Can WooCommerce stores run incrementality tests for AI Mode?

Geographic holdout tests — pausing AI-visible content in one region and comparing revenue lift — are the gold standard for measuring AI Mode’s causal contribution (Salience, 2026). MMM tells you correlation; incrementality tests tell you causation. Together they form a complete measurement framework.

The test design is conceptually simple. Pick two comparable geographic regions — say, two US states with similar demographics and similar historical revenue. In one region, noindex the pages that AI Mode cites (or block the AI crawler). In the other, leave everything as-is. Run both for 4–8 weeks. Compare revenue, branded search volume, and direct traffic between the two regions. The difference is the causal contribution of AI Mode visibility.

For WooCommerce stores, Transmute Engine’s server-side event pipeline feeds clean, channel-level revenue data into BigQuery — exactly the input MMM needs. Server-side events capture the purchase regardless of whether the customer’s browser blocked cookies, rejected consent, or arrived via a zero-click AI Mode session. The data is complete, timestamped, and already structured for the weekly aggregation MMM requires.

The practical barrier is patience. Incrementality tests need weeks of data. MMM needs months. But the combination produces something click attribution never could: a defensible answer to the question “what is AI Mode actually contributing to our revenue?” — backed by both statistical modelling and causal evidence.

Key Takeaways

• Click attribution structurally fails for AI Mode — there is no click path connecting an AI Mode session to the conversion that happens days later.
• 47.5% of desktop CTR flows through AI Mode — this is too large a channel to leave unmeasured.
• MMM measures contribution, not attribution — it uses aggregate spend and revenue data without relying on individual user tracking.
• Open-source MMM tools exist — Google Meridian and Meta Robyn make modelling accessible to mid-size WooCommerce stores with 12+ months of spend data.
• Branded search volume is the interim proxy — track it in Search Console while building toward full MMM.
• Incrementality tests prove causation — geographic holdout tests measure AI Mode’s causal revenue contribution.

References

1. Elevarus. (2026). Last-click attribution and AI Mode in 2026.
2. Discovered Labs. (2026). How Google AI Mode Ads Work Today.
3. Bosco. (2026). ChatGPT vs Google referral traffic analysis.
4. Salience. (2026). Google I/O 2026: The SEO measurement reset.