Stan Consulting · Marketing Atlas · Position · Performance Max

The Shopify PMax Signal Stack.

Five layers of signal that determine how Performance Max performs on a Shopify account. Conversion-goal hygiene, audience signal stack, listing-group structure, asset-group thematic coherence, value rules. The signal stack determines campaign performance more than any individual setting.

Read the Position → Read the Companion Position

01 Section 01 · The claim The claim.

Shopify operators running Performance Max win or lose on the signal stack. The signal stack — the layer of inputs the operator hands the algorithm — determines campaign performance more than any individual setting. Default-configured PMax fails for Shopify operators specifically because the default ignores the data the operator already has.

The claim has two parts. The first part is empirical: in the Shopify accounts the firm has read, the difference between PMax campaigns that produced new revenue and PMax campaigns that recycled existing-customer revenue was not target ROAS, was not budget, and was not creative quality. It was the signal stack. The second part is structural: Shopify operators sit on first-party data Google's default setup does not consume. The data is in the platform. The integration to feed it to Google Ads exists. The signal-stack discipline of using it does not propagate by default.

The position is five layers, configured in order, before PMax is allowed to read traffic.

02 Section 02 · The conventional view What most people believe.

The conventional Shopify-PMax setup is "install the campaign and let Google figure it out." The platform's onboarding flow encourages this. The default Shopify-Google integration pushes a baseline conversion configuration. Google's PMax recommendations push a baseline asset-group configuration. The default works in the sense that the campaign serves and produces conversions. The default does not work in the sense that the conversions it produces are not the conversions the operator's economics actually rewards.

Belief 01

"The platform's defaults are good enough." The argument is that Google has built sensible defaults and that operator-side configuration is the kind of premature optimization that distracts from running the business. Operators who took this argument seriously now sit on accounts where the algorithm has been optimizing against the wrong signal for months.

Belief 02

"PMax handles the signal layer for you." The argument is that PMax will figure out who the right customer is, what the right product is, and what the right bid is, all from the conversion data the operator's site is already producing. This is partly true and entirely insufficient. PMax is a learning system. What it learns is what it is told to learn. Shopify operators have signal Google does not see by default; not handing it over is leaving optimization on the table.

Belief 03

"My Shopify-Google integration is set up; I am done." The argument confuses the integration's existence with its correct configuration. The integration ships with sensible defaults for a generic store. It does not configure a margin-aware listing-group structure, a customer-cohort audience stack, or first-purchase value rules. Those are operator-side decisions the integration cannot make.

Belief 04

"My ROAS is fine; the signal stack does not matter." The argument depends on which ROAS you are looking at. Reported ROAS on a default-configured PMax campaign is almost always higher than purchase-only ROAS reconciled against Shopify revenue. The gap is the signal-stack tax. The operator who has not opened the gap has not seen it; the gap is still there.

Each belief is supported by the platform's framing. None of them are defensible against a clean audit of what the operator actually wants the algorithm to optimize for.

03 Section 03 · Why the conventional view fails Why that belief fails.

The structural argument is that Shopify operators have signal Google's default setup ignores. Post-purchase event data, refund-rate data, AOV-segment data, repeat-rate-by-cohort data, product-affinity data — all of it lives inside Shopify's customer object and order metadata. Google Ads consumes Customer Match audience signals, value rules, and conversion-value adjustments. The Shopify-side data and the Google-side inputs map to each other. The default integration does not perform the mapping. Operator-side configuration does.

Five failure modes follow.

Failure mode one. Add-to-cart counted as primary conversion. The default GTM-Shopify integration pushes add_to_cart as a Google Ads conversion at parity with purchase. PMax optimizes against the combined goal. Smart Bidding learns that cart-abandoner audiences are high-value because they generate cheap "conversions." The bid signal corrupts; the case file The Conversion Goal That Counted Add-to-Carts as Sales documents the pattern.

Failure mode two. Existing-customer Customer Match list attached to prospecting asset groups. The operator uploads a Customer Match list of all purchasers as an audience signal. PMax reads the signal as "find more people like these." The algorithm finds the operator's existing customers searching by brand, and the campaign optimizes toward them instead of toward new customers. Existing-customer revenue gets reattributed to PMax. The case file The PMax Campaign That Ate Its Own Branded Search documents the cannibalisation pattern this produces.

Failure mode three. Single listing group across the entire catalog. PMax's default shopping-surface configuration treats every product equally. High-margin SKUs and clearance SKUs get the same bid. Sale items dilute the bid economics of full-price items. The algorithm optimizes for sell-through rather than for margin, because margin is the signal the operator never handed it.

Failure mode four. One asset group covering every buyer-state. PMax onboarding suggests a single asset group "to let the algorithm learn." On a Shopify account with multiple product lines and multiple buyer-state segments, a single asset group averages across all of them and cannot differentiate which combinations are working. The algorithm chases whatever segment converts cheapest, which is rarely the segment the operator wants to grow.

Failure mode five. No value-rule configuration for first-purchase versus repeat. Without value rules, PMax treats a five-hundred-dollar first-purchase customer and a five-hundred-dollar repeat-purchase customer as identical revenue. The operator's actual unit economics differ across the two; first-purchase customers carry CAC absorption against future LTV, while repeat purchases are pure margin contribution. Without the value-rule layer, the algorithm cannot make the distinction.

The default fails not because the platform is broken. It fails because the platform cannot guess at signal the operator has and has not handed it.

04 Section 04 · The SC position The SC position.

Five layers, configured in order. Conversion-goal hygiene first because every other layer optimizes against it. Audience signal stack second because audience signals are the input to asset-group learning. Listing-group structure third because it is the surface on which Shopping bids happen. Asset-group thematic coherence fourth because it is the unit at which PMax learns. Value rules fifth because they tune what the algorithm has learned against the operator's actual economics.

Each layer is named below with its scope, its requirement, and the test that says it has been configured.

Conversion-goal hygiene

Primary conversions are purchase events with revenue values matched to backend revenue. Add-to-cart, view-item, and view-content are secondary at most. The default GTM-Shopify integration pushes intent events at parity with revenue events; the signal-stack discipline demotes them to secondary.

Primary conversion · purchase event only
Secondary conversions · add-to-cart, view-item, view-content, micro-events
Conversion values · matched to actual order subtotal, audited against Shopify weekly
Attribution model · aligned with operator's revenue-recognition policy

Test it has been configured: reported revenue on Google Ads reconciles against Shopify recognized revenue inside a five-percent tolerance.

Audience signal stack

Segmented first-party customer lists handed to PMax as audience signals: all purchasers, repeat purchasers, high-AOV cohort, recent purchasers (last thirty / sixty / ninety days), churned customers. Each list is a different signal. The signals are attached to asset groups by buyer-state — prospecting groups receive prospecting signals, retention groups receive existing-customer signals.

Segmented Customer Match lists · built from Shopify customer metadata
Custom segments · aligned to the brand's ICP using post-purchase data
Signal-to-asset-group pairing · matched to the buyer-state of each group
Existing-customer signals · excluded from prospecting groups

Test it has been configured: the operator can name, by audience signal, which asset group it is attached to and why.

Listing-group structure

Product groupings inside the Merchant Center feed mirror the brand's margin structure. High-margin SKUs are isolated. Sale and clearance items are isolated. Fast-moving and slow-moving SKUs are separated. The algorithm sees margin and velocity at the group level rather than averaging across the catalog.

High-margin SKUs · their own listing group with its own bid economics
Sale and clearance SKUs · isolated to prevent dilution of full-price bids
Fast versus slow inventory velocity · separated
Product-line grouping · aligned with how the operator already segments the catalog

Test it has been configured: the operator can predict, before any PMax change, how a margin-tier change in Shopify will surface inside Google Ads bidding.

Asset-group thematic coherence

One asset group per buyer-state per product-line, not one asset group per "campaign." Headlines, descriptions, images, and videos in each group share the buyer-state and product-line theme. Audience signals attached to each group match the group's buyer-state. The algorithm learns at the asset-group level; coherence at this layer determines learning quality.

Asset-group taxonomy · one per buyer-state / product-line combination
Creative coherence within each group · thematic, predictable, third-party-readable
Signal-asset pairing · audience signals match the group's creative theme
Listing-group selection · matches the asset group's product-line scope

Test it has been configured: a third-party reader can name each asset group's purpose from its assets and signals alone.

Value rules

First-purchase versus repeat-purchase value adjustments configured against the operator's actual cohort-revenue economics. First-purchase customers weighted higher to absorb CAC against future LTV. Repeat-purchase customers weighted by cohort, not flat across all repeat buyers. Value rules turn the algorithm's bid against the unit economics the business actually rewards.

First-purchase value rule · new-customer conversions weighted against LTV expectation
Repeat-purchase value rule · cohort-weighted, reflecting real margin contribution
Geographic, device, and audience modifiers · layered against revenue-mix data
Monthly reconciliation · value-rule outputs vs backend cohort revenue

Test it has been configured: the operator can show, in writing, the cohort-revenue data each value rule was set against, with the date of last reconciliation.

05 Section 05 · The mechanism The mechanism.

Below is the working spec. Each layer has three numbered moves. The moves are read in order, completed in writing, and signed before the next layer is configured. The whole signal-stack installation completes in under thirty hours of operator time on a typical Shopify Plus account, with most of the time spent at Layer 2 (audience-list construction) and Layer 5 (value-rule reconciliation).

L1 Conversion-goal hygiene Configure first · foundational

Set purchase as the only primary conversion

Open the conversion-goal grouping in Google Ads. Demote add-to-cart, view-item, and view-content to secondary status. Confirm the primary conversion is the revenue-recognition event the operator considers a sale — purchase, completed checkout, or the platform-equivalent event.

Strip cart-value parameter from non-purchase events

Open GTM. Remove revenue-value parameters from the add-to-cart and view-item tags, or set them to zero. Non-purchase events do not carry revenue value into the bidding signal. Verify the change in GTM preview mode against a live test session.

Verify purchase-event accuracy

Run a live test order through GTM preview. Confirm the purchase event fires, that the conversion value matches the order subtotal per the operator's revenue-recognition policy (with shipping and tax handled correctly), and that the event arrives at Google Ads with the right currency and order ID. Reconcile against Shopify weekly.

L2 Audience signal stack Configure second · first-party data

Build segmented first-party customer lists

Export segmented customer lists from Shopify: all purchasers, repeat purchasers (two or more orders), high-AOV cohort (top quartile), recent purchasers segmented by thirty / sixty / ninety days, churned customers (no purchase in last hundred-eighty days). Upload each as a Customer Match list in Google Ads.

Construct ICP-aligned custom segments

Build custom segments from Shopify metadata that match the brand's ICP. Use post-purchase event data (NPS responses, repeat-purchase rates), refund-rate data (low-refund cohorts), and product-affinity data (cross-sell completions). Each custom segment is a hypothesis about who the brand should be acquiring more of.

Attach signals to asset groups by buyer-state

Prospecting asset groups receive prospecting signals: Similar Audiences derived from existing customers, in-market segments matching the product category, custom segments. Existing-customer Customer Match lists are removed from prospecting groups. Retention asset groups receive existing-customer signals.

L3 Listing-group structure Configure third · Shopping surface

Group SKUs by margin tier

Tag SKUs in Shopify (or in the Merchant Center feed via custom labels) by margin tier — high, mid, low. Configure listing groups in PMax to separate the tiers. The algorithm gets a margin-aware target rather than averaging across the catalog.

Isolate sale items

Tag sale and clearance items with a custom label. Configure their own listing group with its own bid economics. Mixing sale items with full-price SKUs distorts the bid signal across both. The fix is structural; tag-and-separate is a one-time setup that compounds correctly.

Segment by inventory velocity

Tag fast-moving and slow-moving SKUs with custom labels reflecting current inventory velocity. Configure separate listing groups. The bid economics differ — fast-movers tolerate higher bids; slow-movers do not — and the algorithm needs the difference exposed at the listing-group level rather than left to learn implicitly.

L4 Asset-group thematic coherence Configure fourth · learning surface

One asset group per buyer-state and product-line

Asset-group taxonomy mirrors how the operator already segments the catalog. Buyer-state on one axis (prospect, recent visitor, existing customer, lapsed). Product-line on the other axis. The default single-group setup is replaced with explicit one-per-segment groups.

Asset coherence within each group

Headlines, descriptions, images, and videos in each group share the buyer-state and product-line theme. A third-party reader, looking at one asset group's creative, can name the buyer-state and product-line it serves. Mixed-coherence groups dilute the algorithm's optimization.

Signal-asset pairing per group

Audience signals attached to each asset group match the group's buyer-state. Misaligned signal-to-creative pairings produce diluted optimization. The audit asks: does the audience signal match the creative theme match the listing-group selection? Three things, one alignment, per group.

L5 Value rules Configure fifth · economics layer

Configure first-purchase value rules

First-purchase customer value rule weights new-customer conversions higher than repeat conversions, by a factor reflecting the operator's CAC-to-LTV expectation. The algorithm bids accordingly, paying more for first-purchase acquisition than for repeat-purchase capture.

Configure repeat-purchase value rules

Repeat-purchase rule reflects the operator's actual LTV economics. Repeat conversions are weighted by cohort, not flat. High-LTV cohorts (defined by AOV, frequency, or product mix) are weighted higher than the operator's average repeat customer. The rule is built from Shopify cohort-revenue data.

Audit value rules against backend revenue monthly

Reconcile value-rule outputs against backend cohort-revenue data monthly. Adjust weights as new LTV data accumulates. Value rules are not a one-time configuration; they are a tuning surface that gets re-fit as the underlying economics evolve.

06 Section 06 · Evidence and case links Evidence and case links.

The Position page is the operator-facing working specification. The links below are where the doctrine has been applied or summarized for an adjacent audience. Each link is a test the doctrine has had to pass.

Primary case

The Conversion Goal That Counted Add-to-Carts as Sales

The composite case file documenting the Layer-1 failure mode. Shopify food and beverage operator, $3.4M annualized, $62K monthly spend. Reported ROAS 3.8x; actual purchase ROAS 1.4x. The signal-stack defect at Layer 1, end to end.

Read the case file →

Companion case

The PMax Campaign That Ate Its Own Branded Search

The composite case file on the Layer-2 failure mode. Shopify Plus apparel brand. Existing-customer audience signal attached to a prospecting asset group. The cannibalisation pattern that audience-signal hygiene prevents.

Read the case file →

Companion position

Performance Max Is Not a Replacement for Search

The firm-doctrine companion to this operator-facing specification. The four conditions for PMax-Search co-existence. The two positions are read together by operators implementing the full installation.

Read the position →

Adjacent position

The Three-Layer Google Ads Diagnostic

The firm's diagnostic methodology for Google Ads accounts. Account integrity, campaign structure, bid-strategy alignment, in that order. Layer 1 of the diagnostic maps to Layer 1 of this signal stack.

Read the position →

07 Section 07 · Where it breaks Where it breaks.

Every working specification has assumptions. The signal stack assumes Shopify Plus or Shopify advanced reporting tier and a brand with sufficient first-party signal volume. When either assumption is false, the specification needs a different first move.

Brands under $1M revenue

Layer 2 (audience signal stack) requires enough first-party signal volume for Customer Match lists to be meaningful. Brands under one million in annualized revenue often do not have lists large enough for Google Ads to use them at scale. The signal-stack discipline still applies; the audience-list layer collapses to "Similar Audiences derived from a small seed list" until volume catches up.

Shopify basic tier without advanced reporting

Layer 5 (value rules) and parts of Layer 3 (margin-aware listing groups) depend on cohort-revenue data and SKU-level margin data the basic Shopify reporting tier does not expose cleanly. Operators on basic-tier Shopify can configure Layers 1, 2, and 4 fully; Layers 3 and 5 collapse to a simpler shape until the operator upgrades.

Subscription-heavy or LTV-driven economics

Brands whose revenue is concentrated in subscription renewals or in long-tail LTV (twelve-plus-month payback periods) need Layer 5 (value rules) to be the heaviest layer in the stack. The default treatment of value rules as a tuning surface becomes inadequate; the rules become the primary optimization mechanism. The signal-stack ordering still holds; the relative weight shifts.

Headless or non-standard Shopify implementations

Operators running headless storefronts (Shopify backend, custom front-end) or non-standard checkout flows have a Layer 1 problem the standard Shopify-Google integration cannot solve. The conversion-tracking layer requires custom GTM configuration, and the signal-stack work begins after the tracking is verified end to end.

08 Section 08 · What it costs to apply What it costs to apply.

The signal stack installs as the Conversion Second Opinion for operators who want the read on its own. It installs as part of an F4 Marketing System Build for operators ready for the full installation across the account. The methodology is the same in either format. The deliverable shape and the engagement length are different.

Diagnostic only

Conversion Second Opinion

$99972-hour verdict

A written diagnostic verdict against the five layers. Read across each layer. Named structural defect. Recommended install order. No restructure, no implementation. The read.

See the engagement →

Diagnostic plus install

F4 Marketing System Build

$5K–$300Kscoped on the read

The diagnostic runs first as the scoping artifact. The Marketing System Build engagement runs the install of all five signal-stack layers and the supporting reporting layer. Pricing is set against the scope of the install, which depends on account complexity, catalog size, and the depth of the audience-list rebuild required.

See the engagement formats →

Five Cents · Stan's note

The reason this position is operator-facing rather than purely doctrinal is that I have stopped believing that doctrine alone changes operator behaviour. Operators read doctrine, agree with it, and continue to run the default configuration because the default is what shipped with the integration and reading the configuration screen is not on anyone's calendar. The signal stack exists as a working specification because the working specification is what gets done.

The thing I keep noticing about Shopify operators specifically is that they are sitting on more first-party data than almost any other category of operator I read for, and they are using less of it inside Google Ads than almost anyone. The Shopify backend has the post-purchase data, the refund-rate data, the cohort-revenue data, the product-affinity data, all of it sitting in a structured database. The Customer Match upload, the value-rule configuration, the listing-group tag, all of it exists on the Google Ads side. The integration to bridge them exists. The discipline of using it does not propagate by default. That is the gap this position closes.

What I keep seeing is the operator who reads this page recognizing their account in Section Three. The recognition is the deliverable. The seventy-two-hour written verdict is what proves the recognition was right and produces the install order to act on. The Marketing System Build is what runs the install when the operator wants the full installation rather than the read.

Stan Tscherenkow · Marketing Atlas · 2026-05-07

10 Section 10 · Related Atlas entries Related Atlas entries.

The Reference pages in the Performance Max cluster, the Reference pages in the Google Ads Waste cluster, the case files this position was written against, the companion position, and the hub. The graph below is the cluster map.

If you read this and recognized your account

Configure the five layers. Then run the campaign.

The Conversion Second Opinion runs this signal stack against your account in seventy-two hours. A written verdict, the structural defect named, the install order set across the five layers in the order they have to be configured. If the verdict says install, the F4 Marketing System Build runs the full installation against your Shopify account. If the verdict says hold, you keep the read and act on it yourself.

Get the Conversion Second Opinion · $999 Or read the firm-doctrine companion →

The Shopify PMax Signal Stack.

Conversion-goal hygiene

Audience signal stack

Listing-group structure

Asset-group thematic coherence

Value rules

Set purchase as the only primary conversion

Strip cart-value parameter from non-purchase events

Verify purchase-event accuracy

Build segmented first-party customer lists

Construct ICP-aligned custom segments

Attach signals to asset groups by buyer-state

Group SKUs by margin tier

Isolate sale items

Segment by inventory velocity

One asset group per buyer-state and product-line

Asset coherence within each group

Signal-asset pairing per group

Configure first-purchase value rules

Configure repeat-purchase value rules

Audit value rules against backend revenue monthly

The Conversion Goal That Counted Add-to-Carts as Sales

The PMax Campaign That Ate Its Own Branded Search

Performance Max Is Not a Replacement for Search

The Three-Layer Google Ads Diagnostic

Brands under $1M revenue

Shopify basic tier without advanced reporting

Subscription-heavy or LTV-driven economics

Headless or non-standard Shopify implementations

Conversion Second Opinion

F4 Marketing System Build

Five Cents

Performance Max

Brand Exclusions

Asset Groups

Conversion Goals

Smart Bidding

Google Ads Wasted Spend

Search Terms Report

Negative Keywords

Quality Score

Negative Keywords

Performance Max Is Not a Replacement for Search

The Conversion Goal That Counted Add-to-Carts as Sales

The PMax Campaign That Ate Its Own Branded Search

Marketing Atlas

Configure the five layers. Then run the campaign.