mifold · Shopify measurement remediation

Repairing Shopify measurement before trusting the next experiment.

Blue mifold Comfort portable booster shown folded in a hand, with its open seat configuration above.

Summary

Investigated duplicate GA4 collection and inconsistent conversion definitions, corrected live reporting configuration, and developed a documented measurement architecture across storefront and checkout. The work established clearer event ownership, product-data rules, and validation gates for the next CRO phase.

Services

Measurement audit & reconciliation
GA4, GTM & Google Ads remediation
Shopify event architecture
Data-quality QA & technical handoff

Industry: Portable child booster seats
Platform: Shopify ecommerce
Evidence period: August–September 2026

Results

Pages/session gap versus Shopify: 3.5%
Non-purchase Ads actions corrected: 6
Custom dimensions configured: 16
Purchase click-attribution window: 30 days

Diagnosis

Original page-view collection: 2×
Third-party hosts inventoried: 58
Visibility gaps mapped: Checkout
Purchase discrepancy diagnosed: Meta

Measurement

Per event and surface: One owner
Defined purchase-counting key: Transaction ID
Product, value & currency consistency: Item rules
Reconciliation before readout: Release gates

01 Client & Context

mifold sells portable child booster seats through Shopify. Its products require shoppers to understand the format, compare options, and move through a conventional ecommerce purchase journey. Reliable measurement matters both for diagnosing that journey and for evaluating changes to it.

The engagement began with a tracking and analytics review before the next CRO phase. Shopify commerce records, GA4, GTM, and advertising-platform reporting needed to be read together. A number in one dashboard could not be treated as a trustworthy explanation without checking its definition and collection path.

02 Challenge & Goals

Challenges

A duplicate GA4 page-view source inflated page-level reporting and distorted the engagement baseline.
Missing ecommerce and checkout detail limited visibility between product interest, cart activity, shipping, payment, and purchase.
Non-purchase Google Ads actions carried monetary values, while Purchase used a 90-day click-attribution window. The reporting rules needed explicit correction and a documented change boundary.
A wide third-party tag footprint, inconsistent event definitions, and platform purchase discrepancies made it difficult to separate collection defects from changes in business performance.

Goals

Repair the verified duplicate page-view source and compare the resulting reporting with Shopify.
Separate purchase revenue from supporting conversion actions and document attribution changes.
Define one producer per event and surface, with consistent product, variant, value, and currency rules.
Create a documented implementation and QA process that distinguishes live fixes, built components, and remaining validation.

03 Discovery

We traced the page-view discrepancy to duplicate collection and compared GA4 engagement with Shopify’s direct reporting. This gave the first correction a concrete verification method rather than treating a lower bounce rate as evidence of better behavior.
The review extended into Google Ads conversion settings, the Shopify-to-GA4 event path, checkout visibility, product parameters, and a 58-host third-party inventory. It also identified a Meta purchase-count discrepancy that required reconciliation before using platform attribution as a decision rule.
A separate checkout-decline investigation compared traffic, conversion patterns, and advertising change history. The delivered analysis identified traffic loss associated with campaign changes as part of the explanation; it did not justify treating every decline as a broken checkout.
The rebuild was organized around an event contract: name the event, assign its source and surface, define the payload, and state how it will be validated. Existing integrations were evaluated alongside the proposed changes so established functionality would not be mistaken for newly delivered work.

04 Solutions

Remove duplicate page views and establish a readable baseline

Before: The August review found each page view being collected twice. GA4 reported 2.77 pages per session over August 11–17.
Correction: The duplicate source was paused on August 18. The change was recorded in the implementation status report and checked against the daily reporting series.
After: GA4 reported 1.37 pages per session over August 19–21, compared with the Shopify figure of 1.42. The report described this as alignment within 3.5%.
Interpretation: The accompanying increase in reported bounce rate reflects a changed measurement baseline. It is not evidence that visitors suddenly became less engaged.

Page-level reporting moved much closer to the Shopify comparison after the duplicate collection path was removed.

Historical comparison table: GA4 pages per session changed from 2.77 to 1.37, against Shopify 1.42; reporting windows are shown. — Original August 22, 2026 status-report table. The after period covers three full days, August 19–21; this is a measurement check, not a conversion-lift result.

The short after window was explicitly identified in the source report. The 3.5% figure describes pages-per-session alignment, not purchase accuracy across every event or a business-performance gain.

Separate revenue from supporting conversion signals

Conversion values: Six non-purchase Google Ads actions were changed to count-only: View Item, Page View, Add To Cart, Begin Checkout, Search, and Add Payment Info. Purchase retained its revenue value.
Attribution: The Purchase click-attribution window changed from 90 to 30 days on August 19. The team documented the resulting break in reported attribution.
Reporting foundation: Sixteen custom dimensions and one custom metric were registered in GA4. Conversion-linker and URL-passthrough configuration were also recorded as live.
Readout discipline: The handoff explained that historical values remain as recorded. Figures across the August 19 change cannot be treated as if the measurement rules were identical.

The live configuration corrections made revenue-bearing and supporting actions easier to distinguish, with the reporting transition documented for the team.

The six non-purchase actions were secondary, so stripping their values did not change the signals used for bidding. Changing the primary Purchase attribution window was different: the source anticipated changed reported ROAS and a bidding relearning period.

Define the storefront-to-checkout measurement contract

Ownership: Assign one producer to each event and surface. Storefront page views belong to GTM; the Shopify custom-pixel path handles checkout events. A DOM click and a successful add-to-cart receive different event names.
Purchase definition: Map checkout completion to purchase with a transaction ID. Define item identity, variant, category, currency, and numeric value so downstream reporting has consistent inputs.
Built components: The August 22 report records theme code on a copy of the live theme, an imported unpublished GTM container, checkout pixel code, and a technical handoff. Access was unblocked on August 24; subsequent project messages refer to the new implementation and a validation period.
Validation: The contract requires checks for duplicate sends, correct payloads, storefront-to-checkout identity continuity, and reconciliation against commerce records before interpreting an experiment.

The team received a concrete event architecture, implementation package, and QA criteria, giving future diagnosis and testing a common measurement specification.

Original event-contract producer table assigning storefront and checkout events to GTM or the Shopify custom-pixel path. — Original August 18 event-contract excerpt showing event ownership. This is the implementation specification, not a screenshot proving every listed event was live.

Original event-contract rules for currency, value, item identity, category, list context, event IDs, and exclusion of personal information. — Original event-contract field rules, cropped to exclude account identifiers. The specification gives each measurement input an explicit definition.

September project records treated tracking as sufficient to move into design and test preparation, while retaining residual regression checks. The available record does not establish that every checkout event, Meta reconciliation item, or regional-consent change was individually verified complete.

05 Implementation & Development

Repair & re-measure: Pause the duplicate page-view source, inspect the change in daily reporting, and compare the post-fix pages-per-session figure with Shopify. Record the dates and short verification window.
Correct reporting settings: Remove monetary values from the six supporting Ads actions, change the Purchase attribution window, configure the GA4 custom definitions, and document which changes affect reported revenue or bidding interpretation.
Build against a contract: Prepare the Shopify theme snippets, GTM rebuild, custom checkout pixel, and deployment instructions around explicit event ownership and payload rules. The August report separates these staged components from already-live fixes.
Carry validation forward: Retain transaction reconciliation, one-send checks, identity continuity, consent behavior, and data-quality limits as release criteria for the next experiment. Later implementation references do not replace an event-by-event acceptance record.

06 Results & Impact

Pages/session after the fix: 1.37 in GA4; 1.42 in Shopify
Non-purchase Ads actions: 6 changed to count-only
GA4 reporting definitions: 16 dimensions + 1 metric
Purchase attribution: 90 days → 30 days

The clearest measured result is the page-view correction: the August 22 report records GA4 pages per session falling from 2.77 before the fix to 1.37 after it, within 3.5% of the Shopify comparison. That made the page-level baseline more useful for subsequent analysis.

The team also delivered live Google Ads and GA4 configuration corrections, a third-party tracking inventory, and a documented measurement architecture with a technical handoff. The value of these changes is clearer definitions and a more disciplined route from diagnosis to testing.

The later project record supports moving beyond the initial access blocker and into implementation and test preparation. It continues to require regression and reconciliation checks, so the study keeps verified corrections distinct from architecture and remaining validation.

Historical evidence: August 18–22, 2026 implementation reporting, with August–September project follow-up. The pages/session after window is August 19–21. The August 19 attribution change creates a reporting break; before/after ROAS is not directly comparable. No conversion-rate, revenue, or ROAS lift is claimed.

Book a free walkthrough

Review your website with Edu in a free 30-minute session. Identify 2–3 conversion priorities and leave knowing what to tackle next.

Book a free walkthrough

The full Shopify Conversion Audit is a separate service, starting at $1,000.