Google conversion counts changed after Shopify’s checkout upgrade. First check whether the old purchase tag depended on additional scripts or an Order status script tag, then verify the path sending purchases today. Before changing spend, compare destination receipt with paid orders. A missing signal and a missing order require different investigations.
Before changing budget because cost per acquisition (CPA) or return on ad spend (ROAS) moved, audit the full path: Shopify page state, the active tag owner, event receipt in Google, duplicate handling, consent, and the paid-order record the business actually uses.
What changed on August 26
August 26, 2026 was the deadline for non-Plus stores to upgrade the Thank you and Order status pages. Check the store’s actual configuration: an upgrade notice in Settings > Checkout means it still uses the deprecated pages. Shopify’s upgrade guide explains that check and the replacement process.
For app owners, the Order status script-tag timeline is more specific: affected script tags stopped running on that page for non-Plus stores on August 26. Tracking needs a web pixel replacement; visual customizations use a customer account UI extension. This is a page-specific change, not a statement that all storefront scripts stopped.
Google’s migration guide recommends moving legacy post-checkout Google tags to the Google & YouTube app, removing duplicates, and verifying with Tag Assistant and conversion-action diagnostics.
These are official platform instructions, not evidence that every non-Plus store broke on the deadline. A store that had already migrated, did not use an affected path, or has a verified compatible implementation may have no cutover-related loss at all.
Start with page state, not the ads dashboard
The first question is whether the store was exposed to the change.
In Shopify admin, review Settings > Checkout and the relevant configuration. Record whether the store was already on the new Thank you and Order status pages, whether an upgrade notice or report remains visible, and which customizations the report identified.
Then inventory the old and current Google paths. Look for additional scripts, Order status script tags, direct Google tags, Google Tag Manager containers on post-checkout pages, custom pixels, and the Google & YouTube app. Name the owner and destination of each path.
This separates three very different states:
The store was not using an affected legacy path.
The store migrated and the current path can be verified.
The store depended on a legacy path and current receipt is missing or unknown.
Only the third state supports a cutover-repair hypothesis. It still does not establish the size of any business impact.
Run the Purchase Signal Cutover Card
The cutover is the switch from the old checkout pages and tracking path to their replacements. Use one Purchase Signal Cutover Card for each materially different purchase path, with a person responsible for each check.
1. Confirm the checkout page in use
Record the Shopify plan, checkout configuration, upgrade state, effective page version, and the legacy customization that previously owned the purchase signal. If the old path was not affected, stop attributing the discrepancy to this cutover.
2. Identify the tag that owns the purchase
Record the current Google tag or conversion identifier, the Shopify app or pixel that owns it, the connected Google account, and the consent and domain conditions it requires. Google notes that one account of each product type can be connected directly through the app; agency, multi-account, and unsupported custom-setting cases need explicit ownership rather than an assumed migration.
3. Trace one purchase to Google
Use a representative test with the required consent. Check the purchase event, its transaction identity, value and currency, and the intended Google destination. Then inspect the conversion action’s diagnostics. A fired tag and a conversion credited to an ad are different checks. If the account does not expose order-level receipt, mark it unknown and use the available tag and action diagnostics; do not invent a match.
One successful order clears only that tested path. Repeat for the checkout configurations, domains, devices and consent states the store actually uses.
Check duplicates at the same time. Shopify warns that an old script and replacement pixel running together can send duplicate events. Verify the replacement’s receipt before removing the overlap. Shopify’s pixel replacement guidance describes this handoff.
4. Reconcile the period with paid orders
For a declared window, compare the Google report you use for decisions with Shopify paid orders under one written rule for order status, tests, cancellations, refunds, currency, timezone, and value. Record latency separately from missing events.
The comparison does not need to force both systems to match. Google attribution and the Shopify order ledger answer different questions. The job is to explain the bridge well enough that an operator knows whether to repair capture, wait for maturity, correct scope, or investigate the commercial funnel.
Consent can lower event counts without indicating a defect
Shopify notes that event counts can fall after scripts are replaced with pixels because the new path observes consent that the old code may not have handled correctly. That population change needs to be separated from a delivery defect.
Do not label every decline as signal loss. Segment the audit by consent state and jurisdiction where the implementation permits it, retain the privacy setting that produced the observation, and keep unauthorized tracking out of the repair plan.
A measurement path is not healthy merely because it recovers the largest count. It is healthy when it operates inside the intended consent boundary, deduplicates correctly, reaches the named destination, and supports a truthful business decision.
A lower event count can be the correct result
Consider a synthetic diagnostic example, not a store result. A permitted test sample contains 50 unique paid transactions and 60 raw purchase-event receipts. After a duplicate path is removed, a comparable sample has 50 unique paid transactions and 50 receipts.
Raw receipts fell by 16.7%, while paid transactions stayed at 50. The useful finding is that the ten excess receipts disappeared. These are diagnostic receipts, not Google Ads attributed conversions, and a before-and-after comparison does not estimate incremental sales.
If only 40 of the 50 eligible transactions had verified receipt instead, the ten unresolved paths would need investigation. Where transaction-level evidence is unavailable, keep the result unknown. Do not turn a reporting gap into a fabricated order match.
Let the failed gate choose the action
What you find | Next action |
|---|---|
No affected legacy path | Look beyond the upgrade: delivery, offer, inventory, site behavior or reporting maturity. |
Affected path; receipt missing or unknown | Repair or verify the named path before acting on its platform metrics. |
Duplicate receipts | Remove overlap and restate the affected reporting window. |
Receipt passes; values differ | Reconcile discounts, tax, shipping, currency, refunds and order status. |
Counts fall with changed consent | Explain the eligible population; preserve the required permission settings. |
Representative paths pass; paid orders weaken | Investigate demand and customer economics while monitoring untested paths. |
Keep receipt, bidding, and sales lift separate
A migration can change what Google receives and what automated bidding can learn from. The migration guides explain implementation; they do not compare incremental customers, revenue or margin at equal spend. A successful repair therefore supports a measurement conclusion before it supports a budget conclusion.
That means three statements must remain separate:
A supported tag fired and Google received it.
Google used eligible conversion feedback in campaign optimization or reporting.
The advertising created an incremental commercial outcome.
The first can be checked with implementation evidence. The second needs account-level platform state. The third needs a credible comparison, such as a controlled experiment or another design that can estimate what would have happened without the advertising.
Evidence about a different platform, matching tool or tag route does not establish what this checkout upgrade did for your store.
When to look beyond the upgrade
The upgrade becomes a weaker explanation when the store was already migrated or representative purchase paths show valid receipt, correct duplicate handling and the intended conversion-action state, with no new unexplained gap in the paid-order comparison. Keep the conclusion limited to the populations you checked.
It should also lose priority when a business decline begins outside the migration window or appears in stores and channels that did not rely on the affected path.
Save the card with the checkout state, purchase trace, consent condition, owner and paid-order comparison. Use the failed check to choose the repair. If those checks pass, investigate other explanations for the performance change. Existing cash, inventory and profitability limits still govern spend while the diagnosis runs.
