← Back to all postsSeptember 25, 2026

Google Ads multi-source conversions: check what the records mean

Matching records only helps when both sources describe the same event. Check the event, identifiers and values before allowing backend data to change reporting.

Google’s bid-strategy report example, showing strategy status, performance history and conversion reporting.

Before connecting CRM records to a Google Ads conversion action, check that both sources describe the same event. A form submission, a qualified enquiry and a booked job are different outcomes.

Google's multi-source conversions beta can supplement an eligible website conversion action with backend data. It is not a reason to rename an enquiry as revenue or assume every Canadian account has access.

The useful first step is to inspect the event, identifiers and values before anything changes live reporting.

Google’s published bid-strategy report example, including strategy status, conversions and performance history.
Google’s report example shows where strategy status and conversion history sit together. It does not show this business’s results or prove a multi-source connection is configured. Source: Google

Confirm eligibility before designing the connection

Google labels the feature as beta and says it may not be available to every account. Its current documentation limits the supported website actions to those manually implemented through the Google tag or Google Tag Manager; imported Analytics and URL-based actions are excluded.

Check the actual action and account with the person responsible for measurement. If the feature is absent, do not create a substitute live action simply to imitate a guide. You can still improve the underlying records.

Agree what each event means

Events report with Total Revenue metric Google’s published Analytics example, with original labels and example data.
Record what an event means before matching data sources. This is Google’s Analytics example, not the Ads beta diagnostics screen. Source: Google. Open full-size image.

Write a plain-language definition for the conversion you intend to measure. What happened, when did it happen and which record proves it?

If the website records an enquiry and the CRM later records a booking, preserve that distinction. A shared customer or job number does not make two different stages the same event.

Have the measurement owner decide how later outcomes should be represented and which actions belong in bidding goals. That requires a deliberate design, rather than copying a mapping from another business.

Match the event as well as the identifier. Two records about the same customer may still describe different outcomes.

Check identifiers, amounts and currency

Google documents matching within a single conversion action using the transaction identifier. Its deduplication does not remove overlapping events across every separate action in the account.

Review the identifier at each step and confirm that it remains stable. Keep retries from silently creating a different event.

Use an explicitly labelled synthetic example to check the mapping before real records are involved:

FieldHypothetical check
EventBoth records refer to the same confirmed transaction.
IdentifierThe same synthetic identifier appears on both paths.
AmountCAD 10.00 maps to CAD 10.00; a source storing 1,000 cents needs an explicit conversion.
Missing amountMissing remains missing unless an authorised mapping says otherwise.
DuplicateA repeat of the same event does not become another transaction.

These are testing concepts, not an API payload. Follow the actual provider schema and validate the implementation separately.

Be especially careful with values: Google's documentation says backend values can overwrite the original tag value for matched transactions. Zero is a real value. Do not use it as a casual substitute for unknown.

Confirm what information the business is authorised to send, how customer choices are represented and who approves the connection. Hashing a contact detail is not a substitute for that decision.

Use only the information needed for the documented purpose. Keep raw customer records out of public analytics, screenshots and general troubleshooting notes.

This article is preparation guidance. It does not grant permission to connect a sender, import customer data or change an advertising account.

Use the trial to inspect diagnostics

Modify event interface Google’s published Analytics example, with original labels and example data.
Names and matching conditions need an explicit definition. Google’s Analytics example illustrates that principle; it is not a multi-source Ads setup. Source: Google. Open full-size image.

Google recommends timely uploads, ideally within 24 hours, for optimal performance. That is different from saying every later upload is universally rejected.

The documented initial 14-day trial also has specific reporting and bidding behaviour. Inspect diagnostics and the actual action state; do not describe the period as having no information at all.

Agree which evidence the owner will review before the connection's data influences normal decisions. Include failures, retries and a way to prevent repeated submissions after an uncertain response.

Review the business outcome separately

A cleaner conversion record can improve measurement. It does not guarantee a profitable job or prove that a later campaign change caused an improvement.

Keep enquiry quality, confirmed work, costs and capacity visible in the business review. For help examining the measurement plan, learn more about Google Ads management. Tell us what your website records and what the office considers a completed outcome.

Free rank audit