Google Ads Conversion Tracking Setup Guide

Start with a measurement specification

Conversion tracking is not finished when a tag fires. It is finished when Google Ads receives the right business event, once, with the right value and enough identifiers to reconcile it with the system of record.

Create a table before opening Tag Manager. For every event, define its source, trigger, value, unique ID, count setting, whether it is primary or secondary, and who owns validation. Primary actions appear in the Conversions column and can guide bidding; secondary actions remain observable without being included in that optimisation set by default.

Choose a collection route deliberately

A native Google Ads website conversion and an imported GA4 key event can both be valid. They differ in collection, processing and reporting. Pick one primary purchase or lead route for bidding unless there is a documented reason to count both. Keeping the other as secondary can help compare systems without double-counting the same outcome.

Use Google Tag Manager when it matches the site’s governance and release process; it is not mandatory. Direct Google tag installation, a platform integration or server-side implementation may be appropriate. What matters is controlled deployment, consent handling and a testable data layer.

Ecommerce setup: preserve transaction truth

1. Create a Purchase conversion action and choose the value/currency behaviour.
2. Fire it only after the order is confirmed, not on a button click that can fail.
3. Pass a stable transaction ID so duplicate page loads can be deduplicated.
4. Define whether value includes tax, shipping, discounts and gift cards.
5. Reconcile daily order count and revenue against the commerce platform.
6. Plan adjustments or downstream reporting for cancellations and refunds.

For a hypothetical test order of £96 including £6 shipping, Google Ads showing £96 may be correct if the business optimises to gross collected revenue. If finance evaluates merchandise revenue, £90 may be the intended value. The issue is not which convention is universally right; it is whether every system uses the same documented convention.

Lead generation: send quality back to Google

A thank-you-page event tells Google that a form completed, not that a sale became likely. Capture a lead ID and attribution identifiers with consent, store them in the CRM, and return qualified or won stages to Google Ads. Google’s offline conversion documentation recommends enhanced conversions for leads for advertisers starting this workflow.

As of June 2026, Google is migrating offline and enhanced-conversion uploads toward Data Manager. Google says enhanced conversions can receive first-party user-provided data from website tags and connected data sources under a unified setting. Review the current enhanced-conversions update before implementing an older API tutorial.

Do not upload every CRM status as primary. A practical funnel might observe Raw lead and Qualified lead, while bidding to Sales-qualified opportunity. If the final sale is too delayed or sparse, assign qualified stages values based on historical close rate and contribution, then revisit those weights as the funnel changes.

Enhanced conversions and privacy

Enhanced conversions supplement existing measurement with hashed, user-provided first-party data such as email or phone number. Google’s best-practice guide describes how this can improve matching and diagnostics. It does not remove the advertiser’s obligation to collect and use data lawfully, provide required notices, honour consent choices and meet Google’s customer-data terms.

Hashing is a security treatment, not a legal basis. Coordinate with the person responsible for privacy and consent in the jurisdictions where ads and forms operate.

Use a validation protocol, not a green status badge

1. Open the site in an appropriate consent state and inspect the data layer or event payload.
2. Complete a test using a unique order or lead identifier.
3. Confirm one trigger in Tag Assistant or browser diagnostics.
4. Confirm the record in GA4 DebugView if GA4 is part of the route.
5. After processing, confirm the conversion action, value, currency and identifier in Google Ads diagnostics.
6. Reconcile a sample against the system of record, including duplicates and failed transactions.
7. Repeat on mobile, a second browser, and relevant consent choices.

A tag can report “active” while firing on the wrong event, sending a constant value, or recording twice. Conversely, Google Ads reporting is not instantaneous, so absence immediately after a test does not prove failure. Keep the timestamp and identifier and check after the documented processing period.

Diagnose discrepancies by pattern

• Google Ads is exactly double the order system: inspect duplicate primary actions and repeated purchase triggers.
• Counts broadly match but revenue does not: inspect value variables, currency, tax, shipping and refunds.
• Forms appear but qualified leads do not: tracking may be technically correct while the optimisation event is commercially weak.
• GA4 and Ads disagree modestly: compare attribution, date of event versus ad interaction, identity, consent and processing rules before declaring either broken.
• Performance changed after a site release: compare release time with tag diagnostics, consent rates and conversion-action status.

Protect measurement after launch

Assign an owner and monitor daily or weekly reconciliation according to volume. Alert on sudden drops, spikes, zero values, duplicate transaction IDs and changes to primary-goal configuration. Annotate site releases and consent changes. Test after checkout, CRM, Tag Manager or analytics updates.

When changing the primary optimisation event, avoid changing bidding and campaign structure simultaneously. Google’s conversion-goal change guidance explains that Smart Bidding must adjust to the new input. A clean implementation includes a transition and validation plan.

Count settings, windows and attribution

The count setting should match event meaning. Ecommerce purchases usually count every transaction; a lead action may count one conversion per ad interaction when repeated submissions do not represent additional value. Confirm the exact current Google Ads option in the conversion action rather than copying a template.

Set click-through, engaged-view and view-through windows according to the sales cycle and channel. A window longer than the observed consideration period can increase credited conversions without increasing real sales. A window shorter than the cycle can hide legitimate influence. Use CRM timestamps and GA4 attribution paths to understand delay, then document the choice.

Phone-call measurement

Separate calls from ads, calls to a number on the website and manually dialled calls where the implementation supports them. Define a conversion by business meaning, not just duration. A 60-second support call can be worthless while a 35-second booking call can be valuable.

Where lawful, connect call IDs to CRM dispositions such as new enquiry, existing customer, qualified and sale. Keep raw call actions secondary if bidding uses qualified calls. Test number swapping, mobile click-to-call, forwarding availability, opening hours and consent/disclosure.

Consent changes can look like marketing changes

Record consent-platform releases and acceptance rates alongside conversion trends. A drop in observed conversions after a consent change does not necessarily mean demand fell. It can reflect collection behaviour, modelling eligibility or implementation faults. Validate tags under each relevant consent state and involve privacy counsel where necessary.

Do not attempt to bypass user choices to recover reported volume. Measurement design must follow applicable law and Google policy. The objective is a lawful, explainable signal, not maximum collection at any cost.

A monthly reconciliation routine

1. Export Google Ads conversions by action and conversion date.
2. Export orders or CRM outcomes using the same business time zone.
3. Match transaction or lead IDs where available.
4. Classify unmatched rows: attribution, consent, duplicate, cancelled, test, timing or unknown.
5. Calculate count and value variance by action and device.
6. Assign fixes and retest with unique identifiers.

Expect legitimate differences between systems because attribution and identity differ. The goal is not forced equality; it is explaining the material variance. The Google Ads audit guide provides an evidence format for recording each issue.

Implementation handover

Store the measurement specification, data-layer contract, tag/container version, conversion IDs and labels, CRM mappings, consent dependencies, test cases and owner contacts. A technically correct setup that only one contractor understands is an operational risk. Give the business enough documentation to validate the next site release.

Tracking incident response

If conversions spike or disappear, freeze unrelated changes, note detection time, compare real orders/leads, inspect recent releases and goal changes, and test one unique outcome. Estimate the affected spend and bidding period. Repair the source, annotate the outage and decide whether data exclusions or temporary bidding controls are appropriate.

After recovery, reconcile the full incident window and document root cause, customer impact, detection gap and preventive monitor. Do not backfill or alter data without understanding Google’s current import and adjustment rules.

Measurement change approval

Require review when a primary action, value rule, attribution window, consent configuration or CRM mapping changes. Record old and new definitions, expected reporting effect, deployment owner, test cases and rollback. A silent measurement change can make genuine campaign performance impossible to reconstruct.

Sources

• Google Ads Help: offline conversion imports
• Google Ads Help: enhanced conversions best practices
• Google Ads Help: 2026 enhanced-conversions changes
• Google Ads Help: changing goals used for Smart Bidding