In short
GA4 conversion tracking records valuable actions such as form submissions, calls and purchases as key events. Set it up by defining which actions matter, implementing events through Google Tag Manager or gtag, marking them as key events, testing them in DebugView, respecting consent requirements and linking GA4 to Google Ads.
Every marketing decision depends on knowing what is working. Without accurate conversion tracking, you are guessing which campaigns generate leads, and ad platforms are optimising blind. GA4 is the foundation for most businesses, but it does not track your most valuable actions automatically. You have to set it up.
Step 1: Decide what to track
List the actions that represent real business value. For lead generation, these usually include:
- Contact or quote form submissions
- Booked meetings or demo requests
- Phone calls from the website
- Email link clicks
- Chat conversations started
Separate primary goals, which you optimise toward, from secondary actions, which you monitor, such as brochure downloads or pricing page views.
Step 2: Choose your implementation method
| Method | Best for | Notes |
|---|---|---|
| Google Tag Manager | Most businesses | Flexible, no code changes for each new event |
| gtag.js in site code | Developer-managed sites | Precise, but needs developer time |
| Enhanced measurement | Basic interactions | Automatic, but not reliable for lead tracking alone |
| Platform integrations | Some website builders and CRMs | Convenient, check what is actually sent |
Step 3: Track form submissions reliably
The most reliable approach is to fire an event when a form is genuinely submitted successfully, not when the button is clicked. Common methods include a thank-you page, a success message trigger or a data layer event pushed by your form on success.
Give events clear names, such as generate_lead, and include useful parameters such as form name or service type.
Step 4: Mark key events
In GA4, mark your primary lead events as key events so they appear in conversion reporting and can be imported into Google Ads.
Step 5: Test everything
Use Tag Manager's preview mode and GA4's DebugView to confirm each event fires once, at the right time, with the right parameters. Submit test forms on desktop and mobile.
Step 6: Handle consent properly
Privacy rules in many countries require consent before setting certain cookies, particularly for advertising. Use a consent management platform where needed and implement Google's Consent Mode so tags respect user choices. Requirements vary by country, so check the rules in your markets.
Step 7: Connect to Google Ads
Link GA4 to Google Ads, then either import key events or use the Google Ads conversion tag directly. Turn on enhanced conversions to improve accuracy. If you can, import offline conversions from your CRM so Google learns which leads become customers.
Step 8: Monitor continuously
Tracking breaks more often than people realise: site updates, new forms and plugin changes can all stop events firing. Check key events weekly and investigate sudden drops.
Common mistakes
- Tracking button clicks instead of successful submissions
- Counting the same lead several times
- Leaving test and spam submissions in the data
- Ignoring consent requirements
- Never checking tracking after site changes
Accurate tracking is the first thing we fix in our Paid Media work, because every other optimisation depends on it.
Frequently asked questions
What is the difference between events and key events in GA4?
GA4 records many interactions as events. Key events are the events you mark as important to your business, such as lead form submissions. Google renamed conversions in GA4 to key events, while conversions remains the term used in Google Ads.
Should I import GA4 key events into Google Ads or use Google Ads tags?
Both work. Many advertisers use the Google Ads conversion tag for primary bidding signals because it supports features such as enhanced conversions, while using GA4 for broader analysis.
Why do GA4 and Google Ads show different conversion numbers?
Differences in attribution models, counting methods, time zones, conversion windows and consent handling all cause discrepancies. Some difference is normal.