The Problem
Automated bidding learns from whatever you mark as a conversion. The signal usually goes wrong in one of these ways:
- Low-value actions such as newsletter signups and button clicks count as primary conversions.
- The same conversion is counted twice, once through GA4 and once through a separate Ads tag.
- Browser restrictions and ad blockers drop conversions before they are recorded.
- Leads that close weeks after the click are never connected back to the campaign.
Each of these teaches the algorithm something false, and it keeps optimizing toward that until the data changes.
How a Conversion Gets Counted
- A visitor clicks an ad
- The landing page receives a click ID (gclid) in the URL
- The Conversion Linker stores the click ID in a first-party cookie
- The visitor converts, now or days later
- The conversion tag sends the event with the stored click ID
- Google Ads matches it to the click, campaign, and keyword
Any of these steps can break. A redirect can strip the click ID, a missing Conversion Linker means nothing gets stored, and denied consent limits what the tag may send. When reported conversions drop suddenly, I walk this chain from the top.
Primary and Secondary Actions
Google Ads lets you mark each conversion action as primary or secondary, and that decides what the bidding algorithm learns from.
| Type | Used for bidding | Typical examples |
|---|---|---|
| Primary | Yes | Purchases, qualified leads, bookings |
| Secondary | No, observation only | Newsletter signups, brochure downloads, add to cart |
The counting setting matters too. For leads, counting “one” conversion per click is usually right, because a person who submits a form three times is still one lead. For purchases, “every” is usually right, because each order has its own value.
Choosing a Tracking Method
Browser-side tags
The standard setup: a tag in the browser sends the conversion directly. It is quick to install and exposed to ad blockers, browser privacy features, and consent restrictions.
Server-side tagging
The browser sends data to a server container you control, and the server forwards it to Google Ads. Match rates improve, more conversions get past ad blockers and ITP/ETP restrictions, and you decide exactly what data leaves your systems. In exchange you pay for hosting and have one more thing to maintain.
Offline and CRM imports
The conversion is recorded later in the CRM and uploaded to Google Ads with the original click ID. It is the only way to connect a deal that closes weeks after the click to the ad that started it.
Most accounts I work on end up with a mix: browser or server-side tags for immediate actions, and offline imports for outcomes that happen later.
What a Good Setup Includes
- Primary and secondary actions are defined so the bidding strategy optimizes toward the actions that bring in revenue.
- Conversions are deduplicated across platforms. A conversion counted through GA4 is not counted again by a separate Ads tag, and purchases carry a transaction ID so repeats are ignored.
- Enhanced Conversions are configured properly. Hashed customer data, such as an email address, recovers conversions lost to browser restrictions and ad blockers while staying within Google’s data policies.
- Server-side tagging is used where the traffic or privacy requirements justify it.
- Offline and CRM conversions are imported, so a lead that closes three weeks after the click is credited to the campaign, keyword, and ad that produced it instead of landing in “direct” or “unassigned.”
- Attribution windows and models match the real sales cycle. The defaults undercount anything with a long consideration period.
Where It Fits
Conversion tracking is the last step in the measurement chain, and the one with money attached.
- A visitor lands on the site
- Consent defaults load, then update on their choice
- The site pushes events into the data layer
- Google Tag Manager fires the tags consent allows
- GA4 records the journey, Google Ads receives the conversions
It can only be as trustworthy as the layers underneath. The conversion events usually share their taxonomy with Google Analytics 4, they fire through Google Tag Manager, and the signals from Consent Mode V2 decide whether they are observed or modeled.
Common Mistakes
- No Conversion Linker: the click ID is never stored, so conversions on later pages cannot be matched.
- GA4 import and Ads tag both primary: the same purchase is counted twice and bidding sees double the value.
- Conversions fired on the button click: failed or invalid form submissions count as leads.
- No transaction ID: refreshing the thank-you page records the same order again.
- Every micro action marked primary: the algorithm learns to find people who click instead of people who buy.
Why It Matters
A campaign that “looks like it is underperforming” is sometimes a tracking problem wearing a performance problem’s clothes. I have seen budget pulled from channels that were working, and poured into channels that only won on paper because the tracking flattered them.
Once the conversion signal is clean, automated bidding finds more of what is working, and finds it faster. That is the whole point of automated bidding, and it depends on correct data going in.
How I Approach It
1. Audit
I review the current conversion setup to find what is firing, what is duplicated, what is missing, and where the reported numbers differ from what the CRM or finance sees.
2. Define the conversion hierarchy
Which actions are primary and which are secondary, how each is counted, and whether it should influence bidding at all.
3. Architect the tracking
Browser-side, server-side, or a mix, chosen from the traffic, the tech stack, and the privacy requirements rather than from what is easiest to install.
4. Wire in Enhanced Conversions and offline imports
Where they improve match rates or attribution enough to be worth the work.
5. Validate against real transactions
I check that what Google Ads reports reconciles with what the business closed.
6. Document and hand off
So the setup survives new campaigns, new landing pages, and new people on the team.
If the bidding algorithm is optimizing toward the wrong thing, more budget only buys more of the wrong thing. If you want a second look at your setup, get in touch.