Fireshift / Study / Clean setups
What 75 clean Google Ads conversion setups do that 811 others don't
TL;DR. On 5 Nov 2026 we read the public tags of 886 sites with Google Ads conversions. 75 (8.5%) were clean: grade A and zero findings. 36 of the 75 (48.0%) send a value with every Ads conversion, against 23 of the other 811 (2.8%). They also count less: a median of 2 Ads conversions, against 4. Enhanced conversions, Consent Mode and a cookie banner were about as common on both sides. We saw the containers, not the Ads accounts, so primary vs secondary is unknown.
What counts as a clean setup?
We call a setup clean when the free scan gives it grade A and finds nothing. On those 75 sites none of this showed up: an Ads conversion on every page or on a plain view of a page that isn't a success page; a conversion on login; GA4 page views as a key event; an Ads conversion on a button or link; the same conversion set up twice; more than 8 live Ads conversions; no value on any Ads conversion; two or more GA4 properties.
Grade A alone is not the same thing. 325 of 886 sites got an A (B 288, C 168, D 81, F 24 for the other grades), and 250 of the 325 still had a finding.
What do the 75 do differently?
| Habit | Clean (n=75) | 95% interval | The rest (n=811) | 95% interval | Intervals apart? |
|---|---|---|---|---|---|
| A value on every Ads conversion | 48.0% 36 of 75 | 37.1 to 59.1% | 2.8% 23 of 811 | 1.9 to 4.2% | yes |
| 1 to 3 Ads conversions | 66.7% 50 of 75 | 55.4 to 76.3% | 42.8% 347 of 811 | 39.4 to 46.2% | yes |
| Exactly 1 Ads conversion | 30.7% 23 of 75 | 21.4 to 41.8% | 18.7% 152 of 811 | 16.2 to 21.6% | no |
| Median number of Ads conversions | 2 | 4 |
Two habits stand out. They price what they count: about half send a value with every Ads conversion, against fewer than 3 in 100 of the rest. They count less: two thirds run 1 to 3 Ads conversions.
Part of this is built into the test. A clean site sends a value on at least one conversion and runs 8 or fewer, or it would have a finding. A value on every conversion is not part of the test. That one is a choice, and 48.0% of the clean sites make it. Counted per Ads conversion instead of per site, the gap holds: 137 of 213 conversions on clean sites carry a value (64.3%), against 750 of 5,531 on the rest (13.6%).
What doesn't separate them?
| Habit | Clean (n=75) | 95% interval | The rest (n=811) | 95% interval | Intervals apart? |
|---|---|---|---|---|---|
| Enhanced conversions on in a tag | 42.7% 32 of 75 | 32.1 to 53.9% | 38.8% 315 of 811 | 35.5 to 42.2% | no |
| Consent Mode signal | 61.3% 46 of 75 | 50 to 71.5% | 69.2% 561 of 811 | 65.9 to 72.3% | no |
| Cookie banner (CMP) found | 72.0% 54 of 75 | 61 to 80.9% | 72.4% 587 of 811 | 69.2 to 75.3% | no |
| Exactly one GA4 property | 60.0% 45 of 75 | 48.7 to 70.3% | 65.1% 528 of 811 | 61.8 to 68.3% | no |
| Server-side tagging signal | 24.0% 18 of 75 | 15.8 to 34.8% | 15.3% 124 of 811 | 13 to 17.9% | no |
Enhanced conversions, Consent Mode, a cookie banner and a single GA4 property are about as common on both sides. They do other jobs: enhanced conversions match conversions to ad clicks, Consent Mode waits for consent. Turning them on does not change what you count. A site can run all of them and still fire its main conversion on a button. Server-side tagging shows on more clean sites, but the intervals touch, so we don't count it as a habit.
Several setup guides we read on 16 Nov recommend enhanced conversions and server-side tagging for a good setup. Both are worth having, but in this data neither is what sets a clean setup apart. What you count and what each count is worth comes first.
What keeps grade-A sites from being clean?
250 grade-A sites had at least one finding. A site can have more than one, so the rows overlap.
| Finding | Grade-A sites | Share of the 250 |
|---|---|---|
| No value on any Ads conversion | 201 | 80.4% |
| Ads conversion on a plain page view | 14 | 5.6% |
| Duplicate conversion | 13 | 5.2% |
| Two or more GA4 properties | 9 | 3.6% |
| Ads conversion on a click | 7 | 2.8% |
| More than 8 Ads conversions | 6 | 2.4% |
201 of the 250 have one finding and only one: no value on any Ads conversion. For them, the step from good to clean is one field per conversion: a rough value that ranks a demo above a whitepaper download. In Google Ads, that value is what value-based bidding learns from.
The checklist to copy
- One Ads conversion per real step. Demo booked, trial started, paid. Most clean sites run 1 to 3.
- Fire on success. A thank-you page matched by its exact path, or the form's success event. Not the button (click conversions), not the demo page view (page-view rules).
- A value on each one. A rough one is fine (values for lead gen).
- One tag per conversion. No second copy in another container or in the page code (duplicate conversions).
- Nothing on login or on every page (login and trial, every page).
- One GA4 property, and page_view not marked as a key event (GA4 key events).
- Then set primary actions in Google Ads (Google Ads Help: primary and secondary actions, our primary vs secondary guide).
How do I check mine?
- Run the free scan on your domain. It runs the same checks as this study.
- Grade A with zero findings puts you with the 75. Anything else is listed with the tag and trigger behind it.
- For the value check by hand: in GTM, open each Google Ads Conversion Tracking tag and look at the Conversion Value field. Empty on any tag means that conversion has no value in the tag.
- Check the account side yourself: Goals, Conversions, Summary. Is each action Primary or Secondary, and does it have a default value? We can't see this.
Or go through the 9-check audit by hand.
What can't we see from outside?
- Primary or secondary. That is set in the Ads account. A clean container can still bid on the wrong action.
- Values set in Google Ads. A default value set in the account doesn't show in the tag. Some of the other 811 may set values there, so the 2.8% is a floor.
- Whether it fires today. This is a static read of public containers, not a test lead. A renamed dataLayer event can break a clean setup without any sign in the tags.
- Pages we didn't read. We read the homepage, /pricing, /signup and /demo. A tag that only loads on a thank-you page doesn't show.
- Cause. These are shares of what we saw. A clean site may also have a better agency or a simpler funnel. The sample is not random.
Method
Static HTTP read of the homepage, /pricing, /signup and /demo (robots.txt respected) and the public googletagmanager.com containers of 1,727 sites on 5 Nov 2026. Base: the 886 with a readable Google Ads conversion. Clean: grade A and zero findings in the free scan's rules. Counts come from the raw rows by script and a separate second-check script, both run again on 16 Nov with the same result. No forms, logins or bot evasion. No site is named. Full rules: method. All-site results: the study.
Would your setup pass with zero findings?
The free scan runs the same checks in seconds. Public data only.
Scan my siteRelated: the full 1,727-site study · B2B SaaS sites · 9-check audit · all guides
FAQ
What does a good Google Ads conversion setup look like?
A few conversions that fire on real success, each with a value. Of the 75 clean setups we read on 5 Nov 2026, the median ran 2 Ads conversions and 48.0% sent a value with every one.
Do enhanced conversions make conversion tracking clean?
No. 42.7% of the clean setups and 38.8% of the rest had enhanced conversions on. They help match conversions to clicks. They don't change what you count.