We launched a new Google Ads search campaign for an Australian trades client and gave it a week. The numbers came back looking like a disaster: a few hundred dollars spent, around a hundred clicks, a near-20% click-through rate — and zero conversions. Not low. Zero.
A near-20% CTR means the ads were exactly what people were searching for. They clicked, they landed, and then — according to Google — every single one of them walked away. On automated bidding with a target CPA, that's the kind of first week that gets campaigns killed.
Except the client's CRM told a different story. Real form leads had been arriving all week, and they had come from this campaign. We could prove it: each lead carried a gclid — the click ID Google stamps on every ad click. The leads existed. Google just couldn't see them.
This post is about why that happens, the specific failure mode almost nobody checks for, and the audit you should run before you touch a single keyword.
The Secondary Conversion Trap
Zero reported. Real leads.
When Ads says zero and the CRM disagrees, the plumbing is lying — not the campaign
Conversions reported all week
Against a near-20% CTR and real form leads in the CRM — a contradiction, not a verdict.
Two Sources of Truth, and One of Them Is Lying
When the Ads interface says zero and your inbox says otherwise, you have a contradiction, and resolving it is cheap. Pull the leads out of your CRM and look for gclids. A lead carrying a gclid is a lead that started as a paid click — that's what the parameter is for.
If gclids are new to you: Google appends one to your landing-page URL on every ad click when auto-tagging is on (it's on by default, and it should stay on). Any half-decent form setup or CRM integration captures it alongside the enquiry. That little string is the forensic link between "someone clicked our ad" and "someone became a lead" — independent of whatever the Ads interface chooses to report. Which is exactly why it's the arbiter when the two disagree.
Now line those leads up against the Conversions column in Google Ads. If gclid-carrying leads exist and the column still reads zero, the case is closed: this is a measurement fault, not a demand fault. The campaign is generating business and reporting none of it.
That ten-minute cross-check should always be your first move, because the two failure types demand opposite responses. A campaign that genuinely converts nobody needs new keywords, new copy, maybe a new landing page. A campaign that converts invisibly needs its plumbing fixed — and every change you make to keywords or budget before fixing the plumbing is a change made on corrupted data. We covered what Smart Bidding does when it has no conversion signal in an earlier post: it doesn't pause and wait. It keeps spending, blind.
The instinct to blame the campaign is strong because the campaign is what you can see. Resist it. In our case the campaign was working from day one. The measurement had been broken for weeks before the campaign even launched — which brings us to the trap this post is named after.
One Week, Two Ledgers
What Google Ads reported
Campaign view, first week
Verdict, if you trust it: kill the campaign.
What the CRM held
Same campaign, same week
Real enquiries, each stamped with the click ID that proves the ad sent them.
Two sources of truth. When they disagree, audit the measurement before the campaign.
The Ten-Minute Test
The Secondary Demotion Nobody Notices
The audit found the account-level root cause quickly, and it's the one that earns the word "trap": weeks earlier, both of the account's form-lead conversion actions had been switched from primary to secondary.
If you haven't met this setting: every conversion action in Google Ads is either primary or secondary. Primary actions count into the Conversions column and — critically — feed Smart Bidding. Secondary actions are bookkeeping. They still record into the "All conversions" column, but they contribute nothing to bidding and nothing to the headline numbers.
To be fair to Google, the setting exists for good reasons. Secondary is where you put micro-conversions you want to observe without optimising toward — newsletter signups, scroll events, duplicate actions you're phasing out. Demoting an action is a legitimate, reversible, sometimes-correct move. That legitimacy is precisely why nothing treats it as an emergency: the interface has no way to know that the action you just demoted was the only lead signal the account had.
That split is what makes the demotion so dangerous. Everything still looks installed. The tags fire. Tag diagnostics show green. Conversions even keep accumulating — quietly, in All conversions, the column nobody puts on a dashboard. Real form leads were landing there at a steady clip the whole time. Meanwhile the Conversions column read zero, and Smart Bidding — which only eats primary signals — had been starved of form data account-wide.
Every automated-bidding campaign in that account was flying blind, not because tracking was missing, but because tracking had been filed in the wrong drawer.
Nothing in the interface shouts about this. There's no banner that says "your only form signal is secondary now." A www-redirect can break your tracking loudly enough to notice; a demotion breaks it silently. You find it one way: open the conversion actions table and read the primary/secondary status of every row, one by one.
Where the Signal Forks
Same tag, same firing event — the primary/secondary flag decides whether anything downstream ever sees it.
Filed in the Wrong Drawer
A secondary action still records into All conversions, so the account looks alive. But Smart Bidding only eats primary signals — demote your only form action and every automated campaign goes blind, with zero warnings anywhere.
Dating the Damage Without Change History
The first question after finding a demotion: when did this happen? You need the date to know how much data is corrupted and what decisions were made on it.
You'd expect Change History to answer that. It won't. Conversion-action edits — including primary/secondary flips — don't show up there, and the API's change-tracking surface doesn't expose them either. The single most consequential settings change in the account leaves no entry in the place every practitioner looks for settings changes.
So you date it from the data instead, and the technique is worth keeping. Chart two series, daily: Conversions (primary only) and All conversions. Before the demotion they move together. After it, the primary series goes flat while All conversions keeps ticking along. The day they diverge is your demotion date.
It's the stopped-clock trick: the power cut isn't logged anywhere, but every clock in the house stopped at the same time. In our client's account, the flatline started weeks before the new campaign launched — meaning the campaign never had a working signal for a single day of its life. Its "failure" was inherited.
Dating a Demotion Nobody Logged
The stopped-clock trick: the change isn't logged anywhere, but the day the primary series flatlines while All conversions keeps moving is the day someone flipped the switch.
The single most consequential settings change in the account leaves no entry in the place every practitioner looks for settings changes.
Three More Leaks on the Page Itself
Fixing the demotion would have restored the account's server-side lead tracking. But the audit kept going down to the landing page, and found three more faults — each one invisible from the account view, each one a lesson in how tracking looks installed without working.
The form event had no conversion label. The page fired a gtag conversion event on form submit, but the send_to value carried only an account ID with no conversion label after the slash. Google doesn't error on this. It doesn't warn. It accepts the event and attributes it to nothing — a silent no-op that has the exact appearance of working code.
The phone event pointed at the wrong kind of conversion action. Phone clicks fired an event aimed at a "calls from website" conversion action — the type backed by a Google forwarding number. Those actions count calls Google itself observes through the forwarding number, and they ignore manually sent events by design. You can fire that event for years and it will never record once. The tag wasn't broken; it was pointed at a letterbox with no slot.
The gtag library belonged to someone else. The base tag on the page loaded under a different Google Ads account's AW- ID — almost certainly pasted from another site during the build. Even a perfectly labelled event would have credited a stranger's account with our client's conversions.
Three faults, one common property: every one of them passes a glance test. The scripts load, the events fire, the console stays clean. If your audit stops at "the tag is on the page", you will certify all three as healthy.
Looks Installed
- Scripts load on every page
- Events fire on submit and click
- Console is clean, no errors
- Tag diagnostics glow green
Actually Counts
- The action receiving it is primary
- send_to carries a real conversion label
- The AW- id on the page is yours
- Call events match the action's type
The Fix Was One Mutation
Here's the part that should change how you budget your debugging time: after an audit spanning two layers and four separate faults, the fix took minutes.
One change promoted the server-side lead conversion action back to primary — and we re-read the setting afterwards to confirm the flag actually flipped, because a fix you haven't verified is a hypothesis. Two small page edits corrected the tag's account ID and removed the dead phone event rather than leaving a known no-op in place. That's the whole repair.
The verify-after-fix step deserves its own sentence, because it's the discipline that separates a repair from a guess. Tracking faults are silent in both directions: they break without errors, and they stay broken without errors. After promoting the action, re-read its status and confirm it now reports as primary and counted-in-conversions. After fixing the tag, trigger a test submission and watch for the event with the right label under the right account. Every one of the four faults in this account would have survived a fix that was applied but never checked.
The Conversions column came back to life — populated by the same kind of leads that had been arriving, uncounted, all along. Smart Bidding had a form signal again, account-wide, for the first time in weeks. And because the leads had never stopped, there was nothing to "rebuild": the campaign that looked dead on day seven was simply allowed to be measured.
That asymmetry is the real lesson. Weeks of silent damage, an afternoon of methodical audit, a one-line fix. The expensive part of conversion-tracking failures is never the repair — it's the time spent optimising a campaign against numbers that were lies, and the working campaigns paused because a column read zero. The Excellent-score trap costs you money the same way: not through one bad setting, but through every decision you make while trusting it.
Signs it's the plumbing, not the campaign
- Healthy CTR with a zero (or near-zero) Conversions column
- Real enquiries arriving while reports show nothing
- CRM leads carrying gclids that never appear in Ads
- All conversions keeps moving while Conversions sits flat
- Conversions died on one specific day — and nothing in Change History explains it
More From the Google Ads Trenches
The other ways accounts quietly bleed — and what actually fixes them.
Want this audit run on your account?
Conversion plumbing is the first thing we verify on every account we take over — before a single keyword changes. AI-managed Google Ads from $1,200/mo.
Explore Google Ads ManagementThe One-Hour Plumbing Audit
If your account reports zero (or suspiciously few) conversions while real enquiries keep arriving, run this before you touch keywords, budgets, or bidding:
- Read the status of every conversion action. In Goals → Conversions, check each action: primary or secondary? If nothing that matters to you is primary, Smart Bidding is starving — whatever the reports show.
- Check the send_to label. View the page source. The conversion event's
send_tomust carry a conversion label after the slash, not a bare account ID. No label, no conversion — and no error either. - Confirm the AW- ID is yours. Compare the account ID in the page's gtag against the one in your Google Ads account. A pasted-in stranger's tag fails silently and forever.
- Know your call-action semantics. "Calls from website" actions with a Google forwarding number ignore manual events by design. If your page fires events at one, those events do nothing.
- Cross-check the CRM. Pull recent leads, look for gclids, compare against the Conversions column. A mismatch proves a plumbing fault and ends the "is it the campaign?" debate on the spot.
- Date any demotion from the data. Chart Conversions against All conversions by day; the divergence point is your incident date. Change History will not help you here.
Forty-five minutes, none of it requiring anything more exotic than the Ads interface, your CRM export, and view-source. Only after the plumbing passes do campaign-level conclusions mean anything.
This is the audit we now run on every account before judging any campaign — it's part of how we run AI-managed Google Ads at Jordan James Media. If your account says zero and your gut says otherwise, your gut deserves the ten-minute gclid check before your campaign gets the blame.
Read every action's status
Goals → Conversions: is anything that matters to you still primary? Secondary actions feed nothing to bidding.
Check the send_to label
View source. A bare account id with no conversion label after the slash is a silent no-op.
Confirm the AW- id is yours
A tag pasted from another site credits a stranger's account — and never errors.
Cross-check the CRM for gclids
Gclid-carrying leads missing from the Conversions column prove a plumbing fault on the spot.
More field notes like this
Real diagnostics from real accounts — tracking, bidding and the traps Google never warns you about.
Choose your interests:
No spam, unsubscribe anytime. We respect your privacy.
Key Takeaway
- 1
Zero reported conversions + real leads in the CRM = measurement fault, almost never a campaign fault
- 2
Secondary conversion actions record into All conversions but feed nothing to Smart Bidding
- 3
Change History won't show a demotion — date it from the Conversions vs All-conversions divergence
- 4
A tag can fire perfectly and still count nothing: check the label, the AW- id, and the action type
- 5
Run the one-hour plumbing audit before you judge keywords, budgets or bidding
Social Media Carousel
7 cards • Download as ZIP (images) or PDF (LinkedIn)
The Secondary Conversion Trap
Zero reported. Real leads.
When Ads says zero and the CRM disagrees, the plumbing is lying — not the campaign
Around a hundred clicks, real form leads carrying gclids — and a Conversions column reading zero
Where Tracking Silently Dies
- 1
Actions demoted to secondary — bidding starves
- 2
send_to fired with no conversion label
- 3
gtag loaded under a stranger's AW- id
- 4
Manual events at a forwarding-number action
Looks Installed vs Counts
Tags fire, console clean, diagnostics green
Primary action, your AW- id, labelled send_to
Date It From the Data
Change History won't show a demotion. Chart Conversions against All conversions — the day they diverge is your incident date.
Plumbing Before Campaign
Run the one-hour audit before touching keywords, budgets or bidding — decisions on corrupted data compound.
Ads Says Zero. Inbox Disagrees
We run this audit on every account we manage
Share This Article
Spread the knowledge