Google Ads Period-Compare Audit: Why Lead Gen Accounts “Fix Themselves” on CPA… But Still Leak Quality

Introduction: same business, same goal, two very different audit windows

This case study comes from a local lead gen advertiser running Google Ads to generate inbound calls and form leads.

We reviewed two audits using the same evaluation lens (“Lead profile”), but in different time periods:

• February 2026: Growth score 49 (Broken) with lower data confidence
• June 2026: Growth score 73 (At Risk) with higher data confidence

On paper, June looks “better.” The selected-action CPA moved from well above target to roughly on-target.

But when we compared the audits side-by-side, the more useful story was: the account’s outcomes improved faster than its foundations.

Key takeaway: Period-over-period improvements can hide persistent tracking and structure risks that will cap lead quality and scaling.

The symptoms in February (why it looked “broken”)

February was a messy diagnostic window. The audit flagged no spend detected in that month, yet the account also showed conversions in the broader selected-action reporting.

That mismatch matters for local businesses because it usually means you’re looking at one (or more) of these situations:

• Campaigns paused or budgets set too low to generate meaningful auction data
• Tracking firing, but the ads that should be driving those leads aren’t consistently running
• Reporting windows don’t line up with what the business considers “active” marketing

The KPI symptom was clear: the selected-action CPA was far above the target.

And the diagnostic symptoms stacked up:

• Too many Primary conversions (6) for Smart Bidding to learn cleanly
• A Maximize Conversions strategy running while the account had 0 goal-aligned conversions in the audit window
• A non-lead action (e.g., WhatsApp click) treated like a primary success signal

If Google is optimizing to the wrong thing (or nothing), it will still spend—but it won’t learn what a good lead is.

What improved by June (and why the score rose)

By June, the audit had a lot more to work with:

• Data confidence increased (meaning the audit could “see” more signal)
• The account produced a moderate amount of conversions in the window
• Selected-action CPA moved close to target

Just as important, the bidding module looked healthier. In February, bidding health was essentially non-diagnosable (no spend + no goal-aligned conversions in-window). In June, Smart Bidding health scored well.

So what likely changed?

• The campaigns were actually running consistently
• The account accumulated enough conversion volume for Smart Bidding to behave more predictably
• Some of the earlier misalignment issues were reduced (fewer primary conversions overall)

However… June still wasn’t “clean.” It moved from Broken to At Risk because foundational items stayed unresolved.

The persistent problem: “Primary conversion bloat” (and why it quietly ruins lead quality)

Across both periods, the #1 recurring theme was conversion tracking integrity.

• February: 6 Primary conversions active
• June: 4 Primary conversions active

That’s improvement—but still above the recommended maximum (typically 3, and often 1–2 for local lead gen).

Here’s the practitioner reality: local businesses frequently track multiple actions—calls, forms, WhatsApp clicks, map clicks, “contact” page views—and then mark several of them as Primary.

The result is that Smart Bidding starts optimizing for the easiest-to-generate action, not the highest-quality lead.

Common patterns we see when Primary conversions aren’t disciplined:

• The algorithm “discovers” low-intent actions (short calls, repeat form submits, chat opens)
• CPA looks stable while lead quality declines
• Sales teams complain (“more leads, worse leads”) but the ad dashboard looks “fine”

In June, the audit still flagged that phone-call and form-submit outcomes were mixed as primary signals.

That’s not automatically wrong—but it’s risky unless you’ve established a clear hierarchy.

What worsened (or became more visible) in June: structure risk

February didn’t flag structure issues. June did.

Two structural findings appeared in June:

• The account relied on a single active campaign
• There was no dedicated branded Search campaign detected

This is a classic “it works until it doesn’t” setup for local businesses.

When one campaign carries the entire account:

• You can’t separate learning by intent (branded vs non-branded)
• You can’t protect your best-performing segment from experiments
• You can’t diagnose whether performance changes came from demand shifts or optimization changes

And when branded isn’t separated:

• Branded demand can inflate overall performance and hide weak generic acquisition
• You lose control over messaging for people already searching for you
• You can’t set different budgets/targets for “capture” vs “prospecting” traffic

Optimization priority shifts as performance improves: once CPA stabilizes, structure becomes the limiter to scaling safely.

The subtle tracking issue that stayed: “Every conversion” on lead forms

Both audits flagged the same medium-severity issue:

• Lead form conversion counting set to Every conversion

For e-commerce purchases, “Every” can be fine. For lead gen, it usually causes inflated numbers because the same person can submit multiple times (or trigger duplicate events).

Why this matters even when CPA looks on-target:

• Smart Bidding may “think” it’s getting more conversions than it truly is
• Automated bidding can become overly aggressive in pockets where duplicates are common
• You’ll struggle to reconcile ad platform leads with CRM reality

For most local lead gen accounts, One per click is a safer default for the main lead submission event.

Creative didn’t change much (and that’s a missed opportunity)

Both periods flagged similar creative hygiene items:

• At least one RSA with weak Ad Strength
• No image assets attached to active Search campaigns

These aren’t the sexiest fixes, but they matter for local businesses competing in crowded SERPs.

When you add image assets and strengthen RSA variety, you typically unlock:

• Better CTR stability (especially on mobile)
• More coverage across query variations
• Cleaner A/B insights because ads aren’t constrained by asset scarcity

This is also the kind of work you can do without changing budgets or landing pages.

What the diagnosis process looked like (the “same lens” comparison)

When we compare periods correctly, we’re not just comparing CPA—we’re comparing the account’s ability to learn and improve.

Here’s the same evaluation lens applied across both audits:

1. Can the account generate clean, goal-aligned conversion signal?

- Improved: fewer Primary conversions (6 → 4)

- Persistent: still too many primaries; call + form mixed without hierarchy

- Persistent: lead form counting still set to “Every”

2. Is Smart Bidding getting stable input?

- Improved: bidding health went from effectively broken to healthy

- February risk: Maximize Conversions without goal-aligned volume in-window

3. Is the campaign structure readable and scalable?

- Worsened/visible by June: single-campaign dependency

- New/visible by June: no branded separation

4. Is measurement tidy enough to trust?

- June flagged: multiple accessible measurement assets (GA4/GTM) increasing the chance of mis-mapping

The big lesson: better CPA doesn’t automatically mean better system design.

The fix plan we’d prioritize for a local business (next 30 days)

If this were a real engagement, the goal wouldn’t be “change everything.” It would be tightening the feedback loop so every optimization is based on the right signal.

1) Reduce Primary conversions to 1–2 true lead outcomes

Pick the most business-meaningful actions:

• Qualified form submission (preferred)
• Qualified phone call (with a duration threshold, if available)

Move everything else to Secondary:

• WhatsApp click / chat open
• Page views
• Micro-engagements

2) Fix counting for the main lead form conversion

• Change lead form counting from Every to One per click

This is a simple setting change that can dramatically improve the honesty of your reporting.

3) Split branded vs non-branded (even if budgets are small)

A pragmatic local structure:

• Branded Search (protect + measure demand capture)
• Non-branded Search (control acquisition efficiency)

If you want a third layer later, add:

• High-intent “service + location” segment (often the real profit driver)

4) Reduce single-campaign dependency with one controlled experiment

Don’t blow up what’s working. Duplicate intelligently:

1. Keep the current “winner” campaign stable
2. Create one additional campaign to isolate a key variable (intent, geo, offer)
3. Run for 2–4 weeks before judging

5) Upgrade RSA and add image assets

• Add unique headlines tied to local proof (neighborhoods served, turnaround time, certifications)
• Add image assets (logo, storefront/team, before/after where allowed)

What this means for optimization priorities going forward

The period comparison gives a clear sequence for local advertisers:

• When performance is broken: get clean conversion signal + stable spend
• When performance is “okay”: fix tracking integrity so Smart Bidding optimizes lead quality
• When performance is near target: improve structure so you can scale without losing control

June’s numbers suggest the account is close enough to target that the next gains won’t come from “more bid tweaks.”

They’ll come from:

• Choosing a single north-star lead action
• Making branded vs non-branded performance readable
• Creating a structure where you can test without risking everything

If you can’t explain which leads you’re optimizing for, you’re not really optimizing—you’re just spending.

Closing thought (and a low-pressure next step)

If you’re a local business running Google Ads and your CPA has improved but lead quality still feels inconsistent, you’re not alone. This is a common pattern: the account recovers faster than the measurement and structure.

If you want, you can try the VEOtool beta to run the same kind of period-over-period audit view, or request an audit and we’ll highlight what improved, what stayed risky, and what to fix first—without reinventing your whole account.