The “43 Conversions” Problem: An Anonymized Google Ads Audit Case Study (And How Agencies Can Fix It Fast)

Introduction

This case study comes from a Google Ads audit we ran for a growth-focused advertiser (think: performance-minded team, active acquisition spend, and a desire to scale).

On paper, the account looked “healthy” in all the places people typically check first: campaign structure was clean, targeting wasn’t doing anything wild, and there were no obvious landing page or creative red flags.

And yet the growth score came back broken.

The symptoms were the kind that agencies get pulled into all the time:

• Smart Bidding wasn’t producing consistent business outcomes
• Reporting looked busy (lots of “conversions”), but pipeline/revenue didn’t match the story
• Optimizations felt random because the account wasn’t learning from the right signals

Key takeaway: When conversion tracking is noisy, everything downstream (bidding, budgets, scaling decisions) becomes guesswork.

Context: What we audited (and what we didn’t assume)

Audit window: the last 12 months.

We treated this like a standard “can this account scale?” investigation. That means we didn’t start with the assumption that ads were bad, the landing page was bad, or the market was bad.

Instead, we asked a simpler question first:

Is Google optimizing toward something that the business actually wants?

Because in 2026, most accounts are not manually managed. They’re guided by Smart Bidding—and Smart Bidding is only as smart as the conversion actions you feed it.

The audit process: How we diagnosed the real issue

When an account claims it has “plenty of conversions” but stakeholders say “those aren’t real,” we follow a tight sequence:

1. Check the account’s Primary vs. Secondary conversion actions
2. Scan for low-intent events mistakenly marked as Primary
3. Verify whether any goal-aligned outcomes have fired recently
4. Check how many conversions Smart Bidding is averaging together
5. Look for measurement/attribution gaps that would make reporting unreliable

In this account, step #1 immediately explained most of the performance confusion.

What the audit revealed: “Primary conversions” were basically app analytics

The headline finding:

• 43 Primary conversions were active (best practice max is usually ~1–3 per goal)
• Multiple PAGE_VIEW-like events were set as Primary conversions
• There were no goal-aligned conversions detected in the last year

To make this concrete, several Primary conversions looked like:

• “medical_consent” (effectively a consent/page interaction)
• “screen_view” (an app screen view)
• “session_start” (an app session start)

These are not inherently “bad” events. They’re useful for product analytics and UX.

But when they’re set as Primary in Google Ads, the system treats them as the thing to maximize.

So the account ends up in a loop:

• Smart Bidding chases the easiest-to-generate events
• CPA looks great (because “screen_view” is cheap)
• Budgets shift toward traffic that triggers those events
• The business wonders why leads/sales didn’t scale with “conversions”

Key takeaway: If page views and sessions are Primary conversions, you’re not running performance marketing—you’re paying to generate analytics events.

Root cause #1: Too many Primary conversions (and Smart Bidding averaged them)

Google’s automation doesn’t “rank” your Primary conversions the way humans do.

If you mark 43 actions as Primary, Smart Bidding can optimize toward any mix of them—often drifting to the cheapest, highest-volume events.

This is especially damaging when:

• Micro-events fire far more often than purchases/leads
• Micro-events happen earlier in the funnel
• Some events are auto-collected or triggered passively

The result is a blurred optimization target.

For agencies, this is the silent killer: you can spend weeks iterating on ads and audiences while the model is fundamentally trained on the wrong label.

Root cause #2: No goal-aligned outcomes were being captured (or recognized)

The audit flagged that no goal-aligned conversions were detected across the full year.

That typically means one of three things:

• The “real” conversion exists but is set to Secondary (so it doesn’t guide bidding)
• The conversion is misconfigured (wrong trigger, wrong domain/app mapping, wrong event name)
• The real conversion happens offsite/CRM-side and isn’t imported (or import broke)

For lead gen businesses, this is often “form_submit” missing or firing inconsistently.

For e-commerce, it’s frequently purchase tracking not implemented correctly, duplicated, or blocked.

For app-based funnels, it’s often the difference between “app_open” and “sign_up_complete” / “subscribe.”

When the real conversion isn’t present (or isn’t Primary), Smart Bidding has nothing meaningful to learn from—so performance becomes unstable and scaling becomes risky.

Root cause #3: Measurement and attribution were effectively unusable

One module score stood out: Attribution & Measurement scored 0, with a large number of issues.

In practice, that usually shows up as:

• Conversions that don’t map cleanly to the business goal
• Event naming chaos (multiple versions of the “same” thing)
• Platform events (app analytics) mixed with acquisition outcomes
• No consistent “source of truth” between Ads, Analytics, and CRM

Even if the campaigns were structurally sound, measurement quality was not.

And for agencies, that’s the uncomfortable truth: if measurement is broken, your performance narrative is fragile. You can’t defend decisions, and you can’t scale confidently.

The fix: How we cleaned the conversion system (without breaking learning)

This is where many teams overcorrect. They delete everything, flip switches all at once, and then wonder why performance tanks for two weeks.

The safer approach is to reduce noise while preserving continuity.

Here’s the practical sequence we recommended.

1) Decide the single “north star” conversion (per goal)

Pick one primary outcome per goal (two at most). Examples:

• Lead gen: “qualified lead” (or “lead” if that’s all you can track today)
• E-commerce: “purchase”
• App: “subscription” or “trial_start” (not “session_start”)

If you can only track early-funnel events today, choose the closest real proxy—but treat it as a stepping stone, not the permanent KPI.

2) Demote micro-events to Secondary

Events like these should almost never be Primary:

• Page views / screen views
• Session starts
• Consent clicks
• Scrolls / time-on-site

Keep them as Secondary for analysis and funnel visibility.

The audit’s explicit recommendation was to set those PAGE_VIEW-type actions to Secondary and keep only true outcomes as Primary.

3) Reduce Primary conversions to ~1–3

We recommended moving from dozens of Primary conversions down to a small set.

A common agency-friendly setup:

• 1 Primary: the main outcome (purchase/lead)
• 1 Primary (optional): a verified secondary outcome (e.g., “purchase” + “first_purchase” or “lead” + “qualified_lead”)—only if both are clean
• Everything else: Secondary

This prevents Smart Bidding from “choosing its favorite easy event.”

4) Verify the goal conversion fires reliably

Before changing bidding strategies, confirm:

• The conversion triggers once (no duplicates)
• It attributes to Google Ads clicks properly
• It’s consistent across web/app journeys if the funnel spans both
• It matches your reporting definitions (what the client thinks a lead is)

5) Make changes in a controlled rollout

We generally recommend:

1. Set micro-events to Secondary
2. Keep the primary outcome(s) stable for 1–2 weeks
3. Watch conversion volume, CPA/ROAS volatility, and search term quality
4. Then evaluate whether bidding strategy needs adjustment

Key takeaway: Fix tracking first. Then evaluate bidding. Doing it in the reverse order creates false conclusions.

What agency owners should learn from this audit

This account is not unusual. It’s a pattern we see when:

• App events are imported without governance
• GA4 auto-events get promoted unintentionally
• Multiple teams touch tracking (product, analytics, paid media) without a shared definition of “conversion”
• Legacy conversion actions accumulate over time

Here are the practical lessons that will save your team hours:

• Conversion governance is a service line. Include it in onboarding and quarterly audits.
• Primary conversions are a scarce resource. Treat them like you’d treat a P&L metric.
• If “conversions” don’t correlate with business outcomes, assume signal pollution. Don’t assume creative fatigue.
• Measurement gaps create client trust gaps. Fixing attribution is often the fastest way to stabilize the relationship.

A simple checklist you can use in your next audit

Use this with any new account (especially ones “running fine” but not scaling):

• Are there more than 3 Primary conversions per goal?
• Are any Primary conversions clearly micro-events (views/sessions/consents)?
• Did the true business outcome fire in the last 30 days?
• Is there one agreed definition of “lead”/“purchase” across Ads and CRM?
• Do conversion names reflect outcomes (not technical event labels)?

If you find even one issue here, fix it before you “optimize.”

Closing: The fastest wins are often invisible

What made this audit valuable wasn’t a clever targeting hack or a new campaign type.

It was correcting what the account considered a “win.” Once you do that, Smart Bidding can finally learn the right thing, reporting becomes defensible, and optimization work starts compounding.

If you want, you can run your own account through VEOtool’s beta audit to surface these issues quickly—or request an anonymized, agency-friendly audit review you can apply across clients.