The Hidden Cost of Broken Conversion Tracking: What a Real Healthcare Account Audit Revealed

Broken conversion tracking costs mid-spend Google Ads accounts an estimated $26,000 per year in Smart Bidding waste — and most of it is invisible until you run a structured audit to find it.

An 8-hour audit on a regional healthcare ad account confirmed exactly that. The campaigns weren't broken. The conversions were firing. The reports looked normal. What was broken was the data those conversions were teaching Smart Bidding to act on.

This post documents the audit — what I found, how I found it, what I fixed, and why this kind of methodical attribution check is becoming more important as ad platforms lean harder on conversion signal quality for Smart Bidding decisions.

If your account is older than two years and has been through any platform migration — CRM, call tracking, attribution tool — there's a real chance some of these issues exist in your stack right now.

Why Conversion Health matters more in 2026

Three things have changed about paid media over the past five years.

Smart Bidding leans harder on conversion data than ever. Manual bid control has shrunk to the edges of every major platform. Maximize Conversions, tCPA, tROAS, Performance Max, Advantage+ — they all use the conversion signal you send as the primary input for every bid decision. Bad signal in, bad bids out. There's no manual override that fixes it.

Conversion data trains more than just bidding. Audience modeling, lookalike expansion, automated targeting, recommendation engines — they're all downstream of your conversion signal. Bad conversions teach the algorithms to find more of the wrong people, which generates more bad conversions, which trains the algorithms further in the wrong direction. Every part of the account compounds the problem.

Privacy changes have removed the redundancy that used to mask attribution problems. iOS 14, GDPR, third-party cookie deprecation — every signal you still have is more load-bearing because there are fewer of them. When the remaining signal is broken, there's no backup.

The combined effect: Conversion Health Score isn't a vanity metric. It's a leading indicator for almost everything else the account does — bidding decisions, audience expansion, ROAS predictions, channel allocation. When it degrades, every downstream system quietly degrades with it. By the time the cost shows up clearly in a dashboard, the account has been losing money for months.

The setup

The client was a regional healthcare lead-gen account spending mid-five figures per month across Google Ads and Meta. They'd brought me in for a CRM migration — moving off Salesforce onto HubSpot — and the attribution issues started surfacing as the migration work progressed.

On paper, the account looked healthy. Conversion actions were active. Conversions were firing. Lead volume was real. But performance had been quietly degrading for 18 months — cost per lead creeping up, Smart Bidding underperforming, nobody able to pin down why.

The migration gave me the reason to look. The audit framework was the way to look thoroughly. Most of those conversions were firing for the wrong reasons, in the wrong places, for the wrong actions.

How I found it: the 7 categories of a Conversion Health audit

The framework examines seven categories of attribution health:

1. Conversion action inventory and source mapping — cataloging every conversion action across every connected ad platform and identifying which tool or integration created each one.
2. CRM-to-ad-platform sync health — validating that lifecycle data flowing through HubSpot, Salesforce, or other CRMs is being sent back to ad platforms in a form Smart Bidding can use.
3. Call tracking integration validation — confirming that CallRail, Invoca, or other call tracking tools are reporting conversions to the correct, active conversion goals in Google Ads and Meta.
4. Smart Bidding signal quality (primary vs. secondary actions, category conflicts) — checking that each goal category has exactly one Primary action firing real conversions, with no conflicting signals splitting the bidding model.
5. Lifecycle stage attribution (lead → opportunity → customer) — verifying that downstream CRM stage changes (lead becomes customer) are being reported back to ad platforms as conversion events for full-funnel optimization.
6. Cross-platform attribution consistency — checking that the same conversion is counted consistently across Google, Meta, Microsoft, and any other connected platforms — no double-counting, no missing platforms.
7. Privacy and consent-mode compliance — confirming that consent mode (Google), Conversions API (Meta), and any HIPAA or GDPR requirements are configured correctly so signal still flows cleanly under privacy constraints.

The Conversion Health Score is a 0-90 composite measure of how well your conversion data is feeding Smart Bidding, rolling up action hygiene, integration completeness, and signal consistency across platforms into a single number. Scores below 30 are Critical. 30-55 is Degraded. 55-80 is Acceptable. 80+ is Automation-Eligible — Smart Bidding has enough clean signal to optimize aggressively without backup.

The initial diagnostic on this account produced a Conversion Health Score of 28 — Critical territory. That score told me something was broken; the audit told me what.

What I found: 5 conversion tracking issues

Five distinct issues, each contributing to attribution waste:

Issue 1: Salesforce-era conversion actions still active. Two Salesforce-sourced conversion actions ("First Time Phone Call" and "Repeat Phone Call") were still listed as active Primary actions in Google Ads. Both had recorded zero conversions in the past 90 days. They were poisoning Smart Bidding by signaling "no leads happening here" while real leads were happening through other paths.

Issue 2: CallRail integration silently broken. CallRail's Google Ads integration showed "Active" in the CallRail dashboard, but its conversion actions in Google Ads were the same Salesforce-era actions that had been deactivated. Calls were being tracked in CallRail but never reported as conversions to Google Ads. Smart Bidding had no idea phone leads existed.

Issue 3: Dual Primary actions in the same category. The Phone Call Lead goal category had two Primary actions — one for Google's native call extension tracking (working, firing 65 conversions per month) and one for the broken CallRail-sourced action (firing zero). Smart Bidding was splitting its optimization signal across both, which means it was optimizing on conflicting data.

Issue 4: Missing HubSpot lifecycle attribution. HubSpot's lifecycle stages (Lead → Marketing Qualified Lead → Patient) were configured properly post-migration. But there was no integration sending those stage changes back to Google Ads. The platform had no idea which clicks turned into actual patients — only which ones turned into raw form fills. This is the kind of account hygiene problem that compounds with bad pacing decisions — Smart Bidding starts making bad calls in both directions.

Issue 5: Orphaned conversion actions from old experiments. Three more conversion actions, all firing zero, all left over from previous tracking experiments. Each one added noise to the attribution model. It's also the kind of thing auto-applied Google recommendations quietly reinforce — the platform offers to "fix" what it sees as missing signal, but the underlying issue is structural.

Estimating the cost of broken conversion tracking

Broken conversion tracking has a quantifiable impact on ad performance. The size of that impact depends on three factors:

• How much you spend (more spend, more waste in absolute terms)
• How aggressively you use Smart Bidding (higher sensitivity to signal quality)
• How broken the tracking is (Conversion Health Score)

For this client's spend level and Smart Bidding configuration, I estimated:

• 30 percent signal degradation to Smart Bidding (typical for Conversion Health Scores in the 20-30 range)
• 0.9x bid sensitivity multiplier (Maximize Conversions strategy)
• Estimated waste: ~$2,160 per month, or about $26,000 per year

The cost-impact model isn't a simple multiplication of those three factors. Signal degradation cascades through bid quality, then auction performance, then cost per lead — each step amplifying the prior. The framework accounts for the cascading effect against historical performance baselines, which is why 30% signal degradation produces roughly 6-8% effective spend waste in this configuration rather than a flat 30%.

This isn't extra money the client paid out. It's optimization signal that Smart Bidding couldn't act on, leading to higher cost per lead than the account should have been hitting with clean data.

For context, Google's own documentation acknowledges that Smart Bidding performance degrades when conversion data is incomplete or noisy. They recommend at least 30 conversions in 30 days for Maximize Conversions to work effectively. They don't say much about what happens when those 30 conversions are split across conflicting signals — which is what was happening here.

How I fixed it: the 2-day rebuild

Once the issues were mapped, the fix took two days of focused work, slotted into the migration timeline.

Day 1 — diagnosis and architectural decisions. Mapped the current state. Identified the target state. Made strategic calls about which integrations to keep, which to decommission, and how to route signals cleanly.

Day 2 — implementation. Removed five dead conversion actions. Set up HubSpot's native lifecycle integration with Google Ads (firing daily on Lead stage with GCLID validation). Validated Meta's native lead form sync. Documented the new architecture.

The result is a cleaner peer-to-peer architecture: HubSpot is the source of truth for lifecycle data, native integrations send qualified lead signals to Google Ads and Meta, and the previously-required middleware layer was decommissioned (saving an additional $399 per month in software costs).

The Conversion Health Score is expected to climb from 28 into the 55-65 range after the first day of validated lifecycle sync data. Reaching the 80+ "automation eligibility" range will require additional work over the following few weeks. Fixing the lifecycle gap was the foundation, not the finish line.

The lesson: attribution decay is real and mostly invisible

Attribution decay is the gradual degradation of conversion signal quality over time. As tools change, integrations break, CRMs migrate, and old conversion actions accumulate unchecked, the data flowing into Smart Bidding becomes noisier even though the dashboards still show conversions firing. Decay is invisible until you audit for it — and the cost compounds because Smart Bidding's bad decisions train the algorithm in the wrong direction.

The pattern is consistent across the accounts I audit. These issues develop in predictable ways:

• A tool gets replaced, but its conversion actions don't get removed
• A CRM migration doesn't carry over all the lifecycle attribution
• A call tracking provider changes, but the old integration stays "active"
• A Smart Bidding strategy gets enabled, but the conversion data feeding it was configured for a different bidding model
• A new conversion goal gets added without auditing how it interacts with existing ones

None of these failures break the account visibly. The conversions keep firing. The reports keep showing data. But the underlying signal quality degrades, Smart Bidding makes worse decisions, cost per lead slowly climbs, and nobody can pinpoint why.

The only way to find these issues is to look. Methodically. Across every system involved. With a structured framework that knows what to check.

How I do this work

I use 20+ years of paid media operations experience plus AI-assisted diagnostic tools to accelerate the audit phase. The pattern recognition — knowing what a Salesforce-decommissioning failure mode looks like, or spotting dual Primary action splitting before checking the data — comes from having seen it many times before. The AI tooling makes it possible to cover more ground per engagement and surfaces inconsistencies that are easy to miss on a tired afternoon.

The judgment is mine. The architectural decisions are mine. The AI is part of how the work gets done, the same way analytics platforms and SQL workbenches are part of how the work gets done. Pretending it isn't would be dishonest. So would pretending it's the part that matters.

Frequently asked questions

What is a Conversion Health Score?

The Conversion Health Score is a 0-90 composite measure of how well your conversion data is feeding Smart Bidding. It rolls up three things into one number: action hygiene (no orphaned or dead conversion actions firing zero), integration completeness (CRM, call tracking, and lifecycle data flowing back to ad platforms correctly), and signal consistency across platforms (the same conversion counted the same way everywhere it's measured). Scores below 30 are Critical. 30-55 is Degraded. 55-80 is Acceptable. 80+ is Automation-Eligible — Smart Bidding has enough clean signal to optimize aggressively without backup.

What is attribution decay?

Attribution decay is the gradual degradation of conversion signal quality over time. As tools change, integrations break, CRMs migrate, and old conversion actions accumulate unchecked, the data flowing into Smart Bidding becomes noisier — even though the dashboards still show conversions firing. The conversions keep counting; the underlying signal quality keeps eroding. Decay is invisible until you audit for it, which is why most accounts older than 18 months have some version of it whether they know it or not.

How do I estimate the cost of broken conversion tracking?

The cost-impact model uses three inputs: monthly ad spend, Smart Bidding signal degradation (correlated to your Conversion Health Score), and bid sensitivity (a function of which bidding strategy you're running). The math isn't a simple multiplication of those three factors — signal degradation cascades through bid quality, auction performance, and cost per lead, each step amplifying the prior. For a Conversion Health Score in the 20-30 range with Maximize Conversions enabled, expected effective waste is roughly 6-8% of monthly ad spend on a sustained basis.

If you think your account has some of these issues

This is the kind of work I do at Corriston Consulting — diagnosing and fixing conversion tracking and attribution issues hands-on, with documented architecture and validation. If your spend is significant enough to make signal degradation a real cost, it's worth a look.

Conversion Health Rescue is the productized version of this engagement. For broader paid media work — pacing, structure, account architecture — see SEM services or marketing audits. If you're not sure which fits, contact me and we'll figure it out.