Free PPC

Toolkit

Smart Bidding is only as good as the signal it gets. If platform conversions are off by 15 percent against the CRM, the algorithm spends the next two months optimizing toward the wrong outcome at an accelerating rate. This is the validation I run before any bidding or budget call gets made on the numbers.

The framework

Every PPC decision starts here. Three sources, one source of truth, a variance band that decides what you can do next.

The hierarchy

CRM (truth) > GA4 (neutral) > Google Ads (self-attributed)

Variance bands

Direction diagnostic

Re-enable Smart Bidding gate

After any tracking fix, all four must be true before re-enabling:

• 14 consecutive days under 15% variance
• 30+ conversions in the clean period
• No additional tracking changes during the wait
• Sales cycle complete for at least one cohort

What this framework solves

When tracking is broken, every optimization built on top of it is wrong. Smart Bidding optimizes toward phantom signals. CPA calculations are systematically wrong. Budget decisions get made on inflated or deflated numbers. This framework treats tracking as the prerequisite, not an optimization, and gives you the thresholds to know when the data is clean enough to trust and when it is not.

The core idea

The CRM is the source of truth, never the platform. Google Ads is self-serving by design because the platform benefits from attributing more conversions to itself. The hierarchy is always CRM (truth) > GA4 (neutral) > Google Ads (self-attributed). Direction matters more than magnitude: under-reporting leaves money on the table, but over-reporting actively burns it because Smart Bidding overbids on traffic that does not actually convert. And Smart Bidding amplifies tracking errors. Manual bidding with bad data wastes money at a linear rate; Smart Bidding with bad data wastes it at an accelerating rate because the algorithm learns from the wrong signal and compounds the error each cycle.

How to apply it

1. Run 3-way validation. Compare Google Ads, GA4, and CRM for the same period. All three within 10 percent: tracking is healthy. Two agree, one diverges: the divergent one points at the problem. All three diverge: the tagging infrastructure is broken.
2. Classify variance against the bands. 0-10 percent: acceptable. 10-25 percent: investigate. 25-50 percent: materially broken, fix before any optimization. Above 50 percent: crisis, stop Smart Bidding immediately and switch to manual or Maximize Clicks.
3. Diagnose the direction first. Over-reporting: check duplicate tags, “Every” counting on lead forms, non-primary actions inflated into the conversions column. Under-reporting: tag firing failures, cross-domain breaks, Enhanced Conversions disabled, conversion window shorter than the sales cycle.
4. Audit the conversion settings, not just the tags. Counting method (One vs. Every), attribution model, conversion window, include-in-conversions toggle, and conversion value all shape what Smart Bidding sees. “Every” on lead forms roughly doubles reported conversions and halves apparent CPA. It is one of the most common configuration errors and lives in 30 percent of audited lead gen accounts.
5. Match conversion window to actual sales cycle. 7-day window on a 30-day cycle drops every conversion past day 7. B2B SaaS standard: 90-day click-through window, plus the Time Lag report to discount the most recent 30-45 days as still-maturing.
6. Move micro-conversions to secondary. Adding page views or content downloads as primary conversions to “give Smart Bidding more data” pollutes the signal. Either differentiate values (50:1 between demo and whitepaper) or exclude micro-conversions from the primary set entirely.
7. Wait for clean data before re-enabling Smart Bidding. After a tracking fix: 14 consecutive days under 15 percent variance, 30+ conversions in that clean period, no additional tracking changes during the wait. Rushing back forces the algorithm to learn from a mixture of corrupted and clean data.
8. Implement offline conversion import for lead gen. Capture GCLID at form submission, store with the lead, feed back qualified or closed-won outcomes to Google Ads. This is the only way to optimize toward business value instead of cheap form fills.

When to use it

Use this as step zero of any audit, before any bidding change, after any website or GTM update, when CRM and platform numbers diverge, and when Smart Bidding performance suddenly degrades. It calibrates by vertical: ecommerce focuses on dynamic value capture, cross-domain tracking, and new-vs-returning customer separation; lead gen lives or dies on the counting method and offline conversion import; B2B SaaS extends the conversion window to 90 days and discounts recent data through the Time Lag report; local services has to define how phone calls are tracked and what minimum duration counts as a real conversion (60 seconds is the standard); high-value verticals add compliance constraints (HIPAA, attorney-client privilege) and require value differentiation across case types. It does not apply when there is no CRM or revenue data at all. In that case, the 3-way triangle collapses to a 2-way comparison with no ground truth, and the recommendation is to build the missing truth source before optimizing further.