Many paid ads accounts have a problem in the settings and with values.

On the surface, nothing looks broken with the ad account. The campaigns run, keywords match intent, reporting appears healthy. But underneath all of that sits a corrupted value signal feeding Smart Bidding the wrong story and data.

When advertisers notice the discrepancy, their immediate reaction is to blame the ad platform for over-reporting. In reality, the platform isn’t broken; it’s following the instructions it was given.

When an account is running tCPA or tROAS on autopilot without validating the underlying conversion math, Smart Bidding optimizes toward phantom value rather than actual revenue.

Over time, what tends to happen is that the paid ad account drifts into an inflated state.

This presents as a wide gap between what ads are generating and what the business sees day to day.

A good phrase for this is “value inflation,” and it is what happens when the dollar figure Smart Bidding is optimizing toward doesn’t reflect what the business actually collects in revenue.

Value inflation shows up as:

• Double-counted micro-conversions padding conversion counts.
• Primary and secondary goals fighting each other for bid signal.
• Offline conversion imports pushing in deal values that were never closed.
• Old conversion value rules created in 2023 and never revisited.

This isn’t the result of fraud or platform trickery. It’s almost always a self-inflicted tracking issue that can happen for several common reasons.

And value inflation matters more now than it did five years ago because Smart Bidding doesn’t grade its own homework.

tCPA and tROAS accept whatever value has been set as truth. If the numbers are inflated, the algorithm will aggressively pursue more inflated numbers, chasing ghost conversions, raising CPCs to win auctions for “high value” clicks that were never actually high value, and burning budget in the exact direction it was instructed to go.

The account looks like it’s performing well on the surface.

This is the audit to run before trusting a single tROAS target in an unverified account.

What Value Inflation Actually Looks Like

Value inflation rarely shows up as one glaring error that is obvious. It’s usually three or four small distortions stacked on top of each other, each one nudging the algorithm a little further from reality.

• Double-counted micro-conversions. A “form submit” and a “thank you page view” both firing as separate conversions for the same lead, each carrying its own value, which is doubling the reported conversion volume.
• Mis-weighted primary vs. secondary goals. Newsletter signups, PDF downloads, or add-to-carts marked as primary conversions alongside actual purchases, diluting the bid signal with noise that has nothing to do with revenue.
• Inflated offline conversion imports. Deal values uploaded from the CRM before a deal is actually won, or sales-qualified leads imported at full contract value when the real close rate is 20%.
• Stale or misconfigured conversion value rules. Location, device, or audience-based value rules built for a promotion that ended eight months ago, still multiplying values in the background.
• Dynamic value feed drift. E-commerce feeds passing cart value instead of actual order value after discounts, taxes, and returns are factored in.

Individually, each of these looks like a rounding error in the ad account. Stacked together each one of the issues  can inflate the reported account value by 20% to 40% without a single alarm going off in the interface, because Google Ads has no way of knowing the “conversion value” isn’t real.

Google Ads and Microsoft Advertising just see a number, and its job of the bidding algorithm is to get more of the goal number.

Why Smart Bidding Can’t Tell the Difference

It is really easy when looking at platform data all day to end up confused: tCPA and tROAS are not wrongly reported. The bidding is doing exactly what it was instructed to do with the data provided in conversion settings.

Target ROAS bidding works backward from a value target. The account will bid and spend more where the expected return-per-dollar clears the bar, spend less where it doesn’t. If half the “conversions” are inflated, the algorithm’s model of what a good customer looks like gets skewed toward whatever behavior generated those inflated values, not toward the best actual buyers. It will bid up placements, audiences, and search terms that produce noise, because noise is what is rewarded.

The compounding problem is the learning period.

Smart Bidding recalibrates continuously based on recent conversion data. Inflated values don’t just distort today’s bids but instead they train the model for weeks, sometimes months, depending on the conversion volume. By the time the CFO asks why CAC is climbing while the platform says ROAS is healthy, the algorithm has already built an entire bidding strategy around a fiction.

The Value Inflation Audit Framework

Run this checklist in order. Each step either confirms the data is clean or points directly at the solution

Step 1: Audit the Conversion Action Weighting

Open Goals > Conversions > Summary and pull every active conversion action with its category, count, value, and whether it’s marked primary or secondary. This helps determine two things: conversion actions marked primary that shouldn’t be influencing bids, and value-based goals sitting next to count-based goals with no differentiation between them.

A clean setup usually has one to three primary conversion actions tied directly to revenue, which would be purchase, qualified lead, booked appointment. Everything else, which might be newsletter signups, chat opens, video views, belongs in secondary, observation-only, or excluded from bidding entirely.

Step 2: Check the Attribution Model

Confirm every conversion action is running data-driven attribution and not still defaulting to last-click on an older action that predates the account’s migration. Mixed attribution models across conversion actions in the same account will produce wildly different value patterns for what should be comparable conversions, and it’s one of the most common issues in inherited accounts that nobody thought to check after Google forced the DDA transition.

Step 3: Hunt for Duplicate Tags and Double-Firing

Pull the Tag Diagnostics view in Google Tag Manager or Google Analytics 4’s DebugView and fire a real test conversion through the funnel. Watch for the same conversion event firing twice, once from a hard-coded gtag snippet still sitting in the page code and once from GTM, or a “confirmation page load” firing independently of the actual form submission event it’s supposed to represent. This single issue alone could be responsible for reported conversion volume.

Step 4: Reconcile Value to Pipeline

Pull the last 90 days of Google Ads-attributed conversion value and put it next to actual closed revenue or actual fulfilled orders from the CRM or order management system for the same window.

• If Google Ads says $400,000 in conversion value and the business recognized $270,000 in matching revenue, then there is a mismatch worth chasing.
• Break the gap down by conversion action because it’s rarely evenly distributed. Usually one or two conversion actions (often OCT imports or a lead-value feed) are carrying almost all of the inflation.

Step 5: Cross-Examine in GA4

Use GA4’s Advertising snapshot and a custom exploration comparing Google Ads-reported conversions against GA4’s own purchase or key event counts for the same campaigns and date range. A large, consistent gap between what Google Ads claims it drove and what GA4 recorded as an actual purchase event is a signal that value is getting inflated somewhere between the click and the conversion action firing, which is not a GA4 tracking problem to dismiss.

Where to Find the Proof: Key Reports to Pull

Before updating a single setting and fixing the entire account, document and build the paper trail. It is important to be able to show, in numbers, exactly where the inflation is coming from.

• Conversion value rules report (Goals > Conversion Value Rules). This shows adjusted value by rule, which can show precisely how much value each rule is adding and whether that lift still matches business reality.
• Conversion action diagnostics. This flags conversion actions with no recent conversions, tracking issues, or unusual value patterns worth a second look.
• Segment conversions by conversion action in the campaigns view. This will show the value contribution per action instead of one blended number.
• GA4 Explore comparisons. Session-level, event-level breakdowns that don’t inherit any of Google Ads’ attribution assumptions.
• CRM or order management export. The only report on this list that documents what actually happened, which is why Step 4 above depends on it.

The Fixes: Rebuilding a Clean Value Signal

Finding the inflation is half the job. The other half is rebuilding the value structure so it doesn’t creep back in six months from now.

Set Primary vs. Secondary Conversion Actions Correctly

Move every non-revenue action, which would be content downloads, video engagement, chat starts, or account signups, out of primary and into secondary or “observation” status.

The bid strategy should only ever be optimizing toward actions that represent real, recognized value to the business. If a lead needs to be qualified before it’s worth anything, don’t let the raw lead volume drive bids on its own; that belongs in an offline conversion import once it’s qualified, weighted appropriately below.

Clean Up Offline Conversion Tracking and Dynamic Value Feeds

For OCT, import value at the stage that reflects reality, not the stage that looks best. If the close rate on sales-qualified leads is 20%, either import at a probability-weighted value or import the full value only once the deal is actually won, with a value adjustment made for deals that fall through later. Google Ads has just sunset new offline conversion imports through the legacy Ads API in favor of the Data Manager API, so if the OCT pipeline hasn’t been touched recently, this is the moment to rebuild it correctly rather than patch the old one.

For e-commerce, confirm the dynamic value feed is passing net order value, which is post-discount, post-tax where applicable, and adjusted for returns via conversion adjustments and not gross cart value at checkout initiation.

That single field is the difference between a tROAS target that’s grounded in reality and one that’s chasing shopping cart abandonment.

Audit or Rebuild Conversion Value Rules

Pull up every active value rule and ask, for each one, whether the business condition that justified it is still true today. A location-based rule built around a regional promotion, a device-based rule from a mobile-conversion test that ended, an audience-based rule for a segment that is no longer a target are things to look at. These settings accumulate and compound over time. Since Google now requires a conversion value on new conversion actions, it’s worth checking that nobody defaulted to a placeholder value just to get past setup, because that placeholder has a way of never getting revisited.

Recalibrating tROAS After the Clean-Up

This is where a lot of otherwise-careful audits can go wrong. When the data is fixed and updated, the reported average order value drops because it’s now accurate, and if the tROAS target doesn’t move with it, it can trigger a bidding shock that tanks impression share overnight.

• Recalculate the real target before updating anything in the account. If the true blended ROAS was 380% on inflated data and the clean number is 310%, the target needs to reflect 310%, not 380%, with the hope that performance improves back up to it.
• Move the target in stages, not one jump. A 15% to 20% adjustment every five to seven days gives Smart Bidding room to recalibrate without a full learning-period reset.
• Expect a temporary dip in reported conversion volume. That’s the inflation leaving the system, not the campaign failing. Track it against real revenue, not the dashboard.
• Hold for at least two full weeks post-cleanup before making a second round of target adjustments. Judging a newly recalibrated bid strategy against a partial data set just reintroduces the same guessing problem that is being fixed.
• Document the before-and-after gap for stakeholders. When leadership sees CPA rise on paper right after a “clean-up,” they will ask questions. Have the pipeline reconciliation from Step 4 ready before they do.

Key Takeaways

• Value inflation is any gap between what Google Ads reports as conversion value and what the business actually collects. Double-counted micro-conversions, mis-weighted goals, inflated OCT imports, and stale value rules are the four most common causes.
• Smart Bidding treats whatever value it has been fed as ground truth. Inflated inputs produce inflated bids, higher CPCs, and budget spent chasing conversions that were never real.
• Audit in this order: conversion action weighting, attribution model consistency, duplicate tag firing, value-to-pipeline reconciliation against actual revenue, and a GA4 cross-check.
• Fix it by separating primary revenue actions from secondary engagement actions, correcting OCT value timing and e-commerce feed logic, and rebuilding conversion value rules around current business conditions.
• Recalibrate tROAS targets in stages after cleanup because a sudden, unadjusted target against newly accurate data is what causes the bidding shock advertisers fear most.

Value inflation doesn’t announce itself. It is easy to ignore.

It is an issue that lives in conversion settings, making every report look a little better than it should, until the gap between reported performance and actual revenue gets too big to explain away.

This audit can be used on a quarterly cadence and not just when something feels off and misaligned with the business results.

By the time performance visibly breaks, the bidding algorithm has usually been trained on bad data for months.

Clean data is what lets Smart Bidding actually do its job in the current era of PPC. Everything else is just a more expensive way of guessing.

More Resources:

Featured Image: Anton Vierietin/Shutterstock