The Hidden Side Of Rising ROAS: What Advertisers Need To Know

A rising ROAS is the kind of number nobody questions. It goes into the monthly deck, it justifies budget increases, and it quietly becomes the benchmark for next quarter. The problem is that a ROAS figure is only as honest as the conversion data underneath it. When that data includes values someone typed into a settings field, leads counted twice, or purchases recorded by two different tags, the “improvement” may have nothing to do with your ads.

A recent PPC Hero column makes this point well. The author describes a student whose account showed a ROAS above 500% alongside zero actual sales. That case was easy to diagnose because the reported number was entirely fictional. The harder cases are accounts that genuinely sell, where only part of the reported value is inflated. In those accounts, a broken conversion setting doesn’t look like a problem. It looks like a great year. Before you debate attribution models or fund an incrementality test, check whether your conversion value was actually measured at all.

Ask Where The Number Came From First

Any ROAS jump should trigger one question before it triggers a celebration.

• Did real revenue in your back-end grow at the same rate as reported conversion value?
• Were any conversion actions added, removed, or switched from secondary to primary during the period?
• Did anyone edit conversion values, counting settings, or goal configurations?
• Were new tags, GA4 imports, or CRM uploads connected recently?
• Is a large share of conversion value coming from non-purchase actions like leads or sign-ups?
• Does the trend line coincide with a site redesign, checkout change, or thank-you page update?

Treat every unexplained ROAS gain as a tracking question before you treat it as a performance win. If you can’t trace the gain to orders in the bank, you don’t have a result yet. You have a hypothesis.

Static Values Are Typed, Not Measured

Static conversion values are one of the most common and least visible sources of ROAS inflation.

• A static value is a fixed number someone entered in a conversion action’s settings, not a figure pulled from a real transaction.
• Google Ads adds those static values to conversion value right alongside actual order revenue.
• ROAS is then calculated by dividing that combined total by cost.
• Reported totals don’t separate measured revenue from typed-in estimates, so you can’t tell the difference at a glance.
• Static values are common on lead forms, calls, sign-ups, and other micro-conversions.
• The value is only as accurate as the assumptions of whoever set it, often months or years ago.

A static value isn’t inherently wrong, but it’s an estimate dressed up as revenue, and it deserves the scrutiny you’d give any estimate.

Changing A Static Value Mid-Period Breaks Your Comparisons

The real damage happens when a static value changes between the periods you’re comparing.

• PPC Hero’s illustration: an account with 1,000 orders worth $100,000, plus 2,000 leads valued at $10 each, on $20,000 of spend reports $120,000 in value and a ROAS of 6.0.
• If someone raises the lead value to $20 before the next quarter, the identical orders, leads, and spend would report $140,000 and a ROAS of 7.0.
• That’s a one-sixth ROAS increase driven entirely by an edited field, with no change in actual business results.
• Because the new value applies only to new conversions, nothing looks broken in the interface.
• Last year’s ROAS and this year’s ROAS end up measured in different units, yet the chart draws them as one continuous line.
• According to the same column, a static value is fine when it reflects your lead-to-sale rate multiplied by your average sale.

If static values changed during a reporting window, any year-over-year ROAS comparison across that change is effectively meaningless without adjustment.

The “Every” Counting Setting Quietly Doubles Leads

How a conversion action counts repeat events matters more than most advertisers realize.

• Google Ads lets you count “Every” conversion per click or just “One” per click.
• “Every” makes sense for purchases, because a single customer can genuinely buy more than once.
• For leads, Google’s own help documentation notes that usually only one unique lead per ad click adds value.
• With “Every” on a lead form, someone who submits, spots a typo in their phone number, and resubmits becomes two leads.
• A thank-you page that loads twice can also generate two conversions from one person.
• The PPC Hero author reports that in four out of five accounts he analyzes that feed leads into bidding, at least one lead action in the conversions column is set to count every conversion, and across all lead actions in that column it’s roughly half.

Counting settings are a silent multiplier on lead volume, and since bidding algorithms optimize toward that volume, the error compounds over time.

Duplicate Purchase Counters Inflate Revenue

Extra purchase tracking is the second major source of phantom conversions.

• According to PPC Hero, more than one in four accounts with purchase tracking have two or more purchase actions in the conversions column.
• Sometimes these actions legitimately measure different sales, such as online versus offline orders.
• Often they count the same order twice: once through the Google tag and again through a GA4 import or CRM upload.
• Duplicates often appear during migrations, when a new tracking method is added but the old one is never removed.
• Agency handovers and team turnover make it easy for nobody to remember why a second action exists.
• Two counters recording one sale will double that sale’s contribution to ROAS.

Every purchase action in your conversions column should have a clear, documented reason to exist, and a different reason from every other purchase action.

The Dangerous Duplicate Is The One That Looks Healthy

Most audits go hunting for the wrong kind of duplicate.

• A counter that drifts away from your real numbers is usually caught, because it obviously looks broken.
• The dangerous duplicate matches your back-end almost exactly, so it looks trustworthy and nobody touches it.
• It can sit as a secondary action for months without affecting reported results.
• The moment someone adds it to a goal or promotes it to primary, conversion value can double overnight.
• Because the numbers “look right” individually, the problem only shows up when both are counted together.
• Bidding strategies then chase inflated value, potentially pushing spend in directions your real revenue doesn’t support.

A close match between two counters isn’t reassurance. It’s a warning sign that you may be measuring the same thing twice.

How To Audit Your Conversion Settings

A focused audit doesn’t need to be complicated, but it does need to be systematic.

• List every conversion action and mark whether it’s primary or secondary and which goals include it.
• For each action, note the source: Google tag, GA4 import, CRM or offline upload, or call tracking.
• Check counting settings and switch lead actions to “One” unless you have a specific reason for “Every.”
• Review every static value and document how it was calculated and when it last changed.
• Pull the change history for conversion settings across your reporting period.
• Flag any two actions that track the same event through different methods.
• Confirm that every primary action represents a distinct, valuable business outcome.

The output should be a simple inventory anyone on the team can read, not a set of assumptions that live in one person’s head.

Reconcile Reported Value Against Real Revenue

The ultimate test of any conversion setup is whether it matches the money that actually arrived.

• Compare reported purchase conversion value against back-end order revenue for the same period.
• Compare reported lead counts against unique leads in your CRM, after deduplication.
• Separate measured revenue from static-value contributions when you report ROAS.
• Calculate a “revenue-only ROAS” that excludes estimated lead values entirely.
• Investigate any period where reported value grew faster than real revenue.
• Recalculate historical comparisons using consistent values whenever a static value changed mid-period.

Reconciliation won’t be perfect, since attribution windows and channel overlap will always create some gap, but large or growing gaps are signals you can’t afford to ignore.

Put Governance Around Conversion Changes

Most of these errors aren’t malicious. They come from small, undocumented edits that nobody revisits.

• Restrict who can edit conversion actions, values, and goal settings.
• Log every change with a date, the person responsible, and the reason.
• Annotate reports and dashboards whenever a conversion setting changes.
• Re-audit after site launches, checkout changes, tag migrations, and agency transitions.
• Apply the same discipline across every ad platform, not just Google Ads, since any channel reporting conversions can suffer the same issues.
• Treat conversion configuration as financial infrastructure, not a one-time setup task.

Governance turns conversion tracking from a fragile setup into a reliable source of truth your budget decisions can actually rest on.

Final Thoughts

ROAS is a ratio, and ratios are easy to move without moving the business. A static value edited between quarters, a lead form counting every submission, or a dormant purchase action suddenly added to a goal can all produce charts that look like progress. As the PPC Hero column points out, the most dangerous errors aren’t the ones that look broken. They’re the ones that look healthy, sit quietly for months, and then inflate results the moment someone flips a setting.

The fix isn’t exotic. Before you pay for incrementality studies or argue over attribution models, audit the conversion layer underneath them: what’s counted, how often, at what value, and from which source. Reconcile it against real revenue, document every change, and make sure your comparisons measure the same thing from period to period. Do that, and your next ROAS increase will mean something.

A great ROAS is worth celebrating only after you’ve proven it came from customers, not configuration.