Siti Muinah
A stark encounter during a recent online educational session revealed a critical flaw in how many businesses track the success of their digital advertising campaigns. A student, while presenting her screen, showcased a meticulously organized spreadsheet detailing her advertising performance. Amidst rows of promising data, a Return on Ad Spend (ROAS) exceeding 500% stood out, a figure that would typically signal exceptional profitability. However, a closer inspection of the table’s final entry revealed a confounding detail: "no sales." This paradoxical outcome immediately flagged a significant issue with the underlying measurement, suggesting that the reported ROAS was not a true reflection of actual revenue generated.
This discrepancy, while alarmingly high in this specific instance, is symptomatic of a broader, more insidious problem within the digital advertising ecosystem: the misinterpretation and misconfiguration of conversion tracking. When a platform like Google Ads reports an astronomically high ROAS alongside zero sales, it’s a clear indicator that the system is counting something as a "sale" that demonstrably is not. This situation, while glaringly obvious in the student’s case, often manifests in more subtle and dangerous ways in professional advertising accounts, where partially inflated numbers can be mistaken for genuine success, leading to flawed strategic decisions and wasted marketing budgets.
The root cause of such inflated metrics frequently lies in the fundamental settings of conversion tracking. The student’s situation was an easy diagnosis because the misrepresentation was total. The more challenging scenarios involve accounts where advertising efforts do generate actual sales, but the reported figures are only partially accurate. In these instances, a faulty conversion setting doesn’t immediately appear as a problem; instead, it can create the illusion of a banner year, masking underlying inefficiencies.
This issue is distinct from the challenges of blended averages, which can obscure the performance of less successful campaign segments. Instead, the problem discussed here resides one level deeper: the very foundation of measurement itself. Before engaging in complex analyses of attribution models or costly geo-holdout tests, advertisers must first verify whether the numbers they are relying on have been accurately measured at all. In a significant portion of accounts reviewed, a crucial part of the performance data has, in fact, not been properly measured.
The Hidden Impact of Manually Entered Conversion Values
A significant contributor to inflated conversion values, and consequently skewed ROAS figures, is the use of manually entered, static values for certain conversion actions. Every conversion type that relies on a static value incorporates a figure that was manually inputted by a person into a settings field. Google Ads, in its reporting, aggregates these manually assigned values alongside actual revenue generated from completed transactions. The combined total is then used to calculate ROAS. Crucially, the reporting interface does not differentiate between revenue derived from actual sales and values that were simply typed in.
Consider a hypothetical e-commerce account that generates 1,000 orders, totaling $100,000 in revenue, and an additional 2,000 leads each quarter. With a quarterly ad spend of $20,000, and each lead assigned a static value of $10, the reported conversion value would be $120,000, yielding a ROAS of 6.0 ($120,000 / $20,000). Now, imagine that before the start of the next quarter, an advertiser decides that leads are now worth more and updates the static value to $20. This change, however, only applies to conversions recorded after the adjustment. Consequently, the actual sales and lead volume remain consistent with the previous quarter, and the revenue generated from orders is unchanged. Yet, the reported conversion value will now jump to $140,000 ($100,000 from orders + $40,000 from leads), resulting in a reported ROAS of 7.0 ($140,000 / $20,000). This seemingly impressive one-sixth increase in ROAS is not a result of improved campaign performance or increased sales, but solely due to a human modifying a single numerical field in the settings.
The practice of assigning static values to leads is not inherently problematic when it accurately reflects the average revenue generated from a lead-to-sale conversion. However, the issue arises when these static values are altered mid-reporting period, or when different periods are compared without accounting for these changes. This creates a scenario where year-over-year ROAS figures are measured in fundamentally different units, even though the reporting platform may present them as a continuous, linear progression. The chart might still depict a single, unbroken line, masking the fact that the underlying metrics are no longer directly comparable.
The Peril of Overcounting Leads and Duplicated Transactions
Another prevalent issue contributing to inflated conversion numbers is the "Every" conversion setting for lead-related actions. In approximately four out of every five accounts analyzed that utilize leads for bidding strategies, at least one lead conversion action is configured to count every instance rather than a unique occurrence per click. Across all lead actions within the conversion column, this "Every" setting accounts for roughly half of the inflated figures.
While the "Every" setting is appropriate for purchase conversions, as a single individual can indeed make multiple purchases, it is often misapplied to lead generation. Google’s own documentation clarifies that for lead actions, typically only one unique lead per ad click adds value. When the "Every" setting is applied to a lead form, a user who submits the form, realizes there’s a typo in their phone number, and resubmits it, will be counted as two separate leads. Similarly, if a thank-you page loads twice due to a technical glitch, this can also result in two lead conversions being recorded from a single user interaction.
A secondary, yet equally problematic, source of inflated conversion counts stems from the presence of duplicate conversion counters. More than one in every four accounts with purchase tracking exhibits two or more distinct purchase actions within the conversions column. In some cases, these may genuinely track different types of sales. However, more frequently, they are counting the same order multiple times. This duplication can occur through various means, such as one counter being triggered by the Google tag directly, while another is fed through a Google Analytics 4 (GA4) import or a Customer Relationship Management (CRM) upload.
Many standard audits focus on identifying the "wrong" type of duplicate, such as a counter that has drifted significantly from actual sales figures and is clearly broken. These are often harmless because their malfunction is apparent, leading to their removal. The truly dangerous duplicates, however, are those that align almost perfectly with the back-end sales data. These appear healthy and robust, leading to them being overlooked. Such a duplicate can persist as a secondary conversion action for months. The moment it is added to a campaign’s goal, the reported conversion value can effectively double, and the bidding strategy for every campaign utilizing that goal will be impacted. Critically, no alert is triggered because, from the system’s perspective, nothing has technically "broken."
Unused Primary Actions: Ghosts in the Machine
A striking statistic reveals that approximately two out of every three accounts analyzed contain at least one conversion action marked as "primary" for bidding purposes, yet no active campaign goal is currently utilizing it. These actions, while not being bid upon by any active campaign, carry a status that implies their importance. They often escape scrutiny because they are not visible in the primary "conversions" column that advertisers typically review.
The original rationale behind the setup of many of these unused primary actions is frequently lost to memory. However, a single edit to a campaign’s goal settings, or the hasty creation of a custom goal, can inadvertently cause Smart Bidding algorithms to begin optimizing for an event that has not been reviewed or validated in years. Custom goals are a common pathway for these dormant actions to re-enter the bidding process. Google’s own support documentation explicitly states that a secondary action included in a custom goal will be used for bidding once that goal is set within a campaign. Audits rarely uncover instances where these custom goals are thoroughly checked for the inclusion of outdated or irrelevant conversion actions.
The Predatory Nature of "Cheap" Placements in Conversion Columns
Smart Bidding algorithms, designed to optimize ad spend for maximum efficiency, can inadvertently direct budget towards low-value interactions when these are miscategorized as conversions. In approximately six out of every ten accounts, page views, engagement metrics, file downloads, outbound clicks, direction requests, or subsequent video views are present within the conversions column – the very actions that campaign goals are set to bid on.
These "soft" events, along with the duplicated conversion counters, do not fire uniformly. "Cheap" placements, meaning those that are easily triggered and require minimal user engagement, can activate these soft events far more rapidly than a typical search ad generates a genuine sale. Consequently, advertising budget begins to shift, incrementally each week, towards the placements that are most effective at "filling the counter." The reporting system then interprets this shift as improved efficiency and performance.
There is one acknowledged exception to this rule. For very small lead generation accounts, where the volume of actual leads is too low for bidding algorithms to effectively learn, a "softer" event might be used as a proxy. This approach can be effective when it is a deliberate, strategic decision, and the proxy event is assigned a value equivalent to its conversion rate multiplied by the average sale value. However, the majority of soft events encountered in conversion columns are not the result of such calculated decisions.
This inherent susceptibility to miscounting is precisely why setting a Target ROAS on an account that cannot accurately report the revenue of each order is ill-advised. With flawed counting mechanisms, a ROAS target does not safeguard profit margins; instead, it provides the algorithm with precise instructions based on fundamentally incorrect data.
The Imperative of Free Pre-Testing Before Costly Campaigns
PPC Hero recently published a comprehensive guide to incrementality testing, a valuable methodology that rightly focuses on measuring the uplift in a business’s own sales, rather than solely relying on Google Ads metrics. However, the critical step of preparing for these tests, including the strategic decisions around bidding and budget allocation, still often relies on the data found in the conversions column. The paramount importance of correcting this column first cannot be overstated. This foundational cleanup requires minimal investment, typically just an afternoon of dedicated effort.
The process begins with a thorough audit of all conversion actions. Each action must be meticulously reviewed to ensure it accurately reflects a genuine, valuable outcome for the business. This involves verifying that:
- Static values are accurate and consistently applied: If static values are used for leads, they should represent a well-researched estimate of their worth based on historical conversion rates and average order values. Any changes to these values should be documented and considered when comparing performance across different periods.
- Conversion counting is appropriate: The "Every" versus "Unique" setting must be correctly applied based on the nature of the conversion. For lead actions, "Unique" is generally preferred.
- Duplicate counters are identified and eliminated: Any conversion action that duplicates another should be removed to avoid inflating numbers. This includes ensuring that conversion tracking is not being managed through multiple, overlapping systems without proper reconciliation.
- Unused primary actions are de-prioritized: Conversion actions marked as primary but not actively used in campaign goals should be reviewed and either removed or appropriately reconfigured if they are intended to be part of future strategies.
- "Soft" events are excluded from bidding: Page views, file downloads, and other low-intent actions should not be included in the conversions column if the goal is to optimize for actual sales or high-value leads.
What survives this rigorous validation process is the growth that can be reliably compared from year to year. This clean, verifiable data should be the cornerstone of budget discussions and the basis for future incrementality testing, ensuring that investments are made based on genuine performance metrics.
The Bottom Line: Precision in Measurement for Profitable Growth
A Return on Ad Spend (ROAS) is a meaningful measurement only when every component contributing to its calculation is accurately measured. A manually typed-in value represents an estimate from a specific point in time, subject to human judgment and potential revision. A second, duplicate counter adds no new information, merely reiterating the same outcome. Both of these inaccuracies become prohibitively expensive when they reside within the conversion column, the very data source used to dictate crucial advertising budget allocations.
Recalling the student with the seemingly stellar ROAS of over 500% but no actual sales, her account’s error was so egregious that it was impossible to miss. However, many other accounts are not so transparent. They present figures that are alluring and appear to align with desired outcomes, making them tempting to accept at face value.
In the pursuit of accurate performance reporting, a crucial step forward is to separate the reporting of actual order revenue from manually entered lead values. By presenting these as distinct lines on reports, stakeholders can clearly discern which portion of the overall ROAS is derived from tangible sales and which part is based on an estimate. This transparency is vital for informed decision-making and for ensuring that advertising investments are truly driving profitable growth, rather than merely inflating vanity metrics. The future of effective digital advertising hinges on the ability to trust the data that guides its strategy, and that trust begins with unimpeachable measurement.
Basiran
Lina Irawan
