Est. reading time: 8 minutes
The most expensive mistakes in Google Ads don’t happen inside the ad account. They happen in the meeting where someone looks at a report, draws the wrong conclusion, and makes a budget decision based on it. We’ve sat in those meetings. A business owner divides total spend by the revenue they can directly trace to Google Ads, decides the channel isn’t working, cuts the budget, and six weeks later revenue declines across the board with nobody connecting it back, because the effects are indirect and delayed. Or the opposite: in-platform ROAS looks incredible, budget increases, but the ROAS was inflated by brand campaigns intercepting organic traffic, and the incremental revenue from the added spend is a fraction of what the dashboard suggested.
Both scenarios play out constantly, and the problem isn’t the dashboard. It’s interpreting the data without understanding what it represents, what it leaves out, and where the numbers lie if taken at face value. Here’s where they lie.
Brand campaigns distort everything
This is the single biggest source of misread Google Ads performance, and almost every account we audit has it. Brand campaigns bid on your own company name: someone types it, your ad appears at the top, they click, they convert, and the dashboard attributes that conversion to the campaign at a very low CPC and a spectacular ROAS. The problem is that most of those people were going to find you anyway. They knew your name, they were navigating to your site, and without the ad the vast majority would have clicked the organic result directly below it. The brand campaign didn’t generate that demand, it intercepted it, and then took credit.
Brand campaigns aren’t worthless. Protecting your name from competitors bidding on it, controlling the top-position messaging, and holding the slot where competitor ads would otherwise sit above your organic result are valid defensive plays. But blending brand and non-brand performance into one report makes the numbers meaningless, because brand will always post lower CPA and higher ROAS, it converts people already looking for you, and the non-brand campaigns actually generating new demand will always look worse beside it.
The businesses that decide well separate the two completely: brand reported alone, evaluated as a defensive cost, held at the minimum budget that maintains coverage; non-brand reported alone, evaluated as the true measure of whether Google Ads is generating new business. We’ve seen an account with a blended 8x ROAS and a delighted leadership team, sitting on top of a non-brand ROAS of 1.2x. The account wasn’t driving profitable growth, it was intercepting branded searches at a great reported return while the growth campaigns barely broke even, and without the split view, that problem was invisible.
Last-click attribution lies in a specific way
Google Ads defaults to crediting the last ad click before conversion, which creates a systematic bias worth understanding. Paid search often captures demand at the bottom of the funnel: someone sees a Meta ad, visits, leaves, thinks for a few days, searches the category on Google, clicks your search ad, buys. Google’s report gives that conversion entirely to the search click. Meta gets nothing, Google’s ROAS looks fantastic, Meta’s looks weak, and neither number describes what happened, because Meta created the interest and Google converted it, valuable work on both ends that spans both channels.
This drives real misallocation. We’ve worked with businesses that shifted budget from Meta to Google on the strength of platform-reported ROAS comparisons, watched Google hold strong for a few weeks, and then watched it decline as the top-of-funnel demand Meta had been generating dried up, fewer people entering the funnel, fewer branded and category searches for Google to capture.
The fix isn’t abandoning platform attribution, it’s supplementing it. Blended ROAS, total revenue over total ad spend across every channel, says whether the overall investment returns acceptably regardless of who claims credit, the role we’ve assigned to MER as the ratio that ends attribution arguments. New customer acquisition cost matters more than overall CPA, since efficiency at converting existing demand is worth less than it appears, so segment conversions by new versus returning wherever possible. And incrementality questions keep everyone honest: the key question is never “what does this channel claim” but “what revenue would we lose if we turned it off,” approximated with controlled spend-down tests, reduce a campaign’s budget for two to four weeks, measure total business revenue rather than attributed revenue, and compare. The gap between the dashboard and the business-level impact is your reality check.
The campaign types tell different stories
Applying one ROAS or CPA target across campaign types is among the most common evaluation mistakes we see, because the types do different jobs. High-intent search, “buy [product],” “[service] near me,” is your closest-to-revenue work and should carry your strictest efficiency targets, and when it underperforms, the issue is usually landing page quality, offer competitiveness, or relevance rather than the channel. Research-stage search, “best CRM for small business,” sits earlier in the decision and will always post higher CPA and lower ROAS, because the searcher needs more touchpoints before converting, so it gets evaluated on pipeline contribution instead, assisted conversions, new users introduced who later convert through other campaigns. A research click that leads to a branded conversion two weeks later contributed real value that last-click misses entirely.
Performance Max earns its own scrutiny. It runs across Search, Display, YouTube, Gmail, Maps, and Discover simultaneously, and its evaluation challenge is transparency, since Google offers limited visibility into which channels drive the results. A PMax campaign can report strong ROAS while most of its conversions are branded queries that would have converted anyway. We run it with cautious optimism, alongside standard search rather than replacing it, watching the Insights view closely for branded cannibalization, the same discipline we detailed in structuring Performance Max the right way. And Display and YouTube are awareness channels, not direct response, so judging them on last-click ROAS will always disappoint, because creating demand for other channels to convert is the job. View-through activity, audience building, and reach are the appropriate metrics, and if leadership expects Display to match branded search ROAS, the problem is the expectation.
Spend efficiency vs spend level: the scaling trap
A subtler mistake, and an expensive one. Every account has a natural efficiency ceiling at a given spend level, because the highest-intent, lowest-competition queries get captured first, and increasing budget expands into more competitive auctions and progressively lower intent. CPA rising as spend rises isn’t the campaign failing, it’s the economics of scaling any auction channel. The mistake is seeing a strong CPA at $10,000 a month, tripling to $30,000 expecting the same number, panicking at a 40 percent CPA rise, and slashing back or restructuring, when $30,000 simply buys a different point on the curve than $10,000 did.
The right question isn’t whether CPA held, it’s whether the incremental conversions are profitable. If the target is $50, the old budget delivered $30, and the new one delivers $45, the campaign is performing within target, more customers at a slightly higher cost each, usually a good trade. We model the curve before recommending increases: the first increment captures the cheapest conversions, each additional one costs more, and the budget should stop where marginal CPA crosses the profitability threshold, not where CPA matches a number it was never going to match at scale. Presented that way, “here’s what the next $5,000 likely produces, at this expected CPA, and here’s whether that’s profitable on your unit economics,” the increase-panic-slash cycle never starts.
What good evaluation actually looks like
The businesses that consistently decide well share a few practices. Brand and non-brand separated in every report, so blended metrics never obscure whether growth campaigns generate profitable new business. Campaign types held to appropriate benchmarks, strict efficiency for high-intent search, pipeline contribution for research-stage, awareness metrics for Display and YouTube, cannibalization scrutiny for Performance Max. Platform attribution supplemented with blended business-level metrics and periodic incrementality checks through controlled spend changes. And the spend-level curve understood, with budget decisions made on whether incremental conversions are profitable rather than on whether CPA matched an unscalable memory.
None of it requires sophisticated modeling or expensive tools. It requires separating the data correctly, asking the right questions, and resisting the temptation to decide based on whichever number feels best or worst on a given day. The biggest risk in Google Ads isn’t a poorly structured campaign. It’s a well-structured campaign killed or starved because someone read the data wrong, and getting the evaluation right protects the investment better than any bid adjustment ever will.
