Reading and improving ROAS honestly

ROASROASReturn on Ad Spend (ROAS) measures the revenue generated for every unit of currency spent on advertising, calculated as revenue divided by ad cost.View full definition → should be one of the simpler numbers in marketing. Revenue attributed to a campaign divided by what you spent on it. A ratio a CFO can check in seconds. The problem is that the revenue figure sitting in your ad platform dashboard is almost never the real number, and most marketing teams either know this and stay quiet, or genuinely do not know it and make budget decisions on fiction.

The pressure in 2026 makes this worse, not better. With tighter cost-of-capital environments and CFOs expecting marketing to justify every dollar, inflated ROAS figures that collapse under scrutiny are a reputational risk for any CMO. Here is how to fix the measurement before you try to fix the metric.

Building a ROAS number you can defend

Step 1: Define which ROAS you are actually measuring

There is platform ROAS, blended ROAS, and contribution-margin ROAS. These are three different things. Platform ROAS (what Google Ads or Meta Ads Manager shows you) counts attributed conversions by their own models. Blended ROAS divides total revenue by total ad spend across all paid channelspaid channelsVisitors arriving via paid ads or sponsored placements, where you pay a platform to display your message rather than earning visits organically.View full definition →. Contribution-margin ROAS subtracts cost of goods and variable fulfillment costs before dividing.

Pick one as your primary number and be explicit about which it is. Most teams oscillate between them depending on which looks better that week. That is the root of the credibility problem.

For CMO reporting, contribution-margin ROAS is the only version that maps to business outcomes. A 4x platform ROAS on a product with 20% gross margins and 15% return rates can be a money-losing campaign.

Step 2: Audit your attributionattributionA framework for assigning credit to the touchpoints that contributed to a conversion, so you can measure which channels and interactions actually drive results.View full definition → window and model

Every platform default is set to flatter the platform. Meta, as of mid-2026, still defaults to a 7-day click, 1-day view attribution window. Google uses data-drivendata-drivenAn approach where decisions are systematically informed by data analysis rather than intuition alone.View full definition → attribution, which it controls and cannot be independently audited. These are not neutral settings.

Run a controlled holdout test (also called a geo-based incrementality test) for your highest-spend channel before you trust any attribution number. Katelyn Bourgoin's work on customer psychology aside, the cleanest test in paid mediapaid mediaVisitors arriving via paid ads or sponsored placements, where you pay a platform to display your message rather than earning visits organically.View full definition → remains: turn off spend in one geographic market for two to four weeks, keep everything else identical, and measure the revenue gap. Companies including Airbnb and eBay have published findings showing that a significant share of clicks credited by last-click attribution would have converted anyway. The exact percentages vary by category and brand maturity, but the direction is consistent.

If a full holdout is operationally difficult, a simpler proxy is to compare platform-reported conversions to actual orders in your CRMCRMCustomer Relationship Management: software and strategy to manage and analyse customer interactions throughout their lifecycle.View full definition → or ecommerce back-end for the same period. If Meta claims 3,200 purchases and Shopify shows 2,800 total, you have overlap from multi-touch attributionmulti-touch attributionA method that distributes conversion credit across all marketing touchpoints in the customer journey, rather than crediting only the first or last interaction.View full definition → and you are double-counting.

Step 3: Separate branded from non-branded ROAS

Branded search (someone typing your company name) almost always shows extremely high ROAS because you are capturing intent that already exists. Mixing branded and non-branded numbers inflates your aggregate ROAS and masks poor performance in the campaigns that are actually doing acquisition work.

Pull these apart in your reporting. Branded ROAS and non-branded ROAS should be two separate lines. If you are running Performance Max campaigns on Google, insist on brand exclusion lists, otherwise Google will allocate budget toward branded queries to inflate reported ROAS and you will have limited visibility into it.

Step 4: Build a weekly revenue reconciliation table

Once a week, a single table should sit in your dashboard:

• Platform-reported revenue (per channel, summed)
• Actual revenue from your source of truth (Shopify, Stripe, your ERP)
• The gap, expressed as a percentage
• A short note explaining the gap (attribution window, cross-device, returns not yet processed)

When this table shows a consistent 20-30% overstatement by platforms, that is your correction factor. Apply it when reporting to the CFO. This one habit does more to build marketing's credibility than any brand campaign.

Step 5: Improve actual ROAS, not just reported ROAS

Once you trust the number, moving it is more straightforward. The highest-leverage actions are:

• Cutting spend on campaigns where incrementality tests show lift below your minimum acceptable threshold, even if platform ROAS looks strong
• Raising average order value through bundling or tiered offers, which improves ROAS without touching CPCs
• Tightening audience targeting to reduce wasted impressionsimpressionsThe total number of times an ad or piece of content is displayed, regardless of clicks. Each display counts as one impression, even to the same person.View full definition → on users with low purchase probability, even if this shrinks reachreachThe number of unique people exposed to your message in a given period. Unlike impressions, reach counts each person once, no matter how often they see it.View full definition →
• Improving post-click conversion rateconversion rateThe percentage of visitors or prospects who complete a desired action (purchase, sign-up, contact form), calculated as conversions divided by total opportunities.View full definition → on landing pages, which improves ROAS faster than bid optimization for most mid-sized advertisers

Pitfalls that collapse this work

The most common failure is running incrementality tests once and treating the results as permanent. Incrementality varies by season, by creative fatigue, and by competitive intensity. A test from Q4 2025 tells you little about Q2 2026 performance.

A second failure is letting platform account managers set your attribution settings. They are measured on your spend, not your profitability. Google's recommendations engine will suggest switching to broader match and higher bids in ways that improve Google's revenue from your account. This is not a conspiracy, it is an incentive structure. Treat their recommendations as a starting point for testing, not a default to accept.

Third: do not let ROAS become a target that gets gamed. If you tell a channel manager her KPIKPIKey Performance Indicator, a measurable value that shows how effectively you're achieving a specific objective, tracked over time against a target.View full definition → is a 4x ROAS, she will pause low-ROAS top-of-funnelfunnelThe customer journey from awareness to purchase, typically Awareness, Interest, Consideration, Decision, Action, with prospects narrowing at each stage.View full definition → spend, concentrate on retargetingretargetingShowing ads to users who have previously visited your site or interacted with your brand, to bring them back and drive conversion.View full definition → (which has high attributed ROAS and low incrementality), and your pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.View full definition → will dry up in 90 days.

Quick wins to start this week

• Pull last month's platform-reported conversions and compare them to actual orders in your back-end. Calculate the gap. Write it down.
• Check your Google and Meta attribution window settings. Document what they are currently set to.
• Separate branded and non-branded spend in your next performance report and show both ROAS figures side by side.
• Identify one campaign where you can run a small geo holdout test in the next 30 days.
• Ask your channel managers how they would respond if ROAS targets were replaced with contribution-margin targets. The answers will tell you where the gaming already happens.

ROAS reported honestly will almost always be lower than the number currently on your slides. That gap is uncomfortable, but it is recoverable. A CFO who discovers the inflation themselves is a much harder problem to manage.