With AI-driven search and hyper-fragmented media channels reshaping how people discover brands, the “set it and forget it” approach to marketing measurement is officially dead.
Measuring impact isn’t a static check of dashboard data. Used strategically, measurement is a virtuous cycle where data informs your ad platform settings and those settings, in turn, generate better data (and business outcomes).
Here’s how to build a measurement flywheel that keeps your growth efficient.
The 4-step measurement cycle
Imagine a Bay Area SaaS company, PowerLoop, selling an AI-powered analytics platform. They’re investing heavily in Google Search, LinkedIn, and some emerging AI publication sponsorships.
Their problem? Google Ads is reporting fantastic ROAS, but their internal CRM shows a significant number of leads and opportunities that can’t be directly attributed to any specific ad campaign, making it hard to prove marketing’s true impact to the board.
1. Platform ROAS
This is your in-engine reality. Whether it’s Google Ads or Meta, platform ROAS uses pixel and conversion API data to tell you what the platform thinks happened. This might go without saying, but platforms don’t have a habit of underestimating their own impact.
The ideal: Use this for real-time optimization.
The limitation: These signals feed your tCPA (target cost per acquisition) or tROAS (target return on ad spend) bidding strategies. It’s the fastest feedback loop you have, but it’s rarely the full truth. This leads us to…
What it looks like in practice (example): PowerLoop’s Google Ads account is configured with a tCPA bid strategy for “Free trial sign-ups.”
Google Ads reports a healthy $50 CPA, well within their target. LinkedIn also shows strong engagement and click-through rates. This looks great on paper, but the unattributed leads are a nagging concern.
Dig deeper: How to avoid marketing mix modeling mistakes that derail results
2. Back-end ROAS
Platform data is optimistic. Your bank account is realistic.
Back-end ROAS, coming from your CRM of choice (Salesforce, Shopify, HubSpot, etc.), connects your ad spend to your actual CRM or internal database. It’ll likely require some data engineering work to properly map back-end performance against ad platform spend, but the effort is well worth it.
The ideal: Clean out the “noise” (refunds, fake leads, or credit card declines), and evaluate marketing efficiency based on your own first-party data.
The benefit: You can use back-end ROAS to validate your account structure. If the platform says a campaign is winning but the back end shows low-quality leads, it’s time to restructure your targeting or creative.
What it looks like in practice (example): When PowerLoop connects their ad spend to Salesforce, they find that many of the “Free trial sign-ups” from Google Ads are either incomplete profiles or come from IP addresses outside their target market and never convert to qualified sales opportunities.
LinkedIn, while showing engagement, has a lower conversion rate than expected. This insight leads them to refine their Google Ads audience targeting and adjust LinkedIn campaign objectives to focus more on high-intent lead forms.
Get the newsletter search marketers rely on.
See terms.
3. Incremental ROAS (iROAS)
This is the “So what?” metric. iROAS answers the question: How many of these sales would have happened even if we didn’t show the ad? This is where marketing mix modeling (MMM) and incrementality testing (geo-lift tests or holdout tests) come into play.
The goal: Identify true value and “halo effects” across channels.
The action: MMM insights tell you where to double down and where you’re just paying for customers who would have converted anyway. Use these insights to prioritize your next round of incrementality tests.
What it looks like in practice (example): PowerLoop conducts a geo-lift test by pausing Google Ads in select non-core markets for a few weeks and measuring the difference in sign-ups between dark areas and similar areas where ads are still running. They discover that while Google Ads drives some incremental sign-ups, a significant portion of those attributed by Google would have signed up organically anyway, through direct traffic or referrals.
Conversely, their MMM suggests that the AI publication sponsorships, while not driving direct “last-click” conversions, are significantly contributing to brand awareness and reducing the overall CPA across all digital channels by driving more organic searches for their brand. This reveals that the sponsorships have a higher iROAS than initially thought.
Here’s an example of overvalued and undervalued channels:
The greater the incrementality factor, the more undervalued this channel has been, such as YouTube and podcasts in this example. The lower the incrementality factor, the more overvalued these channels have been, such as paid review sites in this case.
Dig deeper: Why incrementality is the only metric that proves marketing’s real impact
4. Marginal ROAS (mROAS)
The final frontier is understanding where to spend the next dollar. Every channel eventually hits a plateau where efficiency craters. This truism is called the law of diminishing returns. Understanding when you hit that mark is key to efficient budgeting.
The goal: Estimate the “room for growth” before hitting a performance ceiling.
The benefit: By monitoring mROAS, you know when to pull back on a saturated channel and reallocate that budget into emerging spaces.
What it looks like in practice (example): PowerLoop’s analysis shows that after spending $100,000/month on Google Ads, another $10,000 yields a marginal return of $0.80 for every dollar spent – meaning they’re essentially breaking even or losing money on additional spend.
However, for their AI publication sponsorships, every additional dollar spent is still returning $2.50 in incremental value, indicating significant room for growth. They decide to reallocate 15% of their Google Ads budget to expand their sponsorship program.
Why the cycle never ends
Marketing measurement is a work in progress because the landscape is constantly shifting. Today, you might be perfecting your Google Search strategy. Tomorrow, you’re figuring out how to measure the impact of a mention in a ChatGPT or Perplexity response.
The hypothetical PowerLoop team understands this. They’re constantly evaluating new AI-driven channels and planning how to integrate them into their measurement cycle. They know that what worked last quarter might not work this quarter and that relying solely on platform data is a recipe for wasted spend.
The goal isn’t to find a “perfect” number that stays set in stone. The goal is to use this cycle to stay agile. When your iROAS reveals that a channel is more incremental than you thought, you push your tROAS targets in the platform (Step 1) more aggressively. When mROAS shows you’re hitting a plateau, you start testing new, unproven channels to find different audiences.
Dig deeper: Break down data silos: How integrated analytics reveals marketing impact