By Tobin Thomas
Budget season has arrived on shakier ground than usual. Inflation is sticky, consumers are pickier and CFOs are asking a question that many platform dashboards can’t answer: What revenue did this spend actually cause?
Marketers have moved beyond last-click, and most feel good about their measurement. But I find that few can see how their spending actually performs across channels.
Here are my five predictions for how that gap closes in 2027.
1. Platform-Reported Numbers Stop Counting As Evidence
2026 showed how fast a platform can rewrite its own scorecard. Meta removed its seven-day and 28-day view-through windows in January, then redefined what counts as a click in March. Many advertisers saw reported conversions fall 15% to 40% overnight, with no change in real performance.
Yet Winterberry’s September 2026 survey found that 93% of brands rely on platform-native tools, and 84% call them a core part of their measurement stack. More than three-quarters say they’re satisfied with the in-channel incrementality those tools report.
But I don’t think that satisfaction would survive a thorough review. A platform is graded on the metrics it reports about itself, and it will claim credit for a sale whether its spend caused it or not.
My prediction: By the end of 2027, the number marketing reports will be incremental return on ad spend (iROAS), not platform-reported return on ad spend (ROAS).
2. Traffic Stops Being A Proxy For Demand
When an AI Overview appears, organic click-through rate has fallen from roughly 1.76% to 0.61%. On mobile, most Google searches now end without a single click. As platforms take over the mechanics of media buying, an advertiser’s edge shifts to the data and information outside those mechanics.
Teams that read a traffic decline as a demand problem are likely misdiagnosing it. Often the demand didn’t disappear, only the click did.
I foresee sessions and click-through rate (CTR) being demoted from headline key performance indicators (KPIs) to supporting signals. Geo-lift (which measures the incremental impact of marketing by comparing outcomes across geographic areas), holdout tests and causal media mix modeling don’t care whether a customer clicked a blue link or read an AI summary.
These can all better measure whether the spend moved the outcome.
3. Influence Moves To Surfaces With No Click To Count
Morgan Stanley estimates agentic shoppers could account for as much as $385 billion in U.S. e-commerce by 2030. ChatGPT’s self-serve ads platform crossed a $1 billion annualized run rate within months of opening to all U.S. businesses. Social is becoming a purchasing channel too, yet DHL found that 63% of e-commerce businesses sell through social while only 45% of shoppers buy there. People discover on one surface and purchase on another.
According to another Forbes Council piece, the hard question ahead is no longer how to get seen, but how to get believed. Belief is built through creator reviews, comment threads, group chats and AI answers, and most of that never produces a trackable click.
I believe a new channel’s self-reported attribution will soon count for nothing on its own. Every new surface will need to pass a bounded incrementality test before it earns an always-on budget.
4. Unified Measurement Becomes The Operating Standard
Winterberry’s data shows an industry that has outgrown its old methods but not yet its silos. Only a third of enterprise brands still use first- or last-click attribution. Far more use marketing mix modeling (71%), data-driven attribution (62%) and incrementality testing (52%).
The gap is in putting them together. Unified marketing measurement is a leading priority for 83% of brands, but only 9% say they can practice it “extremely well.” Cross-channel optimization was the top measurement priority, chosen by 40% of respondents. That ranked ahead of data aggregation (33%) and incrementality and causality (31%).
They want the methods to agree with each other and produce one answer about where the next dollar should go.
That’s why I see standalone attribution tools losing to systems that calibrate media mix modeling (MMM), incrementality testing and causal attribution against each other. That calibration is what makes the output audit-grade.
5. Measurement Starts To Act, With Humans Setting The Guardrails
Every major platform is automating the buying side. Google is migrating search campaigns into AI Max and plans to bring Dynamic Search Ads into the same framework in 2027. When the platform runs the mechanics, the advantage goes to the team with better inputs and better judgment about what comes back.
Winterberry’s research suggests that gap is still wide. AI is already present in measurement at most brands, but it remains peripheral for many of them. Only 16% say AI plays a central role in measurement, while just 18% believe they have the people and expertise optimally deployed. The result is a paradox: AI is becoming easier to deploy at the same time that interpreting its output remains a scarce capability.
Dashboards tell you what happened and leave a person to decide what to do. Agents that act on a causal model can close that gap. They can reallocate budget, pause spend that isn’t incremental and flag drift—all within limits a human sets.
My prediction is that the strongest teams in 2027 won’t hand measurement or strategy to AI. They’ll give the measurement model the routine reallocation and reporting work, and keep people making the strategic calls, informed by real data.
Finalizing A Plan
Before you finalize your 2027 plan, here are three changes to consider:
1. Re-baseline every platform metric against post-change numbers. Comparing this year’s Meta or Google figures with last year’s isn’t a like-for-like read.
2. Split iROAS from reported ROAS for each channel. The gap between them is where next year’s budget reallocation opportunity lives.
3. Fund one bounded test in a new surface. Test something like ChatGPT Ads, with an incrementality read before you scale it.
The platforms will keep changing their rules, and AI will keep intercepting more of the click. The answer to both is to build your plan on evidence that doesn’t move every time a platform updates its algorithm.
Feature image credit: getty
By Tobin Thomas
Tobin Thomas is the CEO and Co-founder of Lifesight.io, an enterprise-grade agentic unified marketing measurement platform.
