Marketing Metrics in the AI Era: What Changes and What Doesn’t

Marketing metrics in the AI era are not fundamentally different from what they were before, but the pressure to measure more things, faster, with greater confidence has intensified considerably. AI tools now generate dashboards in seconds, surface anomalies automatically, and produce natural-language summaries of campaign performance. What they cannot do is tell you whether you are measuring the right things in the first place.

That distinction matters more than most teams acknowledge. The question is not whether AI makes measurement faster or easier. It does. The question is whether faster and easier measurement of the wrong things gets you anywhere useful.

Why AI Changes the Measurement Environment Without Changing the Fundamentals

I have spent a long time watching the measurement conversation in marketing shift with each new wave of technology. When I was growing an agency from 20 to nearly 100 people, the dashboards got more sophisticated every year. More data sources, more visualisations, more automated reporting. And every time we added complexity, we had to ask the same uncomfortable question: are we measuring what matters, or are we measuring what is easy to measure?

AI does not resolve that question. It amplifies it. When a generative AI tool can produce a performance summary in thirty seconds, the temptation is to treat that summary as insight. It is not. It is pattern recognition applied to whatever data you fed it. If your data is incomplete, biased toward last-click attribution, or missing offline conversion signals, the AI summary will be confident and wrong in equal measure.

Forrester has written thoughtfully about what to consider when automating marketing dashboards, and the core tension they identify is the same one I have seen in practice: automation makes dashboards faster to build and easier to consume, but it does not make the underlying measurement decisions for you. Those decisions still require human judgment about what the business actually needs to know.

The fundamentals of good measurement have not changed. They are: measure outcomes not activity, connect metrics to commercial decisions, understand what your data cannot tell you, and resist the urge to report on everything just because you can. AI makes each of those harder to hold onto, not easier, because the volume and velocity of available data keeps increasing.

The New Risk: Confident Measurement of the Wrong Things

There is a specific failure mode that I have watched become more common as AI tools have embedded themselves into marketing operations. I call it confident irrelevance. The dashboard looks authoritative. The AI-generated summary sounds precise. The numbers are updated in real time. And none of it is connected to a decision that actually matters.

I saw this clearly when I was working with a client in a category with long purchase cycles. Their reporting was immaculate. Engagement rates, scroll depth, video completion, email open rates, click-through rates across every channel. The AI tools they were using surfaced trends weekly and flagged anomalies automatically. The problem was that none of those metrics had any demonstrable relationship to the thing the business cared about, which was new customer acquisition in a category where the consideration window ran to several months.

When we stripped the dashboard back and rebuilt it around metrics that actually connected to commercial outcomes, the team initially felt like they had lost visibility. They had not. They had traded the illusion of visibility for something more honest and more useful.

Forrester puts this well in their piece on the questions you must ask to improve marketing measurement. The most important question is not “what can we measure?” but “what decision does this measurement support?” That sounds obvious. In practice, most marketing teams cannot answer it for the majority of the metrics on their dashboard.

If you are building your measurement framework from scratch or reviewing what you have, the marketing analytics hub on this site covers the strategic and technical layers in detail, from GA4 implementation to modelling approaches and the metrics worth keeping in 2025.

What AI-Powered Measurement Actually Does Well

It would be lazy to frame this as purely cautionary. AI genuinely improves certain parts of the measurement process, and ignoring that does not serve anyone.

Anomaly detection is the clearest example. In a previous role, we had a client running campaigns across twelve markets simultaneously. Spotting a conversion rate drop in one market, caused by a broken checkout flow on mobile, used to take days. By the time it surfaced in a weekly report, the damage was done. AI-driven anomaly detection catches that kind of issue in hours. That is a genuine operational improvement with real commercial value.

Predictive lead scoring is another area where AI adds something substantive. When you are managing large volumes of inbound leads across multiple channels, the ability to score and prioritise based on behavioural signals, rather than just demographic fit, meaningfully improves conversion rates downstream. The AI is not replacing the judgment about what a good customer looks like. It is processing more signals faster than any human team could.

Custom reporting in GA4 is also considerably more accessible with AI assistance. Moz has done useful work on building custom GA4 reports that surface what actually matters for your specific business, rather than defaulting to the standard views. When AI tools help teams build and interpret those custom configurations, the barrier to meaningful measurement drops significantly.

The pattern across all of these is consistent: AI performs well when the measurement objective is already well-defined and the data quality is reasonable. It performs poorly when it is asked to substitute for strategic thinking about what to measure and why.

The Metrics That Survive the AI Era

Some metrics become more important in an AI-saturated environment, not less. Here is my honest read on which ones hold up.

Revenue contribution, attributed honestly. Not last-click revenue. Not platform-reported revenue that double-counts across channels. Revenue contribution measured with the acknowledgment that attribution is always an approximation, and that the approximation should be as honest as you can make it. AI tools will give you attribution models with impressive confidence intervals. Treat those confidence intervals with scepticism until you have validated them against something real.

Pipeline velocity for B2B. How fast are qualified leads moving through the funnel, and where are they stalling? AI tools are genuinely useful here because they can identify patterns across large datasets that humans would miss. But the metric itself is not new. It has always been one of the better proxies for marketing and sales alignment.

Content performance against commercial intent. Semrush has a thorough breakdown of content marketing metrics worth tracking, and the ones that survive scrutiny are consistently the ones tied to intent signals rather than volume metrics. Page views tell you something. Pages that drive qualified traffic that converts tell you something useful.

Email metrics with context. Open rates became less reliable as a standalone metric after Apple’s Mail Privacy Protection changes, but email as a channel is still worth measuring carefully. Crazy Egg’s analysis of email marketing metrics is a reasonable reference point for thinking about which signals still carry weight. Click-to-conversion rates, list health, and revenue per email sent are more defensible than open rates as primary indicators.

Brand search volume as a proxy for awareness. This one is underused. When your brand search volume is growing, it is a reasonable signal that awareness activity is working. It is not perfect, but it is more honest than most brand awareness metrics because it reflects actual consumer behaviour rather than survey responses or ad recall scores.

The Metrics That Lose Value in an AI Environment

Some metrics that were already questionable become actively misleading when AI tools start optimising for them at scale.

Engagement rate is the obvious one. When AI content tools can produce high-engagement content at volume, and when algorithmic distribution increasingly rewards engagement signals, optimising for engagement rate becomes a game that can be won without any corresponding commercial outcome. I have seen agencies report strong engagement numbers to clients for months while the underlying business metrics flatlined. The engagement was real. The value was not.

Time on page and scroll depth follow the same logic. These were always proxies for content quality, not measures of it. When AI-generated content can produce long-form articles that keep readers engaged without actually informing purchase decisions, the proxy breaks down further.

Impressions and reach in paid media deserve particular scepticism in an AI bidding environment. When automated bidding systems are optimising for reach or impressions, you can hit those numbers reliably. Whether the impressions are reaching the right people at the right moment in the buying process is a different question entirely, and one that impression counts cannot answer.

Unbounce has a useful list of content marketing metrics worth tracking, and what I find instructive about it is the emphasis on metrics that connect to conversion rather than consumption. That distinction becomes more important, not less, as AI tools make it easier to generate content and drive surface-level engagement at scale.

How to Build a Measurement Approach That AI Cannot Undermine

Early in my career, I was refused budget for a new website and built it myself instead. That experience shaped something in how I think about tools and resources: the constraint forces clarity about what you actually need. When you cannot rely on a tool to solve the problem, you have to understand the problem well enough to solve it without one.

The same discipline applies to measurement in an AI environment. If you cannot articulate what decision a metric supports, and what you would do differently based on what it shows, you do not need that metric. AI tools will happily track it, report on it, and surface trends in it. That is not a reason to keep it.

A measurement approach that holds up in an AI environment has three characteristics. First, it is anchored to commercial outcomes that the business already cares about, not outcomes that are convenient to measure. Second, it is small enough that every metric on the dashboard has an owner who can explain what it means and what they would do if it moved. Third, it is honest about what it cannot measure, rather than filling gaps with proxies that look authoritative but are not.

The MarketingProfs piece on preparation in web analytics is older but the core argument holds: the failure mode in analytics is almost always strategic, not technical. You can have the most sophisticated measurement infrastructure in your category and still be measuring the wrong things. AI does not change that. It makes it more expensive to ignore.

For teams working through GA4 specifically, Moz’s guide to GA4 custom event tracking is worth working through. The principle of building your event tracking around the actions that actually matter to your business, rather than tracking everything and sorting it out later, is exactly the right approach in a measurement environment where AI will otherwise surface noise as confidently as it surfaces signal.

When I was judging the Effie Awards, the entries that stood out were never the ones with the most sophisticated measurement frameworks. They were the ones where the team could clearly articulate what they were trying to achieve, how they measured whether they achieved it, and what the commercial result was. That clarity is harder to maintain as AI tools add layers of automated insight between the data and the decision. Maintaining it is worth the effort.

There is more on building measurement frameworks that connect to commercial outcomes across the full marketing analytics section of this site, including approaches to attribution, GA4 configuration, and the metrics worth keeping versus the ones worth cutting.

About the Author

Keith Lacy is a marketing strategist and former agency CEO with 20+ years of experience across agency leadership, performance marketing, and commercial strategy. He writes The Marketing Juice to cut through the noise and share what works.

Frequently Asked Questions