The uncomfortable truth: your measurement stack is out of sync with how people actually buy now
Look at the headlines: “Most marketing metrics are misleading.” “Zero-click searches.” “AI decides what your content means and why it gets you wrong.” “5 priorities for lead gen in AI-driven advertising.” “CFOs scrutinize CTV spend; incrementality emerges as a differentiator.”
Underneath all of that is one blunt reality: the way attention, discovery, and decision-making work has changed faster than most measurement stacks.
You’re still shipping reports built for a click-and-cookie world while:
- AI agents and zero-click SERPs answer questions without sending traffic
- Platforms rewrite your titles, descriptions, and even your copy
- Ad systems optimize to their own black-box “success” metrics
- CFOs demand provable incremental impact, not pretty dashboards
The result: smart teams doing dumb things because the numbers say so.
The core problem: optimization gravity vs. business reality
Most teams are stuck in what I’d call optimization gravity:
- Platforms optimize for what they can easily see (clicks, views, leads)
- Analytics tools optimize for what they can easily track (sessions, last-click revenue)
- Boards optimize for what fits in a simple chart (CAC, ROAS, MQLs)
None of those are inherently bad. They’re just incomplete. And in an AI-shaped ecosystem, “incomplete” quickly becomes “misleading.”
A few concrete ways this shows up:
- Zero-click environments distort channel value. Search, social, and marketplaces increasingly answer queries in-feed or in-SERP. Your “traffic” drops, but your influence might not. If you only measure visits, you’ll under-invest in the content that actually shapes demand.
- AI ranking systems misread your intent. Search and social models decide what your content “means” and who should see it. If your measurement assumes your message is received as written, you’re optimizing to a fiction.
- Automated bidding optimizes to platform goals, not yours. “Smart” campaigns push spend into cheap conversions, branded search, or remarketing. ROAS looks great while net new demand quietly erodes.
- Attribution models cling to decaying identifiers. Cookies, device IDs, and deterministic paths are weaker every quarter. Your model is increasingly confident about an increasingly small slice of reality.
The question isn’t “Which single metric should we use instead?” It’s: How do we build a measurement stack that reflects how people actually discover, compare, and decide in 2026?
The new measurement brief: from tracking clicks to modeling influence
For CMOs and performance leaders, the job now is to treat measurement like product strategy, not reporting hygiene.
A modern stack has to do three things well:
- Capture the real shape of the journey (including the invisible parts)
- Separate incremental impact from “would have happened anyway”
- Give operators fast feedback loops without lying to them
1. Rebuild your journey map for an AI-shaped funnel
Most funnels on decks are still linear: awareness → consideration → conversion. Real behavior now looks more like:
- Prompt in an AI assistant
- Zero-click SERP answer
- Reddit thread or TikTok explainer
- Brand site skim (maybe)
- Marketplace or aggregator
- Dark social (Slack, WhatsApp, DMs)
- Return via branded search or direct
You won’t fully track this. But you can design measurement to acknowledge it instead of pretending it doesn’t exist.
Practical moves
- Run “how did you hear about us?” like a product feature. Make it required on high-intent forms. Use free-text, not a dropdown. Tag and code responses weekly. This is your window into dark social, creator influence, and AI/search assistants.
- Instrument “assist” touchpoints, not just landings. Track views and engagements on buying guides, comparison pages, calculators, and long-form explainers, even if they don’t convert directly. These are often what AI agents and zero-click SERPs pull from.
- Map “AI surfaces” where you can influence answers. Think: FAQ schemas, structured data, product feeds, public documentation, review volume. Treat them as media placements, not hygiene.
- Accept that some influence is intentionally untrackable. Stop forcing every tactic into a last-click ROI frame. Instead, decide which layers are “measurable performance” vs. “modeled contribution” and be explicit with finance about the difference.
2. Make incrementality the spine of your media decisions
As CFOs scrutinize CTV, paid social, and AI-optimized campaigns, the question they’re really asking is: “What would happen if we turned this off?”
That’s an incrementality question, not an attribution question.
Practical moves
-
Standardize holdout testing. For any meaningful channel or tactic:
- Define clear test vs. control groups (geo, audience, store clusters, time-based)
- Run tests long enough to smooth weekly volatility
- Measure impact on business outcomes (revenue, margin, LTV), not just clicks
- Use incrementality to police “too good to be true” performance. When a campaign shows absurd ROAS, assume cannibalization until proven otherwise. Branded search, remarketing, and lower-funnel AI campaigns should all earn their budget with tests, not screenshots.
- Build a simple “incrementality score” per channel. Even a rough 1-5 scale (1 = mostly cannibalization, 5 = highly incremental) forces better budget conversations than arguing over attribution models.
- Teach your team to say “we don’t know yet.” For new AI-driven placements or formats, default to test-and-learn instead of rolling them into your standard ROAS narrative.
3. Fix your metric portfolio: fewer vanity numbers, more decision numbers
The problem isn’t that “most marketing metrics are misleading.” It’s that most teams use them for the wrong decisions.
You need a clear separation:
- Health metrics – Are we present and relevant in the category?
- System metrics – Is the machine working as intended?
- Decision metrics – Where should we put the next dollar?
Health metrics (stop pretending these are performance)
- Share of search / branded search volume
- Category-level impression share (where available)
- Social engagement rate and follower growth
- Review volume and rating trends
These tell you if you’re in the game, not if a specific campaign “worked.” Use them to steer brand and category bets, not weekly budget shifts.
System metrics (for operators, not the board)
- Click-through rate, CPC, CPM
- Landing page conversion rate
- Email deliverability and click rates
- Site speed, error rates, broken flows
These are like engine temperature and oil pressure. Crucial for media buyers and growth teams. Dangerous when they become the headline story in executive reviews.
Decision metrics (what should we scale, pause, or test?)
- Incremental revenue / profit per channel or campaign
- Payback period by cohort and channel
- Blended CAC vs. LTV (by segment, not just overall)
- Contribution margin after media and discounts
These are the ones you put in front of finance and the CEO. Everything else is supporting detail.
4. Design your stack for “AI-mangled” content and traffic
AI systems are now:
- Rewriting your titles and snippets (search, social, email)
- Summarizing your pages instead of sending traffic
- Recommending your product in contexts you never see
If your measurement assumes a clean, direct line from your message to their action, you’re modeling a world that no longer exists.
Practical moves
-
Track “representation quality,” not just rankings. For key queries and feeds, audit:
- How AI summaries describe you vs. competitors
- Which benefits and objections show up
- Whether your brand is mentioned at all
Treat this like share of shelf in a supermarket, not a technical SEO footnote.
-
Instrument content for “assist value.” Tag pages and assets as:
- Discover (introduce problem/solution)
- Compare (vs. alternatives)
- Decide (pricing, proof, FAQs)
Then look at their contribution to eventual conversions over 30-90 days, not just same-session outcomes.
- Accept lower direct traffic from high-quality content. If a buying guide’s traffic is flat but your “how did you hear about us?” mentions of that topic spike, the content is working through AI and social surfaces. Don’t kill it because of last-click myopia.
-
Guard against content cannibalization with intent, not superstition. Multiple pages on a topic aren’t bad if they serve different intents. Measure cannibalization by:
- Query overlap and ranking volatility
- Net total impressions and conversions for the topic cluster
If the cluster grows, you’re fine. If you’re just shuffling deck chairs, consolidate.
5. Give operators honest, fast feedback without encouraging bad behavior
You can’t ask media buyers and growth PMs to wait for quarterly incrementality studies. They need daily and weekly signals. The trick is to make those signals honest about what they are.
Practical moves
-
Define “training metrics” for AI-driven campaigns. For example:
- Cost per qualified lead (with a clear qualification rule)
- Cost per high-intent action (e.g., pricing page view, demo request start)
Use these to steer algorithms day-to-day, but always roll performance up into incrementality and payback for big decisions.
- Use directional tests for speed, formal tests for money. Small-budget, short-run experiments can inform creative and audience direction. But don’t use them to justify major reallocations without a proper holdout or geo test.
-
Make “model vs. reality” a recurring agenda item. Once a month, pick 2-3 campaigns and compare:
- Platform-reported conversions
- Analytics-attributed conversions
- Incremental lift where you have it
The goal isn’t to crown a winner; it’s to understand the bias of each lens.
- Reward teams for killing “high-performing” but non-incremental spend. If someone proves that a beloved campaign is mostly capturing existing demand, celebrate the savings, not the lost ROAS.
What this means for how you run your team
The shift from T-shaped to “M-shaped” PPC careers is a symptom of the same thing: you can’t afford channel savants who don’t understand measurement, and you can’t afford analytics teams who don’t understand how AI platforms behave.
For leaders, that means:
- Make measurement literacy a core competency. Not just for analysts. Media buyers, lifecycle marketers, and even creatives should understand incrementality, attribution limits, and how AI systems optimize.
- Align incentives to business outcomes, not channel KPIs. If the paid social team is bonused on in-platform ROAS, they will happily drown you in retargeting and branded search lookalikes.
- Pair operators with analysts on the same OKRs. Put a growth PM and a data lead on a shared goal like “reduce blended CAC by 15% at constant revenue,” not “improve channel X ROAS by 20%.”
- Keep the stack simple enough to explain to the CFO in 5 minutes. If you can’t sketch your measurement approach on a single slide – journeys, tests, and key metrics – it’s too complex to survive the next budget review.
The AI era doesn’t make measurement impossible. It just makes lazy measurement expensive. The teams that win won’t be the ones with the fanciest dashboards; they’ll be the ones whose numbers actually match how people decide.