The pattern nobody wants to admit
Scan those headlines and a theme jumps out: everyone is quietly admitting that the way we measure marketing is broken.
“Most marketing metrics are misleading.” “How high-growth companies actually measure marketing.” “Your owned content is losing to a stranger’s Reddit comment.” “AI’s trust problem.” “Why your SEO vendor can’t build on a broken foundation.”
At the same time, AI is rewriting how media is created, distributed, and discovered: AI agents managing search, AI creative, AI video editing, AI prospecting, AI everywhere. The machine is doing more of the work, but the humans are still staring at the same shallow dashboards.
The real issue for CMOs, performance leaders, and media buyers right now is simple:
Your measurement stack was built for a world of channels and campaigns. You now live in a world of agents, intent gaps, and content you don’t control.
If you keep optimising to the old metrics, AI will happily help you go faster in the wrong direction.
Three big shifts that make your current metrics mostly noise
1. AI is becoming the interface, not the channel
Google’s CEO is openly talking about search as an “AI agent manager.” Social platforms are rewriting visibility rules. Streaming platforms are building podcast strategies to feed algorithmic discovery. AI is increasingly the gatekeeper between your brand and the customer.
That breaks a lot of the assumptions behind:
- Last-click or even multi-touch attribution that assumes linear journeys
- Channel-based ROI that treats “search,” “social,” and “display” as stable buckets
- Brand vs performance budget splits that map to media lines, not customer realities
In an agent-driven world, your “channel” is whatever the agent decides is the best next action for the user. That might be your site, or a Reddit thread, or a YouTube review, or a podcast mention you didn’t buy.
2. Your owned content is no longer the main character
When Search Engine Journal is publishing pieces like “Your owned content is losing to a stranger’s Reddit comment,” that’s not a cute SEO anecdote. It’s the new normal.
People trust:
- Forum threads
- Subreddit FAQs
- Creator reviews
- Support docs and GitHub issues
more than your polished landing page. AI systems are trained on those signals too.
Yet most marketing reporting still:
- Measures only what happens on owned properties
- Credits only the clicks you can tag
- Ignores the “dark funnel” of research, recommendation, and validation
So you over-invest in the content you can track and under-invest in the content that actually convinces.
3. AI is amplifying bad measurement habits
Tools are shipping “AI prompts for better ad campaigns,” “AI creative in 2026,” and “AI video editing” while copywriting blogs warn about “AI’s trust problem” and broken email experiences.
If your KPI stack is shallow, AI will happily:
- Generate 100 variants that all optimise to the same junk metric
- Overfit creative to cheap clicks or low-quality leads
- Scale broken experiences faster (see: 73% of ecommerce emails being broken)
AI is not neutral. It optimises for what you tell it to value. If your metrics are wrong, your AI is wrong at scale.
What high-growth operators are actually measuring
The interesting thread in the “how high-growth companies measure marketing” and “most metrics are misleading” pieces is not that they use exotic formulas. It’s that they:
- Measure fewer things
- Measure closer to cash
- Measure at the level of customer reality, not channel structure
The pattern across B2B SaaS, ecommerce, and consumer brands that are actually compounding right now looks something like this.
1. A spine of three commercial metrics
Every serious operator should have a simple, non-negotiable spine:
- Payback period on marketing spend (by major motion, not micro-campaign)
- Incremental revenue (or margin) vs a holdout or baseline
- Customer quality (LTV bands, retention, expansion, or repeat rate)
These are not “brand” vs “performance” metrics. They are business metrics. Everything else is diagnostic.
2. Intent-based, not channel-based views
Search Engine Land is publishing on “measuring intent gaps.” That’s the right direction. The unit of analysis needs to move from “channel” to “intent cluster.”
For example:
- “Problem aware, no solution in mind” (early research)
- “Comparing options” (vs competitors)
- “Ready to buy, choosing vendor”
- “Existing customer, expanding or churning”
Across each intent cluster, you should be able to answer:
- What surfaces first for the user? (Search results, AI answers, social, marketplaces)
- How often are we present in that moment?
- When we are present, how often do we win the next step?
That’s a very different dashboard from “ROAS by channel.”
3. Experience metrics, not just exposure metrics
The “73% of your ecommerce emails are broken” headline is the quiet horror story behind a lot of “our CRM program isn’t working” narratives.
Most teams still:
- Count sends, opens, clicks
- Ignore rendering, deliverability, and friction
- Never connect experience quality to revenue impact
High-functioning teams add:
- Experience health scores (e.g., % of sessions with errors, broken flows, or slow pages)
- Message integrity (e.g., % of templates that render correctly across devices)
- Time-to-value (how long from first touch to first meaningful outcome)
Then they correlate those to conversion and retention, and they stop arguing about whether “UX is marketing” because the data answers it.
The new measurement stack: what to build in the next 12 months
If you are a CMO, VP Growth, or head of media buying, your job this year is not to “add AI” to your stack. It’s to rebuild the measurement stack so AI has something sane to optimise against.
Here’s a practical blueprint.
1. Define your “agent reality” map
Assume that for every important intent, an AI system or algorithmic feed is now the first filter. Map:
- The top 20-50 intents that matter to your business (use search data, sales calls, support tickets)
- For each intent, what surfaces first today:
- Classic search results
- AI answers (search overviews, chatbots, agents)
- Social feeds and creator content
- Marketplaces and app stores
- Forums, Reddit, Discord, niche communities
- Where you are absent, or present but losing (e.g., your owned page is outranked by a third-party review)
This becomes your “agent reality” baseline. Review it quarterly, not annually. Your media and content plans should attach to this map, not a channel spreadsheet from 2018.
2. Move from campaign attribution to motion attribution
Most teams still report at the granularity the ad platforms hand them. That’s convenient, and wrong.
Instead, define 4-7 core motions that matter:
- New customer acquisition
- Reactivation of lapsed customers
- Expansion / upsell
- New market or segment entry
- Brand preference shift vs a specific competitor
For each motion, track:
- Spend (all-in, across channels and content)
- Incremental revenue and margin vs holdout or baseline
- Payback period and LTV/CAC
- Mix of surfaces that actually touched the user (owned, earned, paid, AI answers, third-party content)
Then let campaigns and channels roll up into motions, not the other way around. This is how high-growth companies avoid getting trapped in “turn off this one campaign, ROAS drops, panic” loops.
3. Instrument the “dark funnel” with directional, not perfect, data
You will never perfectly track Reddit, podcasts, DMs, or Slack communities. You don’t need to. You need consistent directional signals.
Practical moves:
- Add “Where did you hear about us?” as a required, open-text field on key forms. Code the responses monthly into buckets.
- Run quarterly “path to purchase” surveys with recent buyers. Ask about:
- First awareness
- Key validation moments
- Content or communities that mattered most
- Track share-of-voice and sentiment on the 3-5 communities that actually matter in your category.
Then, marry this with your motion-level metrics. You’ll see which “untrackable” surfaces correlate with better payback and higher LTV, and you can justify investing there without fake precision.
4. Make AI work for measurement, not just media
Everyone is using AI to write more ads. Fewer are using it to make sense of the chaos.
Use AI to:
- Cluster search queries and content topics into intent groups automatically
- Summarise thousands of “how did you hear about us” responses into a clean taxonomy
- Detect patterns in broken experiences (e.g., parse support tickets and session recordings for recurring friction)
- Generate hypotheses about cannibalisation (e.g., when new campaigns steal from organic or brand demand)
But be strict: AI can propose patterns and hypotheses; humans decide what becomes a KPI.
5. Tie creative and content to business outcomes, not vanity metrics
With AI creative, AI video editing, and 10 years of PPC tests showing that “best practices” can be traps, you need a better way to judge creative than “CTR went up.”
A pragmatic framework:
- Define 3-5 creative territories (not 50 random variants)
- For each territory, track:
- Down-funnel impact (qualified leads, purchases, LTV bands)
- Cross-channel performance (does it travel across search, social, email, site?)
- Longer-term effects (brand search volume, direct traffic, win rates in sales)
- Use AI to generate volume within the winning territories, not to constantly chase new shiny objects
This turns creative from a slot machine into a portfolio with clear business returns.
What to stop measuring (or at least, stop worshipping)
To make room for a better stack, you have to kill some dashboard darlings.
- Channel-level ROAS as the north star. Useful as a diagnostic, dangerous as a goal. It ignores incrementality, cannibalisation, and customer quality.
- Impressions and reach as success metrics. They are inputs, not outcomes. Treat them like spend: necessary, not sufficient.
- Engagement for its own sake. “Replying to comments boosts engagement” is nice, but if that doesn’t show up in retention, NPS, or revenue, it’s a hobby, not a strategy.
- “Brand awareness” studies with no link to future cash flows. If your brand tracker can’t predict future demand, it’s a mood board, not a metric.
The uncomfortable leadership job
The hardest part of this shift is not the tooling. It’s the politics.
As a senior leader, you will have to:
- Tell teams to stop optimising for metrics they’ve been praised for hitting
- Explain to finance why short-term ROAS might dip while payback and LTV improve
- Admit that some of your past “wins” were just channel arbitrage, not durable growth
- Resist the temptation to use AI as a volume button instead of a clarity tool
But the alternative is worse: an AI-accelerated version of what you already have-more campaigns, more creative, more content, more dashboards-sitting on a broken measurement foundation.
The operators who win the next five years will not be the ones with the fanciest prompts or the biggest model budgets. They’ll be the ones who did the boring, sharp work of deciding what actually counts, and then made the machines optimise to that.