How to Use AI to improve Marketing Metrics

On a recent episode of the Authentic Business Adventures podcast, host James Kademan welcomed Jeff Greenfield, CEO and co-founder of Provalytics, to unpack one of marketing’s most frustrating and misunderstood challenges: attribution.

Jeff, a 30-year marketing innovator and pioneer in privacy-first measurement, shared why so many brands are flying blind when it comes to knowing what truly drives their sales—and how platforms like Google Analytics may be part of the problem. “You’re not measuring reality,” Jeff explained. “You’re measuring what Google wants you to see.”

🚀 Yes, I Want The 2026 Playbook!

The Broken Model Behind Most Marketing Decisions

Most brands still use last-click attribution—giving all the credit for a sale to the final touchpoint before conversion. Jeff explained why this creates a distorted view of performance.

Take a connected TV (CTV) ad. A viewer sees it during a game, thinks about the brand, and searches for it days later—then clicks on a paid Google ad. Google Analytics credits that last click, while ignoring the influence of the CTV ad that created the demand in the first place.

That’s how entire budgets shift toward platforms that close the sale rather than those that create the sale.

Why Today’s Data Is More Confusing Than Ever

Modern marketers are managing dozens of channels across walled-garden ecosystems like Facebook, Amazon, YouTube, and CTV. These platforms run on real-time auctions—selling every ad impression to the highest bidder in milliseconds.

If you’re not measuring correctly, your costs go up while performance stays flat—or worse.

To make matters more complex, internal marketing teams often operate in silos, with TV, digital, and PR teams optimizing in isolation. Jeff noted that when Provalytics audits a company’s media performance, it’s not unusual for platforms to collectively report 4,500 sales on a day where only 1,000 sales actually happened.

The Provalytics Approach: Math Meets Machine Learning

Jeff shared how Provalytics was built to solve this attribution crisis with a privacy-first, cookie-less, tag-less platform that combines classical statistics with modern machine learning.

At the core of the platform is Bayesian modeling, a 300-year-old method now enhanced by today’s AI and computing power. It works like human intuition—taking new information and constantly refining predictions. No user-level tracking. No PII. Just smarter models that deliver clarity without compromising privacy.

Why Provalytics Starts at the Top

One of Jeff’s key points: You can’t sell better measurement to someone who already looks great on paper. That’s why Provalytics sells top-down—working directly with CMOs and CFOs who feel the pain of rising costs, siloed teams, and upper-funnel investments they can’t justify to Finance.

By providing a unified, cross-channel view of what’s actually driving results—across CTV, digital, social, and beyond—Provalytics empowers leaders to reallocate spend with confidence.

Advice for Smaller Brands: Start With Impressions

While Provalytics typically works with companies spending $10M+ annually on media, Jeff shared a powerful DIY approach for smaller marketers.

The takeaway? Focus less on clicks—and more on impressions. Why? Because impressions drive awareness. Awareness drives clicks. And clicks drive sales.

Jeff’s recommendation:

• Pull your last 12 months of marketing data into a spreadsheet.
• Add a column for daily impressions.
• Graph impressions, clicks, and sales over time.
• Look for delayed cause/effect patterns.
• Identify what drove attention—not just what closed the deal.

“Clicks are the result,” Jeff said. “The driver is attention. That’s what you need to measure.”

The Future of Attribution Is Here—And It’s Privacy-First

Marketers don’t need perfect data. They need useful data. At Provalytics, we believe attribution should reflect how people actually behave—not just how platforms report it.

We help enterprise brands see the full picture, so they can grow smarter, spend more efficiently, and finally give credit where credit is due.