The new challenge for marketers: too much data, too little truth

What “quality” looks like online and why it misleads offline

Programmatic buyers have gotten really good at assessing site-level quality. Open Sincera provides granular data on things like ad-to-content ratio, refresh rates, page weight, and user experience metrics. We used these signals to create a composite quality score for each site. These indicators help minimize fraud and ensure decent user experiences.

But they don’t guarantee sales performance.

When we compared sites with high quality scores against actual retail sales lift, we found major gaps:

• WSJ Video (premium contextual news): Quality score 90+, sales lift +25%
• Integral-Calculator.com (utility site): Quality score 88, sales lift ≈0%
• Free-Spider-Solitaire.com (casual gaming): Quality score 85, sales lift –8% ROI
• Curated High-Quality Average: Quality score 80-92, sales lift +6 to 12%

Bottom line: technical quality doesn’t equal commercial performance.

The experiment: what happens when you connect real retail data to media

We ran three controlled tests with a national beverage brand, a snack manufacturer, and a leading candy brand. Each test compared sites with high quality scores (based on Sincera’s metrics) against sites identified as historically high-performing through Evidnt’s retail sales database.

The difference was striking.

Technically strong sites achieved good engagement and viewability, but they didn’t correlate with incremental in-store or e-commerce sales. Premium, context-rich environments like The Wall Street Journal Video consistently outperformed utility or gaming sites.

The results:

• +25% incremental lift for WSJ Video placements
• Negative ROI for free-game inventory
• 6–12% incremental lift when optimizing using Evidnt’s sales-verified curated deals

How we measure what actually sells

Our measurement framework combines econometric rigor with real-time retail data, allowing brands to isolate incremental impact from media spend across every channel, placement, and audience. (Full methodology details at evidnt.co/brands/impact/mmm.)

Here’s how it works:

Step 1 – Define test and control
We create test groups (audiences, geographies, or time windows exposed to specific placements) and control groups (matched audiences or regions withheld from those placements). Groups take into consideration demographic as well as buying behaviors at the market level.

Step 2 – Collect retail truth data
We ingest daily POS transactions from 28,000 retail partners covering independent, on-premise, and e-commerce channels. Each sale includes product UPC, retailer ID, timestamp, and ZIP code for precise geo-matched modeling.

Step 3 – Normalize and detrend
We adjust for seasonality, promotions, competitive activity, and macro factors like price, weather, and distribution to produce a baseline forecast for what sales would have been without media exposure.

Step 4 – Calculate incrementality
Using Bayesian hierarchical regression and causal-impact modeling, we determine statistical significance and confidence intervals for the lift.

Step 5 – Integrate into MMM

The incremental outputs feed into our ongoing MMM system, quantifying ROI by channel, creative, and context. Through API integrations with our supply partners—Index Exchange, Nexxen, and Equativ—we score every site and placement with an Evidnt performance score based on verified sales data. This allows dynamic reallocation of spend toward the placements generating the highest marginal sales return.

This bridges short-term lift testing with long-term optimization, a major advantage over isolated campaign analytics.

The real issue isn’t fake data…it’s AI-generated noise

Over the past 18 months, AI content creation has exploded. Thousands of algorithmically generated sites, videos, and articles flood programmatic exchanges daily.

Most aren’t fraudulent, they’re just irrelevant. They don’t engage or impact users, and neither do the ads around them.

They look fine to automated systems: clean layout, decent engagement, safe adjacency. But human audiences don’t spend real time or attention there, and the correlation with purchasing behavior is minimal.

This creates a new kind of inefficiency: too much technically “safe” inventory, not enough authentic human attention. Without retail validation, optimization engines over-index on this low-value inventory, chasing engagement that never converts.

So why are we leaning into Curated Deals?

Evidnt’s approach is straightforward: buy media where sales actually happen.

Through API integrations with Index Exchange, Nexxen, and Equativ, we deploy curated deal IDs where every site and placement receives an Evidnt score based on historical sales performance. These deals:

• Filter out low-context and AI-generated content
• Prioritize inventory with proven sales lift across our retail database
• Continuously optimize in-flight using real-time retail data feedback loops

Every impression within these curated deals is evaluated not only for brand safety and traditional quality metrics, but also for sales potential based on actual transaction data. For brands, this means every dollar spent goes toward placements that have proven to move product.

Results across categories

The numbers speak for themselves:

Retail-validated media consistently outperforms traditional buying by 6–12 percentage points.

The path forward: from awareness to accountability

The volume of AI-generated content will only grow, and so will the noise. To maintain efficiency, marketers need to evolve from probabilistic to proven media optimization.

Evidnt helps brands:

• Measure the true incremental effect of every campaign
• Discover which publishers, contexts, and creative types actually sell
• Activate curated media through trusted SSP partners to scale those results

Brand safety and quality still matter, but they’re not the destination. Authenticity and outcome are.

Redefining what “good media” means

In an era where AI can generate infinite impressions, success depends on connecting exposure to real economic activity.

Evidnt’s methodology and retail-based measurement give marketers a single source of truth: proof that advertising not only reached someone, but changed what they bought. Through our supply partner integrations, we translate that proof into action—scoring every available placement and steering spend toward inventory that drives real sales.

Because at the end of the day, the most powerful metric isn’t reach. It’s return.

Contact info@evidnt.co to learn more about our Curated Deals.