Performance & Reporting
What to expect from Classify contextual campaigns and how to interpret the data you get back.
What to expect
Performance varies by vertical, format, and campaign configuration. The metrics and dimensions below will help you understand what's working and where to optimize. See Best Practices for guidance on campaign setup that maximizes contextual performance.
Ramp-up period — new campaigns may take several days to reach steady-state delivery as DSP optimization algorithms learn from initial impressions. Allow at least a full week before drawing conclusions from performance data.
Available metrics
The Segment Reporting API provides these core metrics:
| Metric | Description |
|---|---|
| Impressions | Total ad impressions delivered against the segment |
| Clicks | Total clicks recorded |
| Spend | Total spend in USD |
| CTR | Click-through rate (clicks / impressions) |
| eCPM | Effective cost per thousand impressions (spend / impressions × 1000) |
These metrics can be broken down by multiple dimensions to understand where performance is coming from.
Report dimensions
When you create a report, you specify which dimensions to break data out by:
| Dimension | Granularity | Best for |
|---|---|---|
date | Daily | Spotting delivery trends, ramp-up patterns, day-of-week effects |
device | Category (desktop, mobile, app, tv, stb, game_console) | Understanding which screens are driving performance |
format | Ad format (banner, video, native, ctv, audio) | Comparing format effectiveness, identifying high-CTR formats |
domain | Individual publisher domain | Finding top-performing publishers, spotting low-quality placements |
Recommended starting point: Request date + domain dimensions. This gives you daily delivery trends and publisher-level performance in a single report — the two most actionable views for optimization decisions.
How to read your reports
Domain breakdown
The domain dimension is the most actionable. Look for:
- High CTR + high volume — these are your best-performing publishers. Consider expanding your segment to find more content like theirs using the Segment Optimization API.
- High volume + low CTR — these domains are consuming budget without delivering engagement. Consider blocking them.
- High CTR + low volume — niche publishers that perform well but don't deliver scale. Worth keeping, but don't expect them to carry the campaign.
Device breakdown
Desktop and mobile typically perform differently:
- Mobile often delivers higher volume but lower CTR
- Desktop tends to have higher CTR but less inventory
- CTV/App environments may behave differently depending on format
If one device category is significantly underperforming, it may be worth adjusting bids rather than excluding it entirely — cutting an entire device type reduces reach dramatically.
Format comparison
Different ad formats have naturally different engagement levels:
- Video and native formats generally achieve higher CTR than standard banner ads
- CTV formats have different engagement patterns than web-based formats
Compare eCPM alongside CTR — a format with lower CTR but lower eCPM may still be cost-effective.
Date trends
Daily data helps you spot:
- Ramp-up patterns — delivery should stabilize after the first few days
- Day-of-week effects — some verticals perform better on weekdays vs. weekends
- Delivery drops — sudden decreases may indicate a deal expiration, DSP budget pacing issue, or inventory availability change
Pixel reporting for deeper analysis
If you've attached the Classify pixel to your creatives, you unlock additional reporting capabilities through the Campaign Reporting API:
URL-level granularity — see exactly which pages your ads served on, not just which domains. This is more precise than DSP-side domain reporting.
Contextual classification overlay — enable include_classification: true in your report request to see the IAB categories, keywords, entities, and sentiment of every page where your ad appeared. This tells you whether the contextual environment matched your intent.
Macro data — if you pass DSP macros through the pixel (inventory domain, deal ID, publisher ID), the report includes expanded macro values. Comparing the macro-reported domain against the actual serving URL is a powerful tool for catching inventory misrepresentation.
Publisher bot reporting — if you registered with the publisher_bot_detection use case, the Publisher Reporting API provides bot traffic breakdown by agent type, including LLM crawlers, search bots, and scrapers.