The Quality Score That Isn't Called a Quality Score Explained

Quick Answer

Meta retired the public-facing Relevance Score in 2019 and replaced it with the three-bucket Ad Relevance Diagnostics. But the underlying quality classifier never went away — it became more sophisticated. Today there's a continuous, internal Ad Quality score that feeds directly into the Total Value formula. You just can't see the number.

The Mechanism Explained

Pre-2019, Meta showed a Relevance Score from 1-10 next to each ad. This was a heavily smoothed version of the internal quality classifier, designed to be intuitive for advertisers. It was retired because:

1. Single-number scores are misleading — they collapsed multiple dimensions
2. Advertisers were optimising for the score itself rather than performance
3. The classifier had become too multidimensional to compress to one number

The replacement (Ad Relevance Diagnostics) shows three percentile buckets: Quality Ranking, Engagement Rate Ranking, Conversion Rate Ranking. These are public-facing summaries of the underlying classifier.

The internal quality classifier in 2026 looks at:

• Engagement bait classifiers (sensational language, withholding information, fake urgency)
• Creative quality (image clarity, video pacing, thumbnail entropy, on-image text density)
• Post-click experience (landing page speed, content alignment, manipulative design like fake X buttons)
• Negative feedback patterns (Hide Ad rate, Report Ad rate, time-to-hide)
• Comment sentiment (NLP classifier on user comments)
• Click quality (bounce rate after click, time on page, return-to-Facebook rate)
• Aspect ratio fitness (does the creative fit the placement natively)
• Brand safety (controversial content, restricted categories)

These get aggregated into a single quality value that enters the auction. Quality enters Total Value additively in the most recent formulation, meaning it's not just a multiplier on bid — it's a separate term that can override an otherwise high bid × eAR product.

Practical Implication

The quality score is the lever you have when you can't bid more aggressively. If your three diagnostics are above average, you've maxed out the public-facing signals. The next level is fixing the inputs the diagnostics are summarising — bounce rate, on-image text, fake urgency claims, landing page speed.

Real Numbers

• The full quality classifier ingests 200+ features per ad
• Quality contribution to Total Value is roughly equivalent to 0.3-0.6 bid units at typical scales
• Improving quality from Average to Above Average reduces CPM by 10-20% in same auctions

FAQs

Q: Can I see the internal quality score?
No — only the three buckets.

Q: Does Meta still calculate a single number internally?
Yes — the buckets are derived from a continuous score.

Q: Why was Relevance Score retired?
Because it collapsed too much information into one number.

Q: Does landing page count as 'quality'?
Yes — post-click experience is a quality input.

Q: What's the fastest way to improve quality?
Replace creative that has below-average diagnostics.

Pix-Vu

Quality classifier inputs include image clarity, on-image text density, and aspect ratio fitness — all things Pix-Vu directly improves by generating clean, properly formatted product imagery. Try it at https://pix-vu.com.