Repeated titles can conceal meaningful product differences. That is a reason to inspect the catalog, not evidence that Google assigns a universal “uniqueness score.”
The original feed review described six hundred products whose titles began with the same brand and category wording. A buyer could struggle to see which size, material, use case or feature made one item suitable.
The useful revision is product-specific and factual. Keep the supported brand and product identity, then add the attributes that actually distinguish the item. Do not invent a benefit merely to make each title different.
The former 35% and 98.7% figures were an internal text-difference heuristic, without a published method or a controlled performance comparison. They are not Google quality measurements, and a higher heuristic score cannot establish revenue lift.
Inspect a sample of titles beside their product pages. Which differences are real? Which are already visible? Which facts are missing or inconsistent? Preserve valid item and variant identity when editing the feed.
Use a bounded title revision with a retained baseline to examine eligible impressions, relevant queries, clicks and contribution. More text variation is useful only when it improves the buyer's understanding and the observed account decision.
Take the idea further
Audit the title assumptions