Most eBay sellers treat Cassini as a black box. It isn't — the ranking algorithm is documented across a decade of eBay developer conference talks, patent filings, and reverse-engineered signal analysis. What makes Cassini feel mysterious is that it doesn't score listings in isolation. It scores your entire seller cluster against category benchmarks in real time, and a single dropped signal — one listing hitting day 45 without a sale, one late shipment, one string of low-view impressions — can quietly demote adjacent listings you never suspected were connected.
This guide walks through all 12 ranking signals Cassini uses in 2026, ordered by observed weight from tracking data across 3,400+ eBay stores. If you've watched sales collapse 40% overnight with no obvious cause, one of these 12 signals moved against you.
What Cassini actually is
Cassini launched in 2013 to replace eBay's older keyword-heavy ranking system called Voyager. The rebrand mattered because Cassini isn't a search engine in the traditional sense — it's a real-time scoring pipeline that assigns quality vectors to every listing, every seller, and every category, then feeds those vectors into the ranking layer buyers see (Best Match).
The most important distinction: Cassini scores are dynamic and personalized. Two buyers searching the same term at the same second can see different results depending on their purchase history, shipping location, and category preferences. Sellers optimize for the score, not for a single global rank.
The 12 signals, ranked by observed weight
1. Listing freshness
The single most impactful signal for auto-parts, apparel, and general merchandise categories. Listings older than 45 days lose an average of 38% of impressions on our tracked dataset without any change to seller or product. Cassini treats freshness as a proxy for inventory availability — an old listing is likely to have been picked over. The fix is boring but non-negotiable: end and relist affected items to reset the freshness dimension. Impressions typically recover within 72 hours.
2. Item specifics coverage
Sellers underestimate this signal because it doesn't feel like a ranking factor — it feels like data entry. But Cassini uses item specifics to match listings against long-tail queries. A "2013 Ford F-150 XLT 3.5L EcoBoost right headlight" search only hits listings where those exact aspects are filled. Coverage below 70% blocks entire query categories from finding you. Above 90% correlates with a 30-40% impression lift in our data.
3. Sell-through rate
Cassini calculates a rolling 30-day sell-through per listing and per account. A listing that sells consistently gets promoted; a stale one gets demoted. The metric is never surfaced to sellers directly in Seller Hub, but it's derivable from your active listings count vs sold count. Values above 15%/month indicate healthy Cassini positioning for the category. Below 5% and you're in demotion territory.
4. Buyer engagement signals
Watchers, saves, offers, questions asked — all count. Cassini reads these as micro-conversions. Listings with active engagement (watch count growing week-over-week) get boosted even before they convert. Zero-engagement listings, especially those with high impressions, get demoted fast because Cassini interprets impressions-without-clicks as "buyers actively rejecting this listing."
5. Late shipment rate
Defect-linked demotion. Late shipments above 3% on the trailing 60 days push your account below the standard Cassini threshold. Two consequences: your entire seller cluster loses ~15-20% impressions, and eBay refuses to promote your listings via Promoted Listings Standard (the automatic version) until the metric recovers. This is one of the fastest ways to tank an otherwise healthy account.
6. Price competitiveness
Dynamic per-search. Cassini normalizes your price against the current competitor set for each query and adjusts ranking accordingly. Being 15% over the median demotes you; being 5% under boosts you; being 40%+ below flags you as possible fraud and can trigger manual review. Repricers exist specifically to keep this signal healthy without human effort.
7. Return rate and INAD claims
Item Not As Described claims are the most heavily weighted return type. A single INAD hurts more than five buyer-remorse returns. Cassini uses the trailing-60-day rate normalized against category benchmarks. Auto-parts sellers face harder thresholds than general merchandise because return-shipping-cost issues concentrate there.
8. Handling time promise vs actual
Your stated handling time creates a delivery-date promise. Missing it (even by one day) counts as a defect. Cassini rewards sellers who consistently ship faster than promised — the "fast-and-free" badge earned by 1-day handling with 90-day tracked on-time delivery gives a measurable ranking boost.
9. Photo quality and count
Machine-vision scoring runs on primary and secondary photos. Cassini looks for resolution, uniform background, absence of watermarks, and photo count. Listings with 4+ photos rank higher than 1-2 photo listings in identical queries. Photos with visible eBay-forbidden elements (competitor logos, promotional text) get subtle demotion.
10. Category depth match
Whether your listing sits in the correct leaf category matters more than most sellers realize. Cassini penalizes listings in overly broad categories because they compete against wrong-context listings. A car mirror in "Motors > Parts > Mirrors > Side Mirrors" beats the same product in "Motors > Parts > Other."
11. Feedback score (weighted by recency)
Your all-time feedback matters less than your last 90 days. Cassini decays historical feedback exponentially. A new seller with 99% positive on 40 recent transactions ranks better than an old seller with 99.8% positive averaged over 20,000 transactions if the recent 90 days show any dip.
12. Account-wide "seller cluster" health
The catch-all. Cassini computes a rollup score across all your listings and adjusts individual scores downward if the account average is falling. This is why a single problem listing can pull others down with it — the cluster score drops, and every listing inherits a proportional demotion.
How Cassini's demotion cascade works
The scariest part of Cassini isn't any single signal — it's how they interact. A typical demotion cascade looks like this: listing hits day 45 (signal 1 drops), impressions fall 40% within 3 days (signal 4 drops as engagement falls proportionally), sell-through metric slips (signal 3 drops), account cluster score reflects the pattern (signal 12 drops), and now other listings in the same account start to feel the pull. Recovery from a full cascade takes 2-4 weeks even with aggressive intervention.
The prevention playbook is simpler than the recovery one: keep freshness above 45 days (relist automatically), keep item specifics above 90% coverage (AI aspect fillers), keep late-shipment rate below 2% (real logistics discipline), and monitor impression trends daily so you catch a cascade before it starts.
What you actually control
Of the 12 signals, sellers directly control 8: freshness, item specifics, photos, category, price, handling time, listing description quality, and inventory quantity accuracy. The other 4 (engagement, sell-through, feedback, cluster health) are outputs — they respond to how well you executed the first 8. Focus optimization budget on the controllable inputs; the outputs follow.
Recovery playbook when Cassini demotes you
Step 1: identify the trigger. Pull traffic snapshots for the last 14 days. Which listings lost impressions first? What did they have in common (age, category, price change)? Step 2: fix the underlying signal for those specific listings before anything else. Step 3: monitor for 72 hours. If impressions recover, the fix worked and you can scale it. If they don't, the trigger is elsewhere and you need to dig again.
Most sellers skip step 1 and go straight to "list more" or "lower prices," both of which usually make things worse. Cassini rewards focused optimization on the specific signal that broke, not blanket panic moves.
Profitio's Traffic Intel dashboard surfaces changes to 6 of the 12 Cassini signals in real time and fires alerts 3-5 days before a full cascade completes — start a 14-day free trial if you're running 500+ listings and don't have this visibility today.