Expired domain metrics explained
What every metric on a Revised listing measures — Revised Score, referring domains, backlinks, linking sites, Agent Citability (preview), visitor traffic and archive snapshots — and the exact values each can take.
Every listing in the directory carries the same set of metrics, and every interface — the website, the REST API and the MCP server — publishes them the same way. This page explains what each one measures, the exact values it can take, and where it stops being reliable.
For the vocabulary around the metrics — tiers, reveals, holds — see Concepts. For turning them into a decision, see Evaluate a domain.
What gets a domain into the directory
A metric only means something next to the question of what was screened out first. Every candidate is checked before it is listed, and a candidate is dropped outright when it has:
- Spam history — an archived snapshot flagged as spam.
- Parked-only history — every archived snapshot is a parking page.
- Trademark risk — the name contains a major brand.
- No measurable signal — no link evidence at all.
These are hard exclusions, so you will not see a listing that fails them. It is also why the bottom of every scale means modest history, not bad history. The full screening process is described in How Revised verifies domain quality.
The metrics at a glance
| Field | Measures | Published as | Can be null |
|---|---|---|---|
revised_score | Overall grade, 0–100 | A range, { "low": 65, "high": 75 } | No |
referring_domains | Distinct domains linking in | Band: <10 … 250+ | No |
backlinks | Total links | Band: <100 … 100K+ | Yes |
linked_from | Tracked sites that link in | Array of names, with linked_from_count | Empty array |
agent_citability (preview) | Links from sources AI assistants cite | { "score": 0–100, "basis": ..., "sources": [...] } | Yes |
visitor_traffic_band | How popular the site was with real visitors while it was live | { "rank_bucket": ... } | Yes |
snapshots | Archived copies of the old site | <10, then 10+, 20+ … | No |
age_years | Age in whole years | Integer | Yes |
category | What the site was about | One of 20 codes | Yes |
estimated_value_band | Valuation, coarse range | String, e.g. $250 to $999 | No |
estimated_value_cents | Valuation, low bound | Integer, in cents | Yes (paid plans only) |
When a field is null, the row’s metrics_missing array names it. A null is absent data, never a zero — see Missing is not zero.
Revised Score
A 0–100 grade that lets you compare two listings without weighing five separate metrics against each other. It is a percentile blend of link-graph centrality, referring domains and site history, computed across the whole directory, plus up to 15 points for real visitor traffic, so a 70 means the same thing on every listing.
It is published as a range centred on the exact score — { "low": 65, "high": 75 } on the REST API, "65-75" on an MCP summary row. revised_score_exact carries the integer score. rsMin and rsMax filter on the range.
The website grades the score into five bands:
| Grade | Score |
|---|---|
| Elite | 80 and above |
| Strong | 65–79 |
| Good | 50–64 |
| Fair | 30–49 |
| Emerging | Under 30 |
There is no red band, because a domain that would grade badly is excluded before it is listed. More in What is Revised Score?
Referring domains
How many distinct domains link to the name. It is the most durable signal on the row: a thousand links from one site is one relationship, and a count of distinct sites is much harder to manufacture than a raw link count.
| Published value | Filter id (rd) |
|---|---|
<10 | lt10 |
10-50 | 10-50 |
50-100 | 50-100 |
100-250 | 100-250 |
250+ | 250plus |
The filter takes the ids in the right-hand column, not the values — see Filters. Background: Referring domains vs backlinks.
Backlinks
Total links, from every linking page. Banded on a roughly logarithmic ladder, because link counts span five orders of magnitude: <100, 100-500, 500-1K, 1K-5K, 5K-25K, 25K-100K, 100K+.
Read it against referring_domains. A large backlink band over a small referring-domain band usually means one site linking many times — a sitewide footer or blogroll — rather than broad support.
Linked from
linked_from names the sites we track that link to a domain — Wikipedia, Hacker News, GitHub, universities, government bodies, standards organisations and news outlets — strongest first. It lists every known site, not a sample, and linked_from_count is the headline figure. Filter on the count with linkedFromMin.
It is usually the most informative field on the row. A small referring-domain count with Wikipedia or a university in linked_from often says more than hundreds of anonymous links. Filter by name with source, or see Find expired domains by linking site.
Agent Citability (preview)
A preview 0–100 score for how deeply sources that AI assistants cite link to a domain. It does not predict citation, and it is a different question from a plain link count, so a domain can rate well on one and poorly on the other.
agent_citability carries the score, a basis and the linking sources. basis is seed_links when AI-cited sources link to the domain and are listed in sources, and link_graph when none do: sources is then empty and the score reflects the domain’s position in the link graph alone, capped at 70. It is null when the domain has no row in the current index. On a listing that means no AI-cited links: none of our trusted seed sites link to the name. Older listings stay live without a score. It is never zero and not an error. The methodology is published in full. Background: AI citability and expired domains.
Visitor traffic
visitor_traffic_band is a coarse band on every row: how popular the site was with real visitors while it was live, drawn from a monthly global top-1M list of the sites browser users visit. rank_bucket is 100000 when the site’s best month was inside the top 100k. The detailed month-by-month record, visitor_traffic, is null on any row whose name is not disclosed to you. Both are null for a site that never made the list. Filter with traffic and trafficSustained.
Snapshots
How many times the old site was captured by the public web archive. A registration date proves someone paid for a name; a snapshot record is evidence that a site actually existed.
Published as <10, then the count rounded down to the nearest ten with a +: 40+ means at least 40. Thin archive coverage is not disqualifying, but it is less evidence, and the Revised Score reflects that.
Age
age_years is the domain’s age in whole years, where Revised can establish it. It is one of the fields most often null, which makes ageMin the filter most likely to cut a result harder than you meant — see Filters.
Category
What the site was about, as one of Revised’s own codes (list_categories returns the current set). It is inferred from archived pages, so it can be the broad other, wrong, or missing. The codes and counts are in Filter the directory.
Estimated value
estimated_value_band is a coarse range on every row. estimated_value_cents is the low bound of Revised’s valuation, in cents, falling back to a reference figure on rows the model has not valued, on paid plans only. It is the figure the priceRange filter buckets. To value any name you type, see Checks on any domain. Nothing in the directory is sold by Revised, so read it as a signal about the name, not a price.
Why ranges instead of exact numbers
Two reasons. Exact figures across thousands of listings would let anyone reconstruct the dataset and identify masked names without revealing anything. And the underlying link data shifts week to week: a domain measured at 137 referring domains today might measure 119 next week with nothing about it having changed, so a precise number would claim a confidence the data does not have.
Bands are computed the same way at every tier, so any two listings are comparable. To sort, flatten them to a midpoint and keep the band for display — Evaluate a domain has a parser.