How every number is measured
Liveness
We fetch each directory's listing pages on a regular schedule with an identifying user agent (LineofsitesBot). A directory is alive when its listing pages respond, dead when they stop, and blocks crawler when it refuses automated access (403/429). Being blocked is itself published signal; we never work around a block.
Dofollow or nofollow
We read the actual rel attributes on product links found on real listing pages. A directory counts as dofollow when the majority of observed product links carry no nofollow/sponsored/ugc attribute. Claimed link policy is ignored; observed markup wins.
Indexation rate
The share of a directory's listing pages that Google has actually indexed, from a rolling sample of listing URLs per directory (rechecked at most every 30 days, sample sizes shown as they grow). A check asks one exact URL through a search API as a site: query, and the instrument that answered is recorded on every check. Where a listing page sits on a domain whose owner connected Search Console to us, the check instead asks Google's URL Inspection API through that read-only credential, which is Google reporting the state of its own index; the connect screen names this use, the answer is recorded as gsc-inspection, and traffic figures from Search Console never enter any number we publish. When no instrument can answer, the check records nothing rather than recording a no. A listing that was never indexed sent nobody anywhere; this number is published nowhere else and it is the one to watch.
Listings and link loss
Every observed product link is a timestamped record: first seen, last seen, and (when it vanishes for two consecutive crawls) disappeared. History is append-only; we never rewrite it. That is what makes "your link quietly died in March" answerable.
The rating
Each measured directory gets a score out of 100, so the table answers "is this worth submitting to" instead of leaving you to weigh four raw numbers. The weights are indexation 45, link value 25, reachability 15, corpus size 15. Link value is multiplied by indexation rather than added beside it, because a followed link on a page Google never indexed passes no authority; a floor remains for the visitors who still click it. Corpus size is log-scaled, so a directory with six listings does not score like one with six hundred. A directory we have not measured is shown as "not rated" rather than given a default. There is no editorial adjustment and no paid input: a directory that pays for faster review scores exactly what its data says.
Domain rating
DR is third-party data shown for orientation only, always labeled as such. We do not verify it and we never sort by it by default.
Paid submissions and neutrality
We sell a service where we submit a product to directories by hand. That work cannot move any number on this site: metrics, sort positions and inclusion are computed from what our crawler measures, and submission volume is not an input to any of them. This is enforced in code by an automated check, not by policy. We publish the acceptance and indexation rates of our own submissions, by mode, even when they are unflattering.
Acceptance rates of our own submissions
When we submit a product to a directory we record the submission, the directory, and the mode the work was done in: by hand by us, prefilled by us and confirmed by the customer, or submitted by someone else and reported to us. The mode is written once and never edited, which is what makes the three comparable. Nothing else about the submission is an input to the rate.
A submission counts as accepted when our crawler later finds the listing it produced on a real page of that directory, read out of the same append-only listing history the rest of the site is built on. A state somebody typed, including one of us typing "approved", is never the source. The approval lag is the wait between the submission and the first time we saw the listing, published as a midpoint with the middle half of the waits beside it, because queues are skewed and one average hides that. Directory pages carry their own figures under "What happened when we submitted".
Measuring acceptance by what the crawler found means a directory we cannot read would otherwise look like one that refused us, so submissions to a directory we have never crawled since sending, or one that blocks our crawler, are held out of the denominator and counted separately rather than scored as rejections. Most of the catalog is in that position today. Submissions we recorded but have not sent, and ones we withdrew ourselves, are held out for the same reason: neither is something the directory did. Every held-out count is printed beside the sample it was removed from.
The industry quotes acceptance rates it computed about itself from its own claims. Ours has two honest limits and they are stated rather than buried: we see only what our users submitted through us and only what our crawler can reach, and paid manual work is measured beside the free paths on the same page, so an unflattering number about our own service is published the same way a flattering one would be.
We have not submitted anything to the directories we track yet, so there is no acceptance rate to publish; it appears here once our own records carry enough submissions to compute one, whichever way it reads.
Popularity vs visibility
Signed-in visitors can upvote any product. Votes are opinion with their own label: they never move a visibility score, a directory rating, a default sort, or a place in the index. Keeping the two streams apart is what lets us publish how well one predicts the other, as a rank correlation between vote counts and measured visibility across products that hold both numbers.
Not enough votes yet to publish a correlation. The coefficient appears once at least 10 products hold both community votes and a measured visibility score; today 0 do.
Corrections
Directory operators can claim their listing and correct how it is presented: the name, tagline, description, icon and submission link. Those stay labelled "operator-reported" permanently, because they are the operator's words rather than something we measured. Nothing measured is editable by anyone. If you spot an error in a measured value, tell us and we fix the data, not the story.
Terms used here
This page states how each number is measured. What each term means, and how the product itself works, lives in the documentation; the glossary defines every term in one sentence.