Umbra Systems Network, method

Every check, and the arithmetic under it

27 checks are defined in the analysis layer and run against every domain in the corpus. This page lists all of them. Each has an anchor, and every line of an individual reading links back to the anchor of the check that produced it.

84,023

is what every counted row on this page adds up to

could judge, plus could not judge, plus not applicable, on all 24 counted rows

Corpus generated 2026-08-09, exported 2026-08-09. Question bank v1. Nothing on this page is judged here: criteria are defined once, in the analysis layer, and this site reads its output back.

01 Sources

Where the readings come from

The site does not judge anything. Criteria are defined once, in the analysis layer, which writes one file per domain; this site reads those files back and displays them. If a verdict here disagrees with a verdict there, this page is the one that is wrong.

84,145 lines were read across 11 collection files. 122 of them could not be parsed. Those are excluded from every denominator on the site rather than silently counted as a pass, which leaves 84,023 domains, and 84,023 judgement files were read back to build this page.

02 Denominators

Every denominator, line by line

A check does not run on every domain. Some are gated behind a precondition; some do not apply at all. Both cases leave the denominator, and both are counted here rather than folded into a pass. The last column is the arithmetic: it is the same number on every row and it equals the corpus.

Check	Found	Expected	Detector gap	Could judge	Could not judge	Not applicable	Adds up to
Home page openshard	1,427	76,517	0	77,944	6,079	0	84,023
Home page status codehard	755	78,330	0	79,085	4,938	0	84,023
https enforcedhard	5,447	70,414	0	75,861	8,162	0	84,023
Domain expiry datehard	2,842	79,675	0	82,517	1,506	0	84,023
SPF record counthard	869	83,100	0	83,969	54	0	84,023
Sending IP on a blocklisthard	373	64,815	0	65,188	0	18,835	84,023
Printed address can receive mailhard	159	71,130	0	71,289	12,734	0	84,023
Domain carries MX recordssoft	14,789	68,087	0	82,876	1,147	0	84,023
Last change to the DNS zonesoft	13,560	28,994	0	42,554	41,469	0	84,023
Mobile viewport tagsoft	2,564	68,825	0	71,389	12,634	0	84,023
LocalBusiness structured datasoft	37,363	34,026	0	71,389	12,634	0	84,023
Image alt text, five or moresoft	6,482	64,907	0	71,389	12,634	0	84,023
Image width and heightsoft	37,577	33,812	0	71,389	12,634	0	84,023
Tappable phone numbersoft	11,186	60,203	0	71,389	12,634	0	84,023
Analytics tag on the home pagesoft	0	54,000	17,389	71,389	12,634	0	84,023
Online booking entry pointsoft	0	8,704	62,685	71,389	12,634	0	84,023
Home page HTML sizesoft	20	71,369	0	71,389	12,634	0	84,023
Sitemap named in robots.txtsoft	2,870	65,418	0	68,288	15,735	0	84,023
Sitemap self reported update, three year linesoft	3,154	46,254	0	49,408	34,615	0	84,023
Description on the Google profilesoft	26,768	57,152	0	83,920	103	0	84,023
Cover photo on Googlesoft	18,825	64,199	0	83,024	999	0	84,023
Claim prompt on Googlehard	10,244	73,676	0	83,920	103	0	84,023
Opening hours on Googlesoft	7,506	76,414	0	83,920	103	0	84,023
Unanswered complaints on BBBhard	320	0	0	320	83,703	0	84,023
Image alt text, five or more, broader variantsoft	Defined in the analysis layer, not counted on this pass. No figure is printed for it anywhere on the site.
Sitemap self reported update, three year line, variant at 730soft	Defined in the analysis layer, not counted on this pass. No figure is printed for it anywhere on the site.
Unanswered complaints on BBB, variant at 1soft	Defined in the analysis layer, not counted on this pass. No figure is printed for it anywhere on the site.

24 of 27 checks produced counts on this pass. Every counted row sums to 84,023: could judge, plus could not judge, plus not applicable. That identity is asserted when this page is built, and a failure stops the build rather than printing a page about denominators whose own denominators do not close.

03 Refusals

Conclusions we deliberately do not draw

We did not find it does not mean it is not there. A thing missing from the copy we fetched is recorded as not seen, and not seen is never promoted to absent. Only a source that positively states an absence makes an absence assertable. Across this corpus that is 312,522 readings we hold rather than state.
One record is not one company. The registries carry a practice more than once, under more than one name, at more than one address. Nothing here counts companies; every denominator on this site counts domains, and that is a different thing.
Being refused is not being broken. A run of status codes in this corpus came from our own collection being rate limited, not from anything failing. Those responses are removed rather than counted; a 429 or a 403 tells us about our request, not about whether a patient can open the page.
Home page load time is deleted, not merely unpublished. The readings were distorted by the way we collected them. Twelve hosts were re-fetched by hand and all twelve had read more than three times high; the worst record in the corpus was seventeen minutes, which was our own thread waiting in a queue. The whole axis is gone. It is not in the checks below, it is not behind a caveat, and it will not come back.
A detector gap is our hole, not their absence. Two checks below match against a fingerprint table that does not cover enough vendors. They are marked, they produce no findings, and their rates are never reported as facts about anyone.
We do not score, rank or grade. There is no composite number here, because any weighting that produced one would be our opinion wearing the clothes of a measurement.

Each of these has cost us a published figure at least once. The rule that a rate must reproduce under a stated denominator retired a claim of 95.5% for enforced https, and a storefront rate of 39.00% that had quietly mixed soft readings into a hard count.

Every check

What each one reads, and when it returns nothing

A hard check can carry a factual statement. A soft check is context only and never appears as a finding. A gate is a precondition: when it does not pass, the check produces nothing rather than a guess. Verification time is how long it takes you to check the same thing by hand, which is the only useful answer to whether it matters.

Reachability

Home page opens

#site_deadhard

Counted when the thing is observed directly.

Reads: dns.a, http.error, http.status
Before it runs: One condition has to hold before this runs. If it does not, the check returns nothing.
You can check it by: browser
Which takes you: 20 seconds
When we cannot judge: 6,079 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 1,427 of 77,944 locations it could be judged on, which is 1.83%.

Home page status code

#http_errorhard

Counted when the thing is observed directly.

Reads: http.status
Before it runs: None. This check runs on every domain in the corpus.
You can check it by: browser
Which takes you: 15 seconds
When we cannot judge: 4,938 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 755 of 79,085 locations it could be judged on, which is 0.95%.

Transport security

https enforced

#no_httpshard

Counted when the thing is observed directly.

Reads: derived inside the analysis layer
Before it runs: One condition has to hold before this runs. If it does not, the check returns nothing.
You can check it by: browser
Which takes you: 20 seconds
When we cannot judge: 8,162 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 5,447 of 75,861 locations it could be judged on, which is 7.18%.

Domain registration

Domain expiry date

#domain_expires_30hard

Counted when the thing is observed directly.

Reads: derived inside the analysis layer
Before it runs: None. This check runs on every domain in the corpus.
You can check it by: third_party_url
Which takes you: 30 seconds
When we cannot judge: 1,506 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 2,842 of 82,517 locations it could be judged on, which is 3.44%.

Mail path

SPF record count

#dup_spfhard

Counted when the thing is observed directly.

Reads: derived inside the analysis layer
Before it runs: None. This check runs on every domain in the corpus.
You can check it by: third_party_url
Which takes you: 30 seconds
Threshold: 2
When we cannot judge: 54 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 869 of 83,969 locations it could be judged on, which is 1.03%.

Sending IP on a blocklist

#blocklistedhard

Counted when the thing is observed directly.

Reads: derived inside the analysis layer
Before it runs: None. This check runs on every domain in the corpus.
You can check it by: third_party_url
Which takes you: 30 seconds
When we cannot judge: 0 locations could not be judged on this pass and 18,835 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 373 of 65,188 locations it could be judged on, which is 0.57%.

Printed address can receive mail

#no_mx_realhard

Counted only when a source states the thing is absent. Assertable.

Reads: dns.mx, html.mailto, html.emails_in_text, _meta.host
Before it runs: 4 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: dns_lookup
Which takes you: 25 seconds
When we cannot judge: 12,734 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 159 of 71,289 locations it could be judged on, which is 0.22%.

Domain carries MX records

#no_mx_anysoftinventory only

Counted only when a source states the thing is absent. Assertable.

Reads: derived inside the analysis layer
Before it runs: 2 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: dns_lookup
Which takes you: 25 seconds
When we cannot judge: 1,147 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 14,789 of 82,876 locations it could be judged on, which is 17.84%. 14,789 of these rest on absence of evidence alone, so they are phrased as questions.

DNS zone

Last change to the DNS zone

#zone_stale_2ysoft

Counted when the thing is observed directly.

Reads: derived inside the analysis layer
Before it runs: None. This check runs on every domain in the corpus.
You can check it by: dns_lookup
Which takes you: 30 seconds
Threshold: 730
When we cannot judge: 41,469 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 13,560 of 42,554 locations it could be judged on, which is 31.87%.

Mobile

Mobile viewport tag

#no_viewportsoft

Counted when the thing is missing from the copy we fetched. Not assertable; it is phrased as a question, never as a claim.

Reads: derived inside the analysis layer
Before it runs: 2 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: browser
Which takes you: 20 seconds
When we cannot judge: 12,634 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 2,564 of 71,389 locations it could be judged on, which is 3.59%.

Search presentation

LocalBusiness structured data

#no_localbusinesssoft

Counted when the thing is missing from the copy we fetched. Not assertable; it is phrased as a question, never as a claim.

Reads: derived inside the analysis layer
Before it runs: 2 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: third_party_url
Which takes you: 30 seconds
When we cannot judge: 12,634 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 37,363 of 71,389 locations it could be judged on, which is 52.34%.

Accessibility

Image alt text, five or more

#images_no_altsoft

Counted when the thing is missing from the copy we fetched. Not assertable; it is phrased as a question, never as a claim.

Reads: html.images
Before it runs: 2 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: third_party_url
Which takes you: 30 seconds
When we cannot judge: 12,634 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 6,482 of 71,389 locations it could be judged on, which is 9.08%.

Image alt text, five or more, broader variant

#images_no_alt_anysoftvariant of #images_no_alt

Counted when the thing is missing from the copy we fetched. Not assertable; it is phrased as a question, never as a claim.

Reads: html.images
Before it runs: 2 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: view_source
Which takes you: 30 seconds
When we cannot judge: Not counted on this pass, so there is no figure to give.

Defined in the analysis layer and not counted on this pass. No figure for it is printed anywhere on this site.

Layout

Image width and height

#images_no_dimssoft

Counted when the thing is missing from the copy we fetched. Not assertable; it is phrased as a question, never as a claim.

Reads: derived inside the analysis layer
Before it runs: 2 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: browser
Which takes you: 25 seconds
Threshold: 5
When we cannot judge: 12,634 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 37,577 of 71,389 locations it could be judged on, which is 52.64%.

Contact details

Tappable phone number

#no_tel_linksoft

Counted when the thing is missing from the copy we fetched. Not assertable; it is phrased as a question, never as a claim.

Reads: derived inside the analysis layer
Before it runs: 2 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: browser
Which takes you: 15 seconds
When we cannot judge: 12,634 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 11,186 of 71,389 locations it could be judged on, which is 15.67%.

Analytics

Analytics tag on the home page

#no_analyticssoftmeasures our coverage

Counted when the thing is missing from the copy we fetched. Not assertable; it is phrased as a question, never as a claim.

Reads: derived inside the analysis layer
Before it runs: 3 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: view_source
Which takes you: 25 seconds
When we cannot judge: 12,634 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 0 of 71,389 locations it could be judged on, which is 0.00%. A further 17,389 could not be matched against our fingerprint table, which is a gap in our coverage rather than a reading about anyone. 17,389 of these rest on absence of evidence alone, so they are phrased as questions.

Booking

Online booking entry point

#no_bookingsoftmeasures our coverageinventory only

Counted when the thing is missing from the copy we fetched. Not assertable; it is phrased as a question, never as a claim.

Reads: derived inside the analysis layer
Before it runs: 3 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: browser
Which takes you: 20 seconds
When we cannot judge: 12,634 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 0 of 71,389 locations it could be judged on, which is 0.00%. A further 62,685 could not be matched against our fingerprint table, which is a gap in our coverage rather than a reading about anyone. 62,685 of these rest on absence of evidence alone, so they are phrased as questions.

Page load

Home page HTML size

#heavy_3mbsoft

Counted when the thing is observed directly.

Reads: derived inside the analysis layer
Before it runs: 2 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: browser
Which takes you: 30 seconds
Threshold: 3,000,000
When we cannot judge: 12,634 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 20 of 71,389 locations it could be judged on, which is 0.03%.

Sitemap

Sitemap named in robots.txt

#no_sitemapsoft

Counted when the thing is missing from the copy we fetched. Not assertable; it is phrased as a question, never as a claim.

Reads: site.robots_txt, site.robots_sitemaps, site.sitemap_url_count
Before it runs: One condition has to hold before this runs. If it does not, the check returns nothing.
You can check it by: browser
Which takes you: 15 seconds
When we cannot judge: 15,735 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 2,870 of 68,288 locations it could be judged on, which is 4.20%.

Sitemap self reported update, three year line, variant at 730

#site_stale_2ysoftinventory onlyvariant of #site_stale_3y

Counted when the thing is observed directly.

Reads: derived inside the analysis layer
Before it runs: One condition has to hold before this runs. If it does not, the check returns nothing.
You can check it by: browser
Which takes you: 30 seconds
Threshold: 730
When we cannot judge: Not counted on this pass, so there is no figure to give.

Defined in the analysis layer and not counted on this pass. No figure for it is printed anywhere on this site.

Sitemap self reported update, three year line

#site_stale_3ysoftinventory only

Counted when the thing is observed directly.

Reads: derived inside the analysis layer
Before it runs: One condition has to hold before this runs. If it does not, the check returns nothing.
You can check it by: browser
Which takes you: 30 seconds
Threshold: 1,095
When we cannot judge: 34,615 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 3,154 of 49,408 locations it could be judged on, which is 6.38%.

Google business profile

Description on the Google profile

#gbp_no_descriptionsoft

Counted only when a source states the thing is absent. Assertable.

Reads: derived inside the analysis layer
Before it runs: 3 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: google_maps
Which takes you: 20 seconds
When we cannot judge: 103 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 26,768 of 83,920 locations it could be judged on, which is 31.90%.

Cover photo on Google

#gbp_no_own_photossoft

Counted only when a source states the thing is absent. Assertable.

Reads: derived inside the analysis layer
Before it runs: 3 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: google_maps
Which takes you: 20 seconds
When we cannot judge: 999 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 18,825 of 83,024 locations it could be judged on, which is 22.67%.

Claim prompt on Google

#gbp_claim_offeredhard

Counted when the thing is observed directly.

Reads: derived inside the analysis layer
Before it runs: One condition has to hold before this runs. If it does not, the check returns nothing.
You can check it by: google_maps
Which takes you: 20 seconds
When we cannot judge: 103 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 10,244 of 83,920 locations it could be judged on, which is 12.21%.

Opening hours on Google

#gbp_no_hourssoft

Counted only when a source states the thing is absent. Assertable.

Reads: derived inside the analysis layer
Before it runs: 3 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: google_maps
Which takes you: 20 seconds
When we cannot judge: 103 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 7,506 of 83,920 locations it could be judged on, which is 8.94%.

BBB file

Unanswered complaints on BBB

#bbb_complaints_unansweredhard

Counted when the thing is observed directly.

Reads: derived inside the analysis layer
Before it runs: 2 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: bbb_site
Which takes you: 25 seconds
Threshold: 1
When we cannot judge: 83,703 locations could not be judged on this pass and 0 were out of scope for it. Both leave the denominator; neither is counted as a failure.

Found on 320 of 320 locations it could be judged on. There is no rate to quote for this one: the check never produces a passing case, so its denominator is its own numerator and a percentage here would always read as one hundred.

Unanswered complaints on BBB, variant at 1

#bbb_complaintssoftvariant of #bbb_complaints_unanswered

Counted when the thing is observed directly.

Reads: derived inside the analysis layer
Before it runs: 2 conditions have to hold before this runs. If any fails, the check returns nothing.
You can check it by: bbb_site
Which takes you: 25 seconds
Threshold: 1
When we cannot judge: Not counted on this pass, so there is no figure to give.

Defined in the analysis layer and not counted on this pass. No figure for it is printed anywhere on this site.

One domain at a time

This page is the whole method. If you want to see what it produced for one domain, the lookup on the baseline page takes the domain exactly as it is spelled and opens the reading for it, or says plainly that it has none.

Look up a domain

checks defined

24 produced counts on this pass

Alex Zhang; Umbra Systems. Every figure here comes from one ledger with its source file and denominator recorded. A figure with no ledger entry is not printed.