Relevance: the metric nobody sells you

Lesson 6 of 50Module 2 · Judging link quality6 min read

By Bill Hartzer · Hartzer Consulting · 25+ years in search, domains and expert witness work

Why nobody sells you relevance

Open any link vendor's price list and you will find it sorted by a metric: Domain Rating, Domain Authority, Trust Flow, sometimes estimated traffic. You will not find it sorted by relevance. There is a reason for that, and it is not conspiracy — it is inventory management.

A vendor holds a stock of sites they can place links on. To sell that stock efficiently they need to describe it in terms that apply to every buyer identically. A Domain Rating of 55 means the same thing to a plumbing company and a fintech startup, so it can be printed on a list and priced. Relevance means something different to every buyer: the site that is perfectly relevant to the plumber is worthless to the fintech. A vendor cannot stock relevance, cannot price it, and cannot promise it at volume. So they sell the number instead, and buyers — reasonably enough — start believing the number is the product.

The consequence is that relevance is systematically underweighted across the entire industry. It is the cheapest quality signal to check and the one most often skipped. When I audit a link profile that underperformed, the pattern is almost always the same: the metrics look defensible and the subjects are all over the map. Fifty links from fifty sites about fifty different things does not tell a search engine what the target page is about. It tells a search engine that somebody was buying links.

Three levels of relevance

Relevance is not one property. It exists at three levels, and they can disagree with each other. Check all three.

Site-level relevance

What is this site, as a whole, about? Not what does its home page headline claim — what do its last thirty published pages actually cover? A site that publishes about commercial roofing, roofing materials, and roofing contractor business practices is topically coherent. A site that publishes about roofing, cryptocurrency, CBD, and online casinos is not a site about anything. It is inventory.

Page-level relevance

Site relevance without page relevance is weak. A well-regarded medical journal is a relevant linking site for a healthcare page, but a link from that journal's staff directory is not a relevant link. The page carrying your link should be about the thing your page is about. A link from a page on "how to choose a commercial roofing contractor" to your commercial roofing services page is coherent. A link from that same site's "office holiday party" post is not.

Paragraph-level relevance

This is the level almost nobody checks, and it is the one that separates an editorial link from a placement. Read the sentence and the paragraph the link would sit in. Would a reader who is genuinely interested in that paragraph want to click through? Does the link continue the thought, or does the paragraph swerve to accommodate it? Inserted links betray themselves here: a paragraph about roof inspection intervals that suddenly mentions a city name and a service phrase reads wrong to a human, and the pattern is detectable at scale.

Why a relevant TF 20 beats an irrelevant TF 50

Trust Flow is Majestic's estimate of how trustworthy the sites linking to a page are, on a 0 to 100 logarithmic scale. It is a good metric. It is also a metric about the linking site's own backlinks — it says nothing at all about whether that site has anything to do with yours.

Consider what each link actually does. A link is a citation. Its value to the receiving page comes partly from the authority flowing through it and partly from the context it establishes: this page, about this subject, was worth citing by that page, about that subject. A relevant link contributes to both. An irrelevant link contributes to one and actively muddies the other.

There is a second effect, and in my experience it is the larger one. Links from relevant sites tend to sit on pages that real people read for reasons connected to your subject. Those readers click. That traffic is qualified. An irrelevant TF 50 link on a page nobody in your market reads sends you nothing but a line in a backlink report.

There is a third effect that only shows up over years. Relevant links are usually placed because someone found the resource useful, and those links survive. Placements on unrelated sites are transactional, and transactional links come down when the arrangement ends, the site is sold, or the page is rewritten. On my own site's profile, the median referring domain stops linking after about 1,080 days and 49.6% of all links ever recorded have been lost — and that is on a profile where nothing was ever bought. Links that exist for a reason last longer than links that exist for a fee.

None of this means metrics are useless. It means the ordering matters. Filter for relevance first, then use authority metrics to rank what survives the filter. Doing it the other way round gives you a shortlist of impressive sites that have nothing to do with you.

How to judge relevance in practice

This is a ten-minute process per candidate, and it does not require a paid tool.

1. Read the last twenty posts. Not the categories in the navigation — the actual recent output. Categories are aspirational; the archive is factual.
2. Write the site's topic in one sentence. If you cannot do it without the words "general", "lifestyle", "business and technology", or "news and more", the site does not have a topic.
3. Check the Topical Trust Flow categories. Majestic assigns topic categories based on where a site's own backlinks come from. If a site claims to be about home improvement and its topical categories are Computers and Adult, the claim and the link graph disagree. Trust the link graph.
4. Find the specific page. If the opportunity is a specific URL, judge that URL. If the opportunity is "a link on our site", ask which page, and treat vagueness as an answer.
5. Read the paragraph aloud. Awkwardness you would skim past on screen is obvious when spoken.
6. Ask who the audience is. If you cannot describe the person who reads this site, nobody does.

Fake relevance and how it is manufactured

Because buyers have started asking about relevance, sellers have started manufacturing it. Learn the three common forms.

Keyword matching instead of topic matching. A page that mentions your keyword eleven times is not a page about your subject. Article directories built their entire business on this: dump enough keyword-stuffed articles into a database and every buyer finds a "relevant" page. The pages had no readers, no editorial standard, and no reason to exist. Google dealt with them and the tactic has been dead for over a decade, but the underlying trick — matching strings rather than subjects — is alive in modern guest post marketplaces.

The category page on a multi-topic site. A network site publishes in fifteen unrelated categories and sells you a link in the one that matches your niche. The category exists to make the sale, not to serve an audience. Check whether the other categories on the site have anything to do with each other. If they do not, the relevance is staged.

The relevant paragraph in an irrelevant post. Someone commissions a 900-word article on a topic adjacent to yours, published on a site about something else entirely, purely to hold your link. The page is relevant to you. The site is relevant to nobody. This is the standard shape of a paid guest post, and while guest posting can be legitimate, this version of it is a placement with a costume on.

The test that cuts through all three: would this page exist if nobody were paying for a link on it? If the honest answer is no, you are not buying relevance. You are buying the appearance of it.