Competitor link gap analysis

Lesson 18 of 50Module 4 · Prospecting and research6 min read

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

What a link gap actually is

A link gap analysis — also called a link intersect, and implemented in Majestic as Clique Hunter — compares the referring domains of several sites and returns the domains that link to some of them but not to others. In the standard configuration you supply three to five competitors plus your own site, and ask for domains linking to at least two competitors and not to you.

The logic is worth stating precisely, because people over-read the output. A domain in that result has demonstrated two things: it links out at all, and it considers this subject area within scope. Both are genuinely useful. It has not demonstrated that it will link to you, that it accepts pitches, that the link was editorially earned, or that the link is still there. Those are separate questions and the intersect does not answer any of them.

Treat the output as a candidate list with unusually good odds, not as a prospect list. It still goes through contact-finding and qualification like everything else. Where it beats operator-based prospecting is precedent: somebody at that domain has already made the decision you are about to ask for.

Choosing the competitors

The quality of the output is governed almost entirely by this step, and it is the step most often done carelessly.

Use search competitors, not business competitors. The companies your sales team loses deals to may have no content operation at all. Pull the sites that actually rank for the queries your target pages are meant to win, and use those. If a review site, a trade magazine and a Wikipedia-style reference dominate your results, those are your competitors for this purpose, even though you cannot sell against them.

Match them roughly on scale. Putting a site with two million referring domains into an intersect against your two hundred does not produce a comparison, it produces the larger site's link profile with a few names removed. If the only realistic comparisons are giants, restrict the analysis to a specific section or subfolder rather than the whole domain.

Prefer three to five. Two competitors give you an overlap that is mostly coincidence. More than five and the overlap threshold either admits everything or nothing, and the output stops discriminating.

Include at least one you consider annoyingly good at this. The competitor whose links you cannot account for is the one the analysis exists to explain.

Running the intersect

The mechanics are similar across tools. You enter your domain and the competitor domains, choose whether to compare at domain or URL level, set the minimum number of competitors a referring domain must link to, and export.

Three settings decide what you get.

• Domain level, not URL level, for the first pass. URL-level intersects are for a specific page you are trying to match. Domain level answers the broader question of who covers this space.
• Overlap threshold of two to start. Raising it to three tightens the list to sites that clearly treat this as a beat. Lowering it to one is not an intersect; it is a merged export.
• Exclude domains already linking to you. Obvious, and frequently forgotten, and it is embarrassing to pitch a site that has linked to you for six years.

Run it more than once. Threshold two gives you volume; threshold three or four gives you the short list of sites that repeatedly cover the topic, and that short list is usually where the campaign should start. If your tool supports it, pull the linking page URL and the anchor text as well as the domain — you will need both when you read the output.

Reading the output honestly

A raw intersect export is mostly noise. Strip it before anyone else sees it.

• Collapse sitewide links. One domain in my own profile sent 219,159 links, 16.8% of the entire profile, from a single sitewide placement. In an intersect a sitewide footer or blogroll link looks like enormous endorsement and represents one decision made once, possibly years ago. Deduplicate to one row per referring domain and check whether the link appears in a template.
• Separate editorial from mechanical. Aggregators, scrapers, directory-style listings, comment sections and job boards will all show up as linking to several competitors. They link to everyone. That is not a signal about your topic.
• Check what the link actually is. Open a sample of twenty linking pages by hand. If most turn out to be paid placements, guest post bylines with a bio link, or press release syndication, the intersect has told you your competitors buy links, not that a prospect exists.
• Note the pattern, not just the domain. If three competitors are all listed on the same kind of page — an industry association member directory, a regional supplier list, a university course resource page — the pattern generalizes. There are usually forty more pages of that type your intersect did not surface, and Lesson 4.3 is how you find them.

When the result is empty, and why that is informative

I want to be honest about something the tool vendors are not: a null or near-null result is common, and it is not a failure of technique.

You will often run a clean intersect on well-chosen competitors and get back forty domains, of which thirty-five are aggregators and five are sites you already have. This happens routinely in three situations: markets where nobody links because the audience does not publish; markets where the competitors' profiles are almost entirely bought, and therefore consist of sites that link to anyone paying rather than to a topic; and markets where you are genuinely the incumbent and the gap runs in the other direction.

Each of those readings changes what you do next, which is why the null result is worth having.

• Nobody links in this market. Then outreach volume is the wrong strategy and creating something citable — data, a tool, a definitive reference — is the only route. Digital PR aimed outside the immediate niche often works better than in-niche outreach.
• The competitors bought their links. You have just learned their profile is not a model to copy, and you have a defensible answer for the executive asking why competitor X has four times your link count.
• The gap runs your way. Reverse the analysis. Domains linking to you but not to competitors are your relationship inventory, and Module 5 is about not losing them.

What you must not do is loosen the criteria until something comes out. Dropping the threshold to one, adding six more competitors, and including every scraper produces a list of two thousand domains and a campaign that fails slowly. An empty intersect answered honestly is worth more than a full one produced by torturing the settings.

Turning the gap into a prospect list

What survives the cleanup goes through the same pipeline as everything else, with one addition: record why each domain is on the list.

For every row, capture the competitor or competitors it links to, the URL of the page carrying that link, and the anchor text. That context is what makes the eventual pitch specific. There is a large difference between an email that says I noticed you cover this topic and one that says your supplier directory lists two firms in this category and omits a third — and the second only exists because you kept the source URL.

Then sort into three buckets. Sites linking to competitors from a page you could plausibly be added to — a resource list, a directory, a roundup — go first, because the ask is small and concrete. Sites linking from editorial articles go second, because the ask is a future mention rather than an edit. Sites linking from anything mechanical go in the bin.

One more discipline: date the export. Competitor profiles change, and an intersect from last quarter will contain links that no longer exist. Given that the median referring domain in my own data stopped linking after 1,080 days, an intersect is a photograph of a moving thing, not a map.