How to evaluate a link building agency

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

Why case studies are weak evidence

Every agency has case studies. They are the least informative document in the sales process and the one buyers spend the most time reading.

The problem is not that they are fabricated, though some are. The problem is that a case study describes a result and attributes it to the work. In SEO, results have many causes running simultaneously: the client also improved their site, also published more, also spent more on advertising, also benefited from a competitor's decline, also happened to be in a market where search demand rose. A traffic chart with an arrow on it establishes that traffic rose. It establishes nothing about why.

Testimonials are worse. They tell you a client was satisfied, which is a statement about the client relationship rather than about the links. I have seen entirely satisfied clients whose profiles were, on inspection, dangerous.

What actually discriminates between agencies is not evidence of outcomes. It is evidence of judgment — whether the people you are about to hire can distinguish a good opportunity from a bad one, and whether they will tell you the truth when something does not work. Judgment shows up in how questions get answered, not in what a deck contains.

Ask to see a live link they built

Start here, because it is the fastest test available and almost nobody asks it properly.

Ask for the URL of a page on somebody else's site that contains a link they earned for a client in the last three months. Not a domain name. Not a logo wall. The full URL of the page, so you can open it and read it.

Then read it. You are checking four things:

• Does the page look like something a person would read? Or is it an article on a site whose entire archive is commercially motivated content across unrelated industries?
• Does the link make sense where it sits? A link that a reader might plausibly click, in a sentence where the reference is natural, is a different artifact from a link inserted into a paragraph that exists to hold it.
• Is the surrounding content competent? Thin, generic, or slightly-off writing indicates who produced it and why.
• Would you be comfortable if a competitor found this page and sent it to a journalist? That is the honest standard.

An agency that earns real links will have several to show and will be pleased you asked. An agency that hesitates, offers to send examples later, or explains that client confidentiality prevents it, has told you something. Confidentiality is a real constraint on naming clients. It is not a constraint on showing you a public web page.

Ask what they would refuse to do

This is my favorite question and it is almost impossible to answer well without genuine experience.

Everyone in this industry has a line. The competent ones have thought about where it is, have refused work over it, and can describe the refusal in specifics. The answer you want sounds like: we do not buy links on sites that sell them, we will not run a private blog network, we will not write content under a fake author byline, we will not place links in a subject we cannot write about competently, and last year we turned down a client who wanted us to build to a page we thought would get them into trouble.

The answer you do not want is a version of "nothing, we do whatever gets results" or a fluent recital of Google's guidelines with no specific instance attached. Guideline recitation is free. A named refusal costs something, which is why it is evidence.

There is a related question worth asking immediately afterward: what have you stopped doing in the last few years, and why? The tactics that worked a decade ago mostly do not now. An agency that cannot name something it abandoned has either not been paying attention or is still doing it.

Ask how they measure, and what happens when it fails

Measurement reveals whether an agency understands its own product. Ask what they report on monthly, and listen for whether links are counted at acquisition and then forgotten.

Half of all links ever recorded on my own profile have been lost — 49.6% of them — and 97.4% of those losses happened while the source page remained perfectly reachable. The link was simply removed, or the page was rewritten around it. Only four out of 645,202 lost links were lost to a 404. An agency that reports acquisitions and never checks whether prior placements are still live is reporting a number that overstates the asset it built, possibly by a large margin.

Then ask the failure question, which is the one that actually predicts the working relationship: what happens when a campaign underperforms?

You are listening for a process. A good answer describes how they diagnose — was the asset weak, was the list wrong, was the pitch wrong, was the timing wrong — and what they change first. A good answer also includes an example of a campaign that did not work and what they learned. An agency that has never had a campaign fail is either new or lying, and both are disqualifying in different ways.

The bad answer is that they simply do more outreach, or that they make up the shortfall from other sources. The second version is how clients acquire links they did not want.

The questions, and what a good answer sounds like

Here is the whole set in one place. Take it to the call.

That last row is underrated. An agency that needs nothing from you is planning to produce links that have nothing to do with you.

What to check before the call, and what to check in the contract

Twenty minutes of homework changes the quality of the conversation.

Look at their own link profile. Not because a cobbler's children argument is decisive — agencies genuinely do neglect their own marketing — but because an agency whose own profile is full of the exact placements they sell has shown you their inventory.

Read something they published. Do they write about the work in specifics, or in the vocabulary of a brochure? Specificity is the whole signal.

Find who works there. If the people doing the work are identifiable, have professional histories, and speak publicly about the subject, you know a great deal about what you are buying.

In the contract, three terms matter more than the rest. First, who owns the prospect list and the assets produced — you should, and it should say so. Second, notice period and what happens to work in flight. Third, whether there is any language committing them to a quantity of links, because if there is, you have re-created the problem you were trying to avoid. A contract that commits to hours, activities and reporting cadence is a service agreement. A contract that commits to link counts is a supply agreement.