Audit Your Blog CTAs: One Reader Decision, One Next Step

Most blog archives end every post the same way: a box that says "Talk to us" or "Get started", identical on a beginner's explainer and an expert's comparison, placed after the conclusion where the reader who wanted it has already left. The fix is not a better button. It is a question asked once per post: what decision did this reader come here to make, and what is the one next step that helps them make it? This guide is the method for answering that question across an archive in a day, with the research that explains why the generic button underperforms and the analytics setup that proves whether the replacement did better.

The method rests on three pieces of published usability research from the Nielsen Norman Group, named because they are the kind of source a reader can check. A 2017 study of generic "Get Started" buttons found they attract clicks and mislead the people who click them. A 2016 article set out a five-level trust pyramid describing how much a site may ask of a visitor at each stage. A 2019 article gave four qualities of a link label that tells the truth about the click. Together they say what a post's next step should be, when it may be asked for, and how it should be worded.

It is written for the owner or editor of a working archive, not a new site. If you are deciding which posts deserve attention at all, start with our refresh prioritisation matrix; the CTA audit here is the pass that runs alongside it, because a refreshed post with a wrong next step has only been half refreshed.

The clearest published evidence on generic calls to action comes from a Nielsen Norman Group article by Aurora Harley and Kim Flaherty, published in August 2017, reporting on user studies in which participants met prominent "Get Started" buttons on company homepages. Its summary is blunt: a generic Get Started call to action attracts clicks but also misleads users and acts as a roadblock for those looking to get information about the company. In one study the authors describe, six of eight participants asked to find out what a meal-delivery service offered clicked the button, because it seemed to match their situation, and were sent into a series of sign-up modals with no information in return.

The point that matters for a blog is the one the authors make about measurement. From the organisation's side, they write, those flows would likely be read as successes, because analytics would show the buttons were effective at getting users into the funnel. A blog's end-of-post contact box behaves the same way. It is the most prominent element on the page after the headline, it applies to almost any goal, and a click on it tells you nothing about whether the reader got what they came for. The same study found a participant comparing phone plans who, after a quiz that dead-ended at a request to call, left for a competitor's site and found the price there.

A blog version of the failure: a post explaining what a consent banner is, read mostly by people who have just been told they need one, ending in "Book a strategy call". The reader is at the start of understanding the subject. A call is a level-five ask in the terms of section 03, and the only readers who click are the ones the post did not need to persuade. The replacement is a next step the reader would have asked for, which on that post is probably a checklist of what the banner must do, downloadable or inline.

Every post that earns search traffic does so because someone typed a question, and behind most questions is a decision. The method starts by writing that decision down in one sentence, in the reader's words, for each post. "Should I switch my analytics to server-side tagging?" "Which of these three email tools fits a five-person team?" "Is my traffic drop a Google update or my own change?" "How do I set up a redirect without losing rankings?" The sentence is the post's purpose, and it is the only input the rest of the audit needs.

In practice almost every decision falls into one of four shapes, and the shape tells you the family of next step that fits. A whether-decision wants a way to check the reader's own situation. A which-decision wants the comparison in a form they can take away. A how-decision wants the next instruction or the tool that performs it. A does-this-apply-to-me decision wants a quick test. None of them wants a sales conversation yet, and the one that is closest to it, the whether-decision, is the one most often treated as if it already were.

An example of the sentence doing its work: a post titled "How long does a site migration take?" has the obvious decision "should we start the migration this quarter?" The generic CTA says "Talk to our migration team". The decision says the reader needs a timeline they can put in front of their own stakeholders, so the next step is a planning template. A failure case is a post with two decisions in it, a comparison that ends in a how-to. It needs to be two posts, or one decision has to be chosen and the other demoted to a link, because a post with two next steps has none.

The reason a sales-call button fails on an explainer is not that readers dislike sales calls. It is that the post has not earned the ask. Katie Sherwin's Nielsen Norman Group article on commitment levels, published in March 2016, sets out a five-level trust pyramid in which each level must be satisfied before the one above it can be asked for. The levels run from baseline relevance and trust that needs can be met, through interest and preference over other options, trust with personal information, trust with sensitive or financial information, and finally willingness to commit to an ongoing relationship. The article's rule is direct: do not make demands at higher levels of commitment until you have addressed all the trust needs at the lower ones.

Map a blog post onto that pyramid and the right next step becomes obvious. A reader who has just learned what a term means has cleared level one at most. A reader who has finished a comparison and chosen an option is somewhere around level two. A contact form is a level-three ask, because it requests personal information, and a discovery call that implies a purchase is a level-four or level-five ask. The post's next step should sit no more than one level above where the reader is when they finish, which is why the inline checklist outperforms the booking link on almost every explainer and why the booking link belongs on the post whose reader has already run the self-check.

An example of matching the level: a diagnosis post ("why did my traffic drop") ends with a five-minute test the reader runs in their own Search Console, and only the paragraph after a positive result offers help. A failure case in the other direction: a buying guide whose reader has already chosen, ending with "read our introduction to the topic". That under-asks, and the reader who was ready to act is sent backwards. The pyramid cuts both ways.

The end of a post is where a template puts the call to action and where the fewest readers are. Readers who found their answer in section three leave in section three. So the next step belongs at the point in the post where the reader has what they need to decide, which is usually just after the comparison table, the decision rule or the diagnosis result, and it can be repeated once in the conclusion for the readers who read to the end. One step, placed at the decision and again at the close, is the whole placement rule.

Placement at the decision point changes what the step can be. An inline next step can be a sentence with a link, a short embedded checklist or a single button, and it reads as part of the argument rather than as an advertisement after it. The end-of-post box, by contrast, is visually separated from the content and is the element readers have learned to scroll past. NN/g's 2017 study makes a related point about prominence: a prominent generic button acts like a magnet and causes users to ignore more appropriate content nearby. The same thing happens when a loud contact box sits under a quiet but correct inline link.

A failure case worth naming: three different next steps in one post, one after each section, each pointing somewhere else. The reader is offered a download, then a newsletter, then a call, and the post reads as a sequence of asks. The audit sheet in section 06 has one proposed-step column per post for exactly this reason. If a section genuinely needs its own step, it is probably a post.

The wording of a next step is a link-label problem, and the best short treatment of link labels is Kate Moran's Nielsen Norman Group article from March 2019, which gives four qualities: specific, sincere, substantial and succinct. In her framing, a link's primary purpose is to tell users what they will find on the other side of a click; it must set expectations that are accurate as well as specific; it must be able to stand alone; and it should not waste words. "Get started" fails the first two on its own, which is the mechanism behind the 2017 findings.

Applied to a blog's next step, the rule is to name the object the reader receives and, where the step asks for anything, what it asks for. "Download the migration planning template (XLSX, no email required)" is specific, sincere about the cost, substantial enough to make sense out of context and still short. "Learn more" is none of those. "Book a call" is specific but, on an explainer, not sincere about where the reader is, because it implies a readiness the post has not created. The label and the commitment level are the same decision seen from two sides.

One example of a good label on each decision type, and the generic version it replaces, are in the table below. These are illustrative, written for this guide rather than taken from any site, and the point is the shape of the label rather than the exact words.

The whole method fits in a spreadsheet with six columns, and the spreadsheet is what makes it a day's work rather than a quarter's. For each post: the URL, the reader decision in one sentence, the current CTA as it appears on the page, the proposed next step, the commitment level of that step from the table in section 03, and the GA4 event that will measure it. The first and third columns can be filled by a script or an export. The second is the only column that needs an editor, and it is the column that decides the other three.

To get through an archive in a day, sort by topic cluster rather than by date, because posts in one cluster share a decision shape and usually share a next step. A cluster of twelve how-to posts on one platform will mostly want "the next guide" as the step, and the proposed-step column is filled twelve times with the same pattern and different targets. Then handle the exceptions: the post whose traffic is ten times its cluster's median, the comparison post that every other post links to, the diagnosis post that is the archive's real front door. Those get individual thought; the rest get the cluster's pattern.

One failure case to design out before you start: filling the proposed-step column with the thing you want rather than the thing the decision implies. The test in the right-hand card applies to every row. If the reader who just finished the post would not have asked for what the column proposes, the column is wrong, and the fix is usually to step down one commitment level.

A changed next step is a hypothesis, and the measurement that tests it is per post, not per site. Google Analytics 4 gives you most of the raw events without any extra code. Its enhanced measurement, as Google's help documents it, records a click event each time a user clicks a link that leads away from the current domain, a file_download event when a user clicks a link leading to a file with a common extension such as PDF or XLSX, a form_start event the first time a user interacts with a form in a session, and a form_submit event when the user submits a form. Scroll depth is recorded the first time a user reaches 90% of the page. Google notes that the form parameters appear in reports only after you create custom dimensions for them, a setup step that is easy to miss.

The step that turns an event into a measurement is marking it as a key event. Google's help defines a key event as an event that measures an action that is particularly important to the success of your business, and describes the mechanism as creating or identifying an event that measures the action and then marking it as a key event in the Admin area, with a counting method chosen at that point. A key event can then be reported per landing page, and it is the per-landing-page view that answers the audit's question: did this post's new next step get taken more often than the old one? Google also notes that a Google Ads conversion can be created from a key event, which matters if a step is later used in paid campaigns.

Two cautions. Internal links, the commonest next step on an explainer, are not an enhanced-measurement event, because the click event is defined as leaving the domain; measure them with a custom event on the link or by reading the next page's views filtered by referrer. And a key event that fires on every page, such as a site-wide newsletter form, cannot distinguish the post's step from the template's; name the post's step with its own event or parameter so the comparison is clean. Connecting Search Console and GA4 to one analysis, as in our GA4 and Search Console connector build, makes the per-post view routine.

Most of what the audit sheet finds is not a per-post problem. It is one template component, rendered on every post, carrying the same level-five ask regardless of what sits above it. That means the first fix is structural: the template's end-of-post box should accept a per-post next step and fall back, when none is set, to the lowest-commitment generic option, a link to the topic's hub or the newsletter, rather than to a sales call. With that in place, the audit sheet's proposed-step column becomes a field the editor fills, and the day's work becomes data entry plus the handful of posts that need their own paragraph.

The second fix is the inline position. A template cannot know where the decision point in a post is, so the inline step is authored, one sentence with a link or one short component, placed by the editor after the table or rule. On a mature archive this is the part that takes real time, and the cluster-first ordering from section 06 is what keeps it to a day: the high-traffic posts get an authored inline step; the long tail gets the template fallback and is revisited when it is refreshed. Pair this with the dating rule in our guide on when a post earns a new date: a changed CTA is not a substantive content change, so it moves no dates.

Then measure for a month, per post, against the key events chosen in section 07, and expect the result to be uneven. Some posts will show a next step taken far more often; some will show nothing because the post never helped anyone decide anything, which is a finding in its own right and feeds back into the refresh decision. Building the template fields, the event plan and the per-post reporting is the kind of work our content engine service sets up, so the audit becomes a standing practice rather than a one-off.