Landing Page Primary Action: One Page, One Job

The cost of a second call to action is not the click it takes. It is the pause it creates.

Hick's law, from W. E. Hick's 1952 work on choice reaction time and Ray Hyman's 1953 follow-up, describes how the time to make a choice rises with the number of options presented. It was measured on simple, equally weighted alternatives, so applying it to a whole web page is a stretch: a page with strong visual ranking gets scanned, not deliberated. The honest version is narrower and still useful. When two elements look equally important and sit at the same point in the visitor's reading path, the visitor has to compare them, and comparison costs attention you already paid to get.

There is a second cost that nobody feels until reporting week. Two calls to action split the same population of visitors. Each one now converts a fraction of the traffic, so each looks mediocre, neither reaches a sample size you can act on, and the page owner ends up defending both because neither can be proven worse. A page with one action produces one number that moves.

The ratio people reach for here is what Unbounce popularized as "attention ratio": clickable destinations on the page divided by conversion goals. Treat it as a framing device rather than a law, since it counts a footer privacy link the same as a hero button. It is still the fastest way to notice that a page asking for one thing offers fifty-three ways to do something else.

Most "too many CTAs" advice fails because it counts buttons. The useful unit is distinct destinations, sorted into three roles.

One test separates a real secondary action from a disguised second primary: does it feed the same funnel? "Watch the two minute demo" placed under "Start a trial" is a secondary action, because a person who watches is one step closer to the primary. "Talk to sales" sitting beside "Start a trial" is not. It is a second primary wearing a smaller button, it forks the road, and the two options will quietly compete for the rest of the page's life.

An escape route is not a failure either. If the page cannot serve the person who landed on it, the honest outcome is that they find the page that can, or return to the results. Bolting on a capture form to intercept them converts a bad fit into a bad lead. The related read is pogo sticking, an industry coinage rather than a published search engine metric, describing that return trip to the results page.

The most common way to get this wrong is to choose the action from what the business needs this quarter. The traffic did not read the quarterly plan. It arrived on a query, and the query says what stage the visitor is at.

Search Console will tell you the mix in about five minutes.

1. Open Google Search Console and go to Performance, then Search results.
2. Set the date range to Last 3 months. A 28 day window is too thin for a query breakdown on most pages.
3. Open the Pages tab, find the URL you are auditing, and click it. That applies a page filter to the whole report.
4. Switch to the Queries tab with Clicks, Impressions, CTR and Position all enabled.
5. Export to a sheet.
6. Sort by clicks, not impressions. Impressions include people who never came. Clicks are the visitors your page actually has to serve.
7. Bucket the top fifty queries by intent, then sum clicks per bucket.

Here is a worked read on an example page. The figures below are illustrative, not SparkCliks data.

A page like that, with roughly half its clicks arriving on definitional queries, is being asked to teach. A "Buy now" hero button on it is not a primary action, it is a mismatch, and the page will read as pushy to three quarters of its own audience. The defensible primary action is the next step a learner would take, and the commercial ask belongs on the page you send them to next.

Do this before you touch the design. Choosing a primary action that contradicts the query is the one mistake no amount of button styling repairs, and the mismatch usually starts before the click, in the snippet. If your title promises one thing and the page asks for another, the visitor notices in the first second: see matching your title tag to the landing page headline and satisfying search intent above the fold.

A button label is a promise about the next screen. Vague labels ("Submit", "Learn more", "Get started") push the work of predicting what happens onto the visitor, and a visitor who cannot predict the next screen either hesitates or clicks and bounces. A click followed by an immediate exit is worse than no click, because it burns the one moment of intent you had.

Two rules keep labels honest. First, one verb per page: if the primary action says "Start", nothing else on the page should say "Start", because a repeated verb on a different destination reads as the same offer. Second, the label and the destination heading should share words. If the button says "See all seven plan tiers" and the next page is headed "Pricing", rename the heading.

Repetition and multiplication get confused constantly. Placing the same button five times down a long page is repetition, and the page still has one job. Placing five different buttons is multiplication, and the page now has five jobs.

The count that matters is distinct destinations, so a page with one destination repeated at four scroll depths still scores one.

Keep the label and the href identical at every repeat. Changing the wording between repeats splits your click data across labels, so no single destination looks dominant in reports, and it reads to the visitor as a slightly different offer each time. A sticky bar follows the same rule: one sticky action, the same one, or none at all.

Trust marks belong beside these repeats rather than in a badge row at the bottom, because the moment of doubt happens next to the button. There is more on that in trust signal placement.

Count the clickable destinations on your own page and the number is usually startling. A typical marketing template ships with six to eight primary nav links, two or three utility links, a search field, an in-body link every few paragraphs, and a footer carrying forty or more. That is fifty-plus destinations set against one.

The instinct is to cut the footer. That is mostly the wrong cut, and here is the part standard conversion advice skips: distraction cost depends on position in the reading path, not on raw count.

A footer link sits after the decision point. By the time somebody reaches it, they have already scrolled past the final call to action and declined it, so the footer is catching people who were leaving anyway. A link two paragraphs above the primary button is a different animal. It intercepts exactly the reader who had built up enough interest to keep scrolling, at the moment before the ask.

Popups deserve their own line. An interstitial that fires before the primary action has been seen replaces your one job with a different one, chosen by a timer. If you run them at all, gate them behind scroll depth or exit intent rather than a five second delay: popups and interstitials after a search click.

Paid search playbooks say remove the navigation from landing pages, and for paid search they are right. That traffic arrives on one ad, with one promise, at a known cost per click, so every escape route is money leaving.

Organic landing pages are not the same animal, and copying the paid advice onto them is a common and expensive mistake. An organic page collects queries you did not choose. Some of those visitors want something the page does not cover, and navigation is how they find the page that does. Take it away and their only remaining option is the back button.

The last row matters more each quarter. Visitors sent by an assistant often land on a page that answers part of a longer question, and a page with no way out either satisfies them completely or loses them completely.

Before the symptom list, a setup detail that stops most of these audits dead. GA4's enhanced measurement fires a click event for outbound links and a file_download event for documents. It does not automatically track clicks on internal links or on buttons. So the competition you are trying to measure, your own nav against your own hero button, is invisible in a default install.

To make it visible: add a Google Tag Manager trigger on all link clicks, send a GA4 event carrying link_text and link_url, then register both as custom dimensions under Admin, then Custom definitions. Give it 24 hours, then build a Free form exploration in Explore with link_text as the row dimension, Event count as the metric, and a filter on the landing page path.

Those percentages are illustrative thresholds for a first pass, not benchmarks. Read them against your own page's history rather than against a number from a blog post.

The fourth row saves the most wasted effort. A page can have a perfect primary action and still convert badly because the form behind it asks for eleven fields. Work out which side of the click the problem sits on before redesigning anything: form friction without losing fields covers the after-click half.

Run this on one real page that already receives organic clicks. Do not start with the homepage.

Step 1. Write the job before you open the page. One sentence, in a doc: "When this page works, the visitor ______." If two people on the team write different sentences, stop here, because you have found the actual problem.

Step 2. Count destinations. Paste this in the browser console on the page. It lists every clickable thing so you can count rather than estimate.

// Lists every clickable destination on the page, in DOM order.
[...document.querySelectorAll('a[href], button')]
  .map(el => (el.innerText || el.getAttribute('aria-label') || '').trim())
  .filter(Boolean)
  .forEach((label, i) => console.log(i + 1, label));

Step 3. The five second test. Show the first viewport to somebody outside the team for five seconds, then ask what the page wants them to do. If they name two things, you have a tie.

Step 4. The grayscale test. Screenshot the page and desaturate it. Anything that still reads as a button competes with your primary. In the first viewport there should be exactly one.

Step 5. The verb and destination pass. List every CTA label on the page in a column with its href beside it. Count distinct hrefs. Count distinct verbs. Both should be one for the primary.

Step 6. The adjacency check. For each repeat of the primary CTA, read the two paragraphs directly above it. Any link there gets removed or moved below the button.

Step 7. The query check. Run the Search Console export from earlier and confirm the primary action matches the dominant intent bucket.

Step 8. Write one change. One change per page per measurement cycle, or you will never know which one worked.

Score the page against this list. Anything you cannot answer yes to is a finding.

A one job rebuild usually changes several things at once, so the measurement has to be stricter than the change.

Baseline window. Twenty-eight days before the change, which covers four complete weekly cycles. Weekly cycles matter here because business pages behave differently at weekends, and a 21 day window weights the days unevenly.

Change window. Twenty-eight days after, starting the day after deployment, not the day of it.

Control set. Three to five comparable pages on the same template with similar traffic, which you deliberately do not touch. Read their conversion rate over both windows. If the controls moved by a similar amount, something seasonal moved them and your change is not the cause. Skipping this step is how teams end up crediting a button color for a demand shift.

Metric. Conversions divided by sessions from organic search only. Total sessions mixes in email and paid, which carry different intent and different volumes week to week.

What counts as signal. On a page with a few hundred organic sessions a week, weekly conversion rate wobbles by a surprising amount on its own. As a working rule rather than a statistical test, treat a move smaller than about a fifth of the baseline rate as noise, and require the direction to hold for three consecutive weeks with the control set staying flat. If you have the volume for a proper split test, run one instead, because everything above is a substitute for the test you cannot afford.

One thing not to do: never validate a call to action with purchased visits. SparkCliks' SERP Clicks visits search for your keyword, scroll the results, click your listing, stay on the page for around a minute, optionally open one internal page you nominate, and never press back. That is exactly what those visits are configured to do, and none of it is an opinion about your button. Keep purchased visits in a separate segment in reporting so they do not dilute the organic conversion rate you are reading. The one genuine benefit for this exercise is the second-page setting, since naming the single page you want visitors to reach forces you to say out loud what the primary action is.

The rule is strong, not universal.

Homepages are not landing pages. A homepage routes several audiences and its job is orientation. Judge it on how quickly people reach the right section, not on a single conversion.

Catalog and category pages. Here "reach a product" is the action, so a hundred product links are the primary action expressed a hundred times, not a hundred competitors. The real competitors on those pages are the banners.

Documentation. The primary action is often "keep reading" or "copy this snippet", which is real, measurable, and not a button.

Two audiences on one URL. A page serving recruiters and buyers usually wants to be two pages. If it cannot be split, use one explicit fork high on the page with two clearly labeled paths, and accept a lower ceiling on both.

Regulated pages. Disclosure and consent links are obligations, not distractions. Style them low, keep them, and do not count them in your budget.

Everywhere else the discipline holds. Name the job, pick the action from the query, say it in the language of the outcome, repeat it, and delete whatever else was standing beside it.