What drives clicks and retention

Four forces explain a visitor’s behavior: a clear promise, a visible button, little effort and trust. They are tested with a hypothesis: if, then, because.

From “it seems” to “let’s test”

You have already done a lot for the site: chosen the words, built the pages, invented the flows. Now you have to answer honestly what of it works. Without measuring, every argument about a site turns into a question of taste: “it seems to me the button should be bigger”. This part teaches measuring, and this lesson is about what is worth measuring and changing at all: how a person behaves on a page, where their attention ends, why they click or leave.

In the part on texts you saw how people read from a screen (they don’t read, they scan), and in the part on copywriting you met Fogg’s model: an action happens when motivation, ability (how simple it is) and a prompt (a button you can press) coincide. Here we turn that into assumptions that can be tested.

What is known about attention

One of the biggest dangers of analytics is that numbers give a feeling of knowing. So below is only what rests on research, with a note of what kind.

• The first seconds decide. An NN/g article drawing on research by Microsoft Research, which analyzed page-visit durations for about 205,000 pages and more than two billion dwell times, showed that the chance of leaving is highest in the first ten seconds, and that those who stay beyond thirty seconds often stay for minutes. The average time on a page is a little under a minute, but it hides a wide spread: some leave at once, some stay long. The conclusion: show the page’s value within ten seconds.
• Attention falls down the page. In 2018 NN/g measured that people spend about 57% of their viewing time on the first screen and about 74% on the first two. For comparison, in 2010 the first screen got 80%: over the years people have grown more willing to scroll, but they still look mostly at the start.
• A person doesn’t read it all. By an earlier NN/g study (2008), in an average visit a person has time to read at most 28% of the words on a page, and 20% is more likely, so they need to be given what the eyes can find: headings, lists, emphasis (remember the lesson on scanning).

Open the diagram: it shows these shares on an invented page. Move the call to action between screens.

Two warnings. These are averages over the pages that were studied. On your site, with your people, the picture may differ, and you’ll see it in the part on heatmaps. And attention is not decision: a person who is ready will find the button at the bottom too. That is why a button is often put both at the start and at the end.

What drives clicks

If you gather what you already know, several things affect a click. They are also the places where you look for hypotheses:

1. A clear promise. In ten seconds a person must understand what the page is and why they need it (the heading, the four U’s, the inverted pyramid).
2. A visible and understandable button. It can be seen (hierarchy: size, contrast, place), and the text on it says what will happen (“Book a consultation”, not “Submit”).
3. Effort. Every form field, every extra step is a cost. Ability, in Fogg’s terms.
4. Trust. Next to the call is what removes fear: a review, a document, a price, clear terms.
5. Expectation and the “scent of information”. A person arrived by a link with a promise and looks for its continuation. If there is no trail, they leave. Researchers call this “information scent” (information foraging theory; a link is in the sources).
6. Speed and ease on a phone. A page that loads slowly or is awkward for a finger loses people before the first glance.

What drives return

Return (retention) for a site with a one-off service works differently than for an app: a person doesn’t come back to the site every day, they come back when they need it, or stay in touch (subscribed, follow a channel). What affects it:

• Value on the first visit. If a person got an answer, they will remember where they got it (the lead magnet and the value ladder).
• A prompt to return: an email, a post, a reminder (the flows of the first part). In the Hooked model it is the “external trigger” you read about in the lesson on influence.
• No irritants. Every pushy email and pop-up lowers the chance of returning.
• A rhythm you can keep. Consistency matters more than frequency.

A hypothesis: how to write one

A hypothesis is an assumption that can be tested and disproved. A good hypothesis looks like this:

If we make (one specific change), then (a measure) will rise (or fall), because (a reason about the person).

And with it, always: what we’ll test it by (which measure) and whether the traffic is enough.

Examples of bad and good wording:

• Bad: “Improve the home page”. It doesn’t say what to change or what should change.
• Bad: “Make the button red, because red converts best”. The reason is made up and isn’t about your person.
• Good: “If we move the ‘Book a consultation’ button onto the first screen, then the share who click will rise, because a person who has already decided won’t have to look for it”.

What distinguishes a hypothesis from an order: the “because” can be disproved. If the button is raised and the share didn’t grow, the reason was wrong, and you learned something new about people.

Honest about traffic. On a page that forty people open a month, the difference between 10% and 12% is four people and five: such a difference proves nothing. So estimate for each hypothesis in advance whether there are enough visitors. If not, don’t pretend to be doing science: test with bigger changes, by comparing months and by talking to people (the lesson on A/B tests covers this).

Practice

Write three hypotheses for your project.

1. Each in the template “If… then… because…” with one specific change.
2. At least one is about the main action from the brief, at least one about a section of the site from the brief.
3. For each, a measure for the test and the number of visitors a month as a number.

The check looks at form: three hypotheses, a word of change in the “then”, a reason, a measure, numbers. Whether your hypotheses are clever is for you to decide: read each as if you were going to fail it, and ask yourself what you’ll learn about people if it isn’t confirmed.