👈 Home

Where do ideas come from? (Part 4)

Earlier this week we talked about the mythical “range of all feasible options” where A/B test variations float in the ether, waiting to be tested. We looked at a trivial example of optimizing a button’s border radius. Today, let’s talk about something way messier: optimizing copy.

It was easy to generate the range of all border radius options. Start at 0, add a pixel at a time until the button’s as round as it can get. Copy’s not so simple, but we can still wrangle it.

The range of all feasible options: Homepage heading copy

Let’s take my current homepage heading copy as an example. It could probably do with some optimization:

What does the range of all feasible options look like for this heading?

If we limit ourselves to English, limit the length to, say, 80 characters, and require that the words make reference to CRO or marketing analytics, we’ve reduced the possibilities from infinite to probably just millions.

Try to form a nice visual image of this vast plane of hypothetical headings - scattered with all the possible ways of saying “I do data markety stuff” that all the copywriting monkeys on typewriters could ever produce.

Those headings are out there; we just need a way to find them, and figure out which ones to test. To do that, we’ll bring some structure to the mess.


The best way to organize iterations of a message is to lay them across a spectrum or three. Here are a some examples:

Terse vs wordy

The current heading is a little on the wordy side. If I wanted to try for something on the other end of the spectrum, I might go extreme:

Formal vs informal

As it is, my heading is fairly informal. (“Hey, I’m Brian.”) I could go further, though:

🤔 Not entirely sure what that means, but it’s definitely informal.

On the other end of the spectrum lie options like “I write, speak, teach, and practice data driven marketing” 😴


This dichotomy was introduced in ProActive Selling. It refers to whether the language encourages the reader to think of attaining positive outcomes or avoiding negative outcomes. My current heading is fairly neutral on this spectrum, but let’s try TOWARD:

A little over the top, but you get the idea. Here’s AWAY:

Pretty heavy handed, but hopefully it taps into something.

Here are a few more spectra you might consider using to sort (or generate) copy ideas:

Other sources

Several of the sources of ideas mentioned yesterday can help populate the range of options.

Competitor analysis

Don’t just steal their phrasing, but if their site conveys a message more clearly than yours, find a way to learn from it without plagiarizing.


The exact phrasing of your actual customers is a powerful thing. Do they use words that your website doesn’t? Wedge those words into a heading, by all means.

User testing

Depending on how qualified your test subjects are, they can bring the same benefit as surveys. But even if they’re not industry mavens, and are just uncovering basic usability issues, pay attention to what they say.

Ideas that come to you in a dream

Yes, seriously, if you dream about heading copy you should definitely do something about it. Running a test might be the healthiest option.

As with the button border radius example, we first define and organize the range of feasible options. Then we choose test variants from among them, making sure the variants are as different from each other as possible.

So make sure not all your options use insider language (try something more general). Don’t test only terse headings - try something wordier. Don’t only test language about product benefits, even if they came to you in a dream.

With these guidelines, you should be able to produce more testable copy than your website traffic can possibly handle. Feeling stuck? Hit Reply and let’s chat. (Or just try adding the 🚀 emoji, that always works.)

© 2022 Brian David Hall