The 3 Questions of Microcopy

Occasionally, I’ll run a lecture or workshop on writing microcopy. Sometimes this is for junior writers; other times, it’s for stakeholders who need to be able to plug in decent-enough words on their own.

Experience has taught me that, regardless of who’s doing the writing, three questions stand in the way of good copy. If you can confidently answer each of them, you’re at least 80% there.

Question 1: How does the feature work?

The rule I set up day one at any gig is this: I don’t write what I don’t know. What is it we’re building? Why are we building it? Who requested it? What problem does it solve?

Let’s say I’m building a nail salon booking flow. The answer for why is pretty obvious — so that customers can book an appointment without calling or dropping in — but even laying that out brings up some good points. Since we don’t want them to come in or call, we need to provide all the necessary information.

Sometimes the questions go deeper into how the logic works in the backend; I don’t necessarily need to know exactly how the gears turn (and wouldn’t understand most of it anyway) but I do need to understand what exactly happens at each step. Are we showing only available appointment slots? Does the user get to choose their technician and, if so, before or after choosing their service? Is the appointment confirmed automatically or manually?

Surprisingly, this step is a tempting one to skip. Or, at least, it’s one that PMs and designers will sometime dangle in front of you. Can you just take a quick look at this screen for me? Well, sure — if the end result of my review is to make sure the text is compliant with the content design system, yeah, a quick glance can accomplish that. But that same quick glance isn’t going to provide enough context to make sure the text is accurate and relevant.

💡Can AI help with this? Ehhh. If you multiple or complex sources of truth, plugging it all into a GPT can produce a decent summary. But 1) we remember more when we actually find the information ourselves, and 2) a summary is just a summary. AI may not have the same definition of “important” as you.

Question 2: What does the user need to know?

Congrats: You now know more than you ever wanted about your feature. But you’re not going to tell all that to the customer. Oh no, of course not.

Yesterday I told my husband that I’d be busy in the afternoon for a manicure. I didn’t tell him that I made the reservation online, or that the salon is a twenty minute walk away, or that last time I was there they offered me red wine and coffee. Is all of that true? Absolutely. Is it relevant? Does he care? Absolutely not.

The whole point of that conversation was for him to know that we shouldn’t make plans. That’s what affected him; everything else, while true and relevant to my experience, had no bearing on his day.

Run through the same mental checks with your users. What do they need to know? What do they want to know? What is actually important, and what has a direct effect on them? I’ve recently gotten into the habit of making a physical list of what needs to be communicated.

For example, a confirmation screen for a nail appointment needs to communicate:

1. That the appointment is indeed confirmed;

2. The service(s) booked;

3. The technician assigned to the booking; and

4. The time and date of the appointment

Only copy that hits those two points will be approved, no matter how good it sounds or how well it follows the style guide.

💡Can AI help with this? Surprise: It’s another Ehhh. AI is good at picking these out if you give it a *!#$ ton of context, but 1) you’ll probably naturally find these while doing your research in step one, and, again, 2) making the list yourself often gets you halfway to decent copy.

Question 3: How can we best communicate this?

Now — and only now — you can start thinking of crafting sentences. Now you can play with titles and subtitles and CTAs, now you can start to think about following the style guide, now you can wonder if it’s worth adding a moment of delight to the moment.

Since I now know what information has to be included, a first draft is actually fairly easy to draft:

• Title: Your appointment with {technician} is confirmed

• Subtitle: We look forward to seeing you at {time} on {date} for your {service}.

💡Can AI help with this? Say it with me: Ehhh. First of all: Why would you want it to? The actual writing is half the fun. But I’d also caution that the more you ask AI to write, the more your own creative muscles will atrophy. It’s not awful to use it for alternative ideas, but as backup tool, not a go-to.

Bonus Question: That’s 80%. What about the last 20?

Answering those 3 questions got us to decent copy. Is it winning any awards? No. (And not that microcopy should be flashy, anyway). But it works. 80% there is good enough.

If we want to be 100% happy with our copy, though, there are extra steps we can take. Part of it comes down to experience; maybe after working as a writer for several years, you know that a certain sentence can be tightened up.

Or maybe since you’ve actually done your research, you know that customers actually care more about the service than the technician. In that case, it would make sense to have the title include the former rather than the latter.

And maybe in user interviews, or while combing through the data, you realize that a decent amount of customers make a mistake and want a CTA that lets them make immediate edits. And since you’re really good at your job, you know that there’s a reason for so many people making a mistake, so you take a fresh look at the flow.

This isn’t to say that research or additional thought/experience isn’t necessary; it’s to say that if you at least ask yourself those three questions, you can get to the point of good enough copy. And with so many companies not having enough writers, or moving at breakneck speed, sometimes good enough is all you can provide.

💡Can AI help with this? Look, mate, stop asking. I know this article has been pretty negative on AI, but that’s where I am right now. In a recent chat with some writerly friends, we agreed that if not using AI means being left behind, we’re happy to stay put. It’s not a journey I want to take. Are there uses for it? Yeah. Does it take half the fun out of it? Absolutely.