8 Steps to Build a Voice Skill That Aligns With Your Brand’s Vision

Unfortunately, Google may not have all the answers to your questions

Building a voice skill that accurately represents your brand can be tricky. Getty Images

If you’re a producer, designing a conversational experience for voice may lead you into uncharted territory. While we can’t fully protect you from a few sleepless nights of quality assurance testing, these eight pointers may help you get to where you need to go with this new and burgeoning marketing tool.

Start by identifying the consumer need your app will solve

Designing your voice experience strategy is all about understanding your users. Identify the need your app will solve for them. For the first iteration of your app, it’s best to assume that things will be at least twice as challenging as you expect. Select a user need that’s feasibly solvable in a one- or two-minute conversation. Articulate how solving those user needs will help you achieve your business objectives. This rationale will serve as your experience strategy and become the project’s north star.

Consult experts and observe this need in the wild

You need to find out if there are existing solutions for the problem you are trying to solve. If there’s an expert in helping people handle this need, say a mechanic or a psychologist, ask them how they do it. Go into the wild and listen carefully to conversations in order to learn how to recommend a cocktail, find a nice bar and listen to how the bartender imparts his or her mixologist magic. It’s also really helpful to start with existing script material. Do you have call center scripts? Start there and build tactics with which your voice experience will help the consumer solve their problem.

Don’t count on everything going smoothly— it won’t and it’s going to be rejected a bunch of times before it’s ready for the big show.

Nail down the happy path

From launching the experience to signing off, the happy path is the series of steps your user will take if everything goes according to plan. Map out what happens when a user activates your app and provides an ideal, predetermined response at every prompt. The happy path should be committed to code immediately, using the computer-generated voice at hand. This step is important, because submitting updates to Google and/or Amazon early and often will save you headaches deeper in the process.

Read your script out loud

Just like a table reading for a movie, eliminate technology entirely and read your voice script out loud among the team. Hearing it read makes it easier to catch hiccups that would otherwise be hard to identify.

If you can, hire a voice actor

Though the tech is rapidly improving, it’ll be years before a computer-generated voice feels truly natural. Deciding on intonation (like which word to emphasize in a sentence) requires computers to make assumptions based on context they don’t have. Using a voice actor adds a more human touch to your app, but also ups the cost and complexity considerably. Consider how many minutes of dialogue will need to be recorded. If the app is going to change in a few weeks or months, will the talent be able to return and record? Don’t jump into recording sessions until you’ve locked down and extensively stress-tested the anticipated flow of the app.

Plan for three rounds of audio recording

Round 1 involves recording the happy path voiceover. Round 2 will include corrections and the audio for your alternate user flows. Round 3 is where you’ll record any final corrections. You should schedule this as close to the end of production as possible, since swapping out audio can be done fairly quickly.

Submit chunks of audio early and often

Submitting updates to Google or Amazon early and often will save you headaches deeper in the process. When your app is rejected—and your app will be rejected—you can’t rely on the app reviewer to tell you why. Submit for approval with every major update to limit the scope of what’s changed and it’ll be easier to figure it out. As you incorporate alternate flows, keep submitting your updates to the virtual assistant platform for review. This will help ensure that as much of your app is approved as possible as you near your live date.

Do not skip quality assurance

Recruit as many people as possible for user testing to identify points of friction. You’ll use this critical feedback to identify bugs, iterate skill development and refine your design. Though you’ll be tempted to shorten or skip this phase—don’t. The assistant platform approvals process routinely misses even the most glaring problems, and skipping testing provides the opportunity for you to ship a product that’s riddled with them. Align with a QA team to test ease-of-use assessment and refinement tests before deployment.

Creating a voice app can be a pretty rad—and challenging—experience for a producer, because you’ll likely encounter problems you can’t turn to Google to answer. Don’t count on everything going smoothly— it won’t and it’s going to be rejected a bunch of times before it’s ready for the big show.

It’s going to take weeks and weeks of stress testing and rerecording, and hours and hours of QA. And if your experience goes anything like ours, there might even be some heightened emotions stirred by the intense work.

But fear not: You may end up creating something no one’s ever seen, or heard, before.

This story first appeared in the May 21, 2018, issue of Adweek magazine. Click here to subscribe.

Margaret Barnard is integrated producer at 360i.
@MikeMohammed Mike Mohammed is senior content strategist at 360i.
@bigtbigtbigt Tony Landa is senior technical director at 360i.
@sabu_91 Sabreen Jafry is project manager at 360i.