Multi-modal System Design

Principal Designer — Bose

Three product categories. Wearable audio, portable out-loud audio, and in-home — speakers and soundbars. Each with its own form factors, its own input types, its own feedback mechanisms, its own reasons to be picked up in the first place. Each scaling up and down by price point, which multiplies the whole thing again.

That was the surface our team was responsible for. And it was only going to get bigger.

The failure mode in a portfolio that size isn't bad design. It's re-decision. Every new product arrives and somebody has to work out, from scratch, what a long press means here. Whether this surface should take a swipe. What confirmation ought to sound like. The decisions get made well, individually, by good people — and they don't add up to anything, because nobody ever wrote down what was already true.

So while we had the window, we took it. I worked with a small team — an industrial designer and a couple of other UX and UI designers — and we planted a stake: the principles and patterns we had already earned through research, testing, and customer feedback, written down at the point we'd actually reached.

Not to constrain anyone. To spare them. So the next designer starting the next product wouldn't feel like they had to invent the language before they were allowed to speak it.

Three things it was meant to buy. Efficiency for the designers, who could stop re-solving solved problems. Consistency across a portfolio about to get considerably wider. And confidence for the person who bought the thing — because when you buy a Bose product, it should act and react like one.

That was the case, and we didn't get to make it only to other designers. Research and the business — program and product owners — sponsored the work and sat as the stakeholders we read out to at every checkpoint and at the end. Which meant every principle had to survive an audience with no particular affection for design vocabulary. If a rule couldn't be defended to someone accountable for a program, in terms that made the cost of ignoring it plain, it wasn't a rule yet.

A core area of focus was gesture. We defined what makes a gesture intuitive and low-effort — favoring single-vector, short-duration interactions over ones with multiple twists and turns, and grounding new gestures in metaphors people already carry, like a swipe to dismiss or a sliding scale for volume. Skeuomorphic reference to previously learned behavior wasn't a compromise. It was the on-ramp.

We also wrote for the inputs that don't have names yet — head nods, gaze detection, conversational cues — pointing at a future where interaction feels less like operating a device and more like an exchange. And throughout, the governing idea was restraint. Fewer gestures per surface. Clearer hierarchy. A vocabulary that transfers across the ecosystem instead of flattering a single device.

The document lived in the design studio, with the intent that it would become a living digital reference — updated as the technology moved, as the industry's patterns shifted, as the portfolio grew. Principles like these are never finished. They're maintained.

The hardware that occasioned these particular rules has long since been replaced. The rules haven't. That was the point of writing them down.

When a user buys a Bose product, let’s make sure it acts, and reacts like one.

Over time, out User Interfaces have evolved from computer-readable, to human-readable, making interactions more natural, and implicit.

Gestures that are short in duration, and focused in are most easily accessible and executable — for instance, single-vector gestures are more easily afforded than gestures with multiple twists and turns.

Consider how you can use familiar metaphors to support gestures that feel natural to the user, like a flick or a swipe away from the user to exit something, or confirm you want it out of your focus.

Avoid supporting too many gestures on a single surface, or multiple surfaces in close proximity. Doing so will decrease the risk user confusion, and false positives. Prioritize primary controls and think about what other modalities may be best suited for secondary controls.

Affording a set of common gestures across our ecosystem allows our users the ability reapply their knowledge to new product, which builds their confidence in themselves, and in Bose.

When considering additions to the set of gestures, consider how that gesture could be applied to other ecosystem, rather than a single product.

Consider the positive in skeuomorphic reference to previously learned behavior. Although a 1-to-1 solution may not apply, we can use nuances of a previous interface to help a user make sense of a new interaction method.

As examples, users already use sliders to adjust volume or swipe away at things they don’t like, so creating gestures that follow a sliding scale for volume, or a flick to dismiss build upon those familiar metaphors.

Human-based gestures take advantage of actions a user typically performs in conversation. For example, a head nod is a common action that has obvious connotations around acceptance. Gaze detection can also be employed to determine focus and attention.

Multi-modal System Design

Let’s Connect