The problem
PGi's flagship product was GlobalMeet, and by 2017 its meeting room looked its age. New sales were hard to win in a category full of products with modern interfaces, and there wasn't much public trust in ours. The company mandated a full refresh.
Then it hesitated. GlobalMeet's existing customers were loyal, and executives were afraid that changing an interface those customers had already learned would push them somewhere else. The head of product held that position hardest. His view was that the product worked, so there was nothing to fix, and he didn't think design belonged in a decision like that.
Those are two claims that can't both be settled by argument. One says the interface is costing us new business. The other says the change is the risk. No data existed that could decide between them, and I had two months to produce a redesign concept.
Design management gave the project to me because I hadn't worked on GlobalMeet before. They wanted someone with no stake in the decisions that had made the interface what it was.
What I did
Measured the product I was being asked not to change
Before I drew anything, I tested the meeting room PGi was already selling.
The tests ran unmoderated through usertesting.com. Each participant got a persona and a set of tasks: you're a manager at a mid-sized company, hosting a meeting for your colleagues. Then they worked through what people actually do in a meeting: turning the microphone and webcam on and off, adding a participant, muting someone, chatting, running a poll, sharing content. I wrote every script, ran every test, and scored every recorded session myself.
I scored each task three ways. A success meant the participant finished quickly and confidently. Friction meant they finished, but hesitated or took longer than the task warranted. A failure meant they did it wrong or gave up.
The shipping product came back at 61% success, 14% friction, 25% failure. One task in four failed outright.
The obvious approach was to skip all of that and present a redesign. That's how most refreshes get sold. It wouldn't have worked on this one. Better-looking screens are an opinion, and an executive who doesn't think design belongs in the decision can decline an opinion without owing anyone a reason. Measuring first cost me weeks out of a two-month timeline, and those were weeks I'd rather have spent designing. It bought a claim he couldn't answer with a preference.
Tested four competitors to find where nobody was winning
Then I ran the same protocol against Zoom, WebEx, GoToMeeting, and UberConference.
I did that for two reasons. The first was the argument, since a number about your own product only proves it's imperfect and every product is. The second mattered more to the design. I wanted to see where the competition was strong so I could stop losing there, and where it was weak so I could find something to win on.
GoToMeeting scored best of the four, at 68% success, 17% friction, 15% failure. Nobody in the category was good. That changed what the project was for. Catching up to a competitor's interface is a defensive goal with a ceiling on it, and the benchmark said the ceiling was low. Winning the category outright on usability was available instead, because none of them had claimed it.
Set the passing grade before I ran a concept
I fixed the definition of success before any concept existed: 95% success, 3% friction, 2% failure. I set those from published benchmarks for what counts as a usable product, since a perfect score isn't a realistic target for any interface.
Fixing it early was the part that mattered. A target you set after you've seen a result is a target you moved, and the executives I'd be presenting to would have been right to say so.
Let the failures choose the next round
Round one tested 12 concepts across 135 tests. The best came back at 78% success, 11% friction, 11% failure. That beat the shipping product by 17 points and beat every competitor, and it still missed the target by a wide margin.
I could have presented it there. It was already the strongest score in the category. I didn't, because a result that clears the competition and misses the bar tells you the bar was the useful one.
So I aggregated friction and failure by task and let the clusters pick the work. Three came out on top: changing the meeting type, the participant panel, and the microphone and webcam controls. Round two tested six concepts across 130 tests and returned 93% success, 3% friction, 4% failure. A final pass on the mic and webcam controls, on My Files, Q&A and polling, and on returning to the home view produced 94% success, 5% friction, 1% failure.
One constraint shaped all of it. The redesign had to keep feature parity with the commercial product, so I couldn't ask whether a feature still deserved to exist. Every gain had to come from arrangement, labeling, and what the interface disclosed when. That's harder than removal, and it's also why the result didn't read to customers as a different product.
What changed
The head of product changed his position. I walked him through the baseline, then the competitors, then the concept scores, in that order, and he told me he couldn't argue with it. He'd been protecting something real. A customer base had stayed with GlobalMeet through years of better-looking competition, and the risk of changing what those people had learned was his to carry. The measurements gave him a version of that risk he could weigh instead of one he had to feel.
The redesign shipped in 2018.
Against the product it replaced, success went from 61% to 94%, and outright failure went from 25% to 1%. Against the best-scoring product in the category, it finished 26 points ahead on success.
Customer response was positive. One longtime GlobalMeet user, in a recorded session, put the whole project in one sentence:
"I like that it is more simplified, clean, not as boxy as GlobalMeet. It looks newer even though it has the same feature sets. More simplified. Very Apple-like."
Modern enough to change the impression, familiar enough that nothing had been taken away.
The work outlived the meeting room. Many of the interface decisions from this redesign became the basis for PGi's first design system, which I built next. That let the rest of the product ecosystem get refreshed in line with GlobalMeet instead of one screen at a time.
I didn't have access to adoption, retention, or sales figures after launch, so I can't tell you what the redesign did to the business. The usability numbers are what I measured, and they're all I'll claim.
What I'd do differently
I'd have tested something interactive. Every concept was validated on static images, so participants answered from what a single screen showed them, and nothing time-based got measured. A meeting room gets judged on joining, on a connection dropping, on finding the mute control while thirty people wait. My protocol couldn't reach any of that. I proved the interface was legible and left its hardest minutes untested. For round two I'd have built a clickable prototype, even a rough one, to better capture an authentic meeting experience.
