Voice — discovery / 2018–2019
Orange — smart speaker
New ways to test a voice product with no UI: Wizard of Oz before launch in Spain.
01
The business problem
In 2018, from Globant, I led an Orange brief that had to be resolved before designing anything: whether the company should launch its own smart speaker in Spain. Google and Amazon had just brought their voice assistants into the Spanish market, and Orange needed to know if a telecom operator had a place in that category before investing in building the product.
It was a business decision with very little evidence behind it. There was barely any data on how voice was used in Spanish homes. My job was to reduce that uncertainty until Orange could decide with data, and then defend whatever recommendation the data supported.
02
My role and decisions
The immediate problem was methodological. You can't run a usability test on a voice product that doesn't exist yet. There's no screen, no prototype to click through, nothing for a user to react to naturally. So before designing the product, we had to design a way to research it.
I led a team of three through two phases. The first was pure segmentation: sixteen user interviews to map the territory before touching any functionality. Who would actually use this? What was their existing relationship with voice assistants? What usage contexts made sense in a Spanish household? From that we identified three distinct user mindsets that shaped the entire product strategy going forward.
My role and decisions
“What I learned about how people talk to machines in 2018 still informs every conversational interface decision I make today.”
The second phase was harder to execute: validation through Wizard of Oz methodology. A team member operated the voice assistant invisibly while the user believed they were interacting with a real system. I designed the VA personality, the conversation flows, and the test scenarios directly from what we'd learned in phase one. We iterated across five blocks of four tests each, not chasing perfection but trying to understand where the interaction broke down and why.
03
Project achievements
A go / no-go discovery project in unmapped territory. No product to test, no UI to evaluate, so we built the research method before we built anything else.
Interviews
24
Round 1: refine the population segment. Round 2: understand their needs. Round 3: test how useful the device was for the defined features.
Screenless product testing
Wizard of Oz
There was no product to test, so we adapted Wizard of Oz: someone ran the assistant unseen. Five blocks of four tests, to see where the conversation broke.
Report conclusions
No-go
The final analysis of every tool pointed the same way: better to wait to launch. The market was not ready.
04
Outcome
The recommendation was no-go: the Spanish market was not ready, even though the technology was.
A strategic document outlining market perception, three validated target profiles, and the specific conditions under which a launch would make sense. Sometimes the most valuable design output is the product you advise against building.
What does your product need now?
A hard decision, a team, or a product that has to move. Let’s start there.
Contact: chema.a.job@gmail.com