Zya aimed to bring legit-sounding music creation to the masses. It was initially built for PCs, then was moved to mobile devices.

Anecdotally, people of all ages found the product's abilities exciting, groundbreaking, and creative. However, we weren't getting much retention. Marketing could drive people to download the app, but they tended to leave and not return. This also made it difficult to monetize.

Executives had defined the target market very broadly, noting that everyone from eight year olds to eighty year olds enjoyed the demos they did.

I was the sole UX designer out of about 50 people at the company at that time. The first thing I wanted to do was understand our audience more specifically. App reviews indicated excitement mixed with confusion and disappointment. I quickly trained a colleague so we could do double duty on observational interviews. We did rounds of in depth interviews with varying ages, genders, purchase power, and musical ability.

I reviewed the various interview videos and analyzed them using a rainbow grid. It's my favorite method for interpreting a lot of qualitative input. With it, I'm able to look for patterns easily, and weight opinions fairly. It's useful for trying different color coding, and sorting a variety of ways (e.g. look at males versus females, or positive versus negative reactions). I also worked with Marketing and ran surveys.

"Feels polished," reported one tester. Actually, we got top marks from almost everyone for ease of use and visual appeal. There were issues though. The conversations kind of went like this:

 A few things came through clearly. First, it didn't seem to matter whether people had money or not: they didn't want to pay much for music. Apple had set the price expectation for a song to $1, and this was a challenge; Zya had negotiated difficult licensing deals with music publishers that made it possible for people to use individual instrument tracks from radio songs in their own creations.

Next, even people with singing skills were too shy to sing alone into a phone in public. We required a vocal to generate music. One of Zya's most impressive features was the ability to correct timing, pitch, and quality, then compose around that vocal. Very young users enjoyed the voice effects, but older ones were very uncomfortable. Even the word 'vocal' felt imposing.

Third, there was confusion over whether this was a game or not. Zya was based on some flawed assumptions. The app had been designed as a fun creation tool, but segmentation played a role: musicians preferred more serious tools, and non-musicians needed more reason to use it.

Executives decided it was up to Marketing to educate users about the value of the special abilities our songs offered. Meanwhile, I suggested leaning into a game structure, so the main activity loop would include progression and rewards, one of which could be music. If people could earn songs, they might value them differently.

As a game, we could address people's motivations. Non-musicians had no intrinsic motivation to create songs, so I suggested giving them extrinsic ones: scaffolding, leveling up, badge and asset collection, daily rewards, spectacular completion animations, and so on. (see on YouTube). I did wireframes (Adobe Illustrator) and interactive prototypes (JustinMind) as low fidelity ways to get feedback and buy-in. Gameplay designers were in charge of fleshing out these features.

I suggested developing text-to-singing technology, or at least making vocals optional. This was significant, because song tech had been built to rely on vocals being recorded first, so a song could be magically composed around them. We ended up adding pre-recorded vocal options, and making it possible to record vocals after the music had been generated (people were somewhat more comfortable singing to music).

I was able to do usability testing every couple sprints and got a lot of feedback that we were usually able to address during each subsequent sprint. The app reviews ended up showing positive improvements.

We added and iterated. Each new game feature encouraged players to invest more time, and allowed Marketing to send meaningful notifications that brought players back. With each improvement, we improved retention by days, weeks, and ultimately months.

There were plans to add online competitions as a way to tie up the experience, add social challenges, and get song creators feedback on their creations.

Text-to-singing (remember that?) was developed on some of the server tech, and a new, simpler app named Ditty was released. TMZ uses Ditty on television and online. (Example on Twitter).

There were about 80 pages in the design document I prepared for engineers.

Interested in a high level Zya case study? Check this out.