Brighton Kamen

the project: use ibm watson technology in a 2-day hackathon

Interns at IBM tend to jump into problem-solving head first. As part of my internship at IBM's Mass Lab in Littleton, MA, I took part in IBM's annual Bluehack, where interns are given 24 hours over the course of three days to develop a working prototype and business model for a new product. Along with five software development interns, I helped create Styl.us which lets users find clothing to match what they already have in their closet. 

Homepage Thumbnail Credit: Zippypixels.com.

the product: styl.us

Three days and not much sleep later, our team entered the Video Comprehension challenge and built Styl.us. Styl.us is an API that uses IBM Watson Visual Recognition to match products (in this case, clothing) to each other. Users can take a photo of an item in their closet, upload it to the app, and the visual recognition technology will use that item to create an outfit. From there, users can browse looks, see details on products, and even purchase directly from the app. 

Click an image below to see an app screen. 

the process: everything happened on day 2

With barely 24 hours total, our team really had to focus on our minimum viable product rather than making sure everything was perfect. I dove right into user research to better target who would most likely use this app and how that would drive the design. Although our final product was the API, our actual deliverable would be a downloadable app. I used Adobe Illustrator and Sketch to create the screens, while also researching similar apps to pinpoint their successes and failures. 

In the US, the majority of clothing shopping tends to be done by women, but how people shop has changed, especially in the digital age. Modern shoppers (like Jenny, our main persona below) make use of mobile devices in stores. Retailers are currently focused on connecting with their consumer base through digital means, and a successful shopping app will help bridge that gap. 

Once I had a general idea of the target market, I consulted an IBM Offering Manager, Sam Bobo, on how to develop a business plan and estimate revenue. Estimating revenue is not an exact science, but with a little more research I came with a plausible (albeit optimistic) business plan. While the app's end users would be consumers themselves, the clients we would sell Styl.us to were the retailers. As a team, we came up with three main ways to monetize Styl.us:

• Users pay for the app or we offer it as a "freemium." We decided not to go with this option because users tend to not like spending money, especially when apps are usually offered for free. We also weren't sure how many users would actually pay for more advanced features. 
• Advertisers pay to have their products featured first or more often in the Browse category. This option would be useful if we were only building the app, but we are offering the API as an IBM product, and IBM typically doesn't sell advertising as a business model. 
• IBM sells the technology we developed so retailers can integrate the API with their own apps. The Styl.us app is only a deliverable, but stores like Macy's or H&M could use the technology in their own apps to interact with users in the store. Retailers would also be able to collect data on what users like and what outfits they tend to search. 

Ultimately, we went with the third option, as IBM tends to sell services to businesses, not directly to consumers. Styl.us would be offered as part of IBM's Global Business services and could be included in a package.

the progression: from design to programming to final demo

While I worked on the design and business plan, my teammate Andrew set up a server. When the user uploads a picture, the picture is sent to the server, where it is analyzed through Watson visual recognition. Watson uses classifiers to determine what type of clothing is in the picture and then my teammate Jordan's suggestion engine offers outfit matches. On Day 1, we were not able to build an app on mobile, so Dan and Clark configured an Android to support the app. By Day 3, we were unable to connect back end and front end, but were able to show the app in our video. 

the critique: what went well, what went wrong, and what i learned

As this was my first hackathon, it took a minute to get into the swing of rapidly designing, executing, and creating a business plan for a completely new product. Thankfully, my talented team of developers and I were able to pull off Styl.us (or, at least, get very, very close to pull it off). Here is what I would tell pre-hackathon me if I could go back to the first day:

• Details take time, so focus on the MVP. With so little time, it was virtually impossible to create a perfect app on the first pass, so I had to focus on what features were most important. The design also came out more minimal than I would have intended, but it would be a great project to revisit someday.
• Fake it til you make it, but you can always come back. Unfortunately, as we were unable to connect front end and back end in the time allotted, we did not have a fully functional product. However, I showed footage of each half working in the video to prove that with a little more time, it would work completely. A week later, we reassembled as a team, connected front end and back end, and were able to make Styl.us fully functional.
• It's not just what you sell, but how you sell it. One fourth of the score from the judges was how sustainable our hackathon idea would be as a product, so we had to prove its worth through a business plan. Andrew and I consulted with an offering manager and I later used that to formulate the business plan. 

In the end, it was all worth it, because we won Second Place!