Written by Jiwon Hong and Jessica Erdman, YesPlz.AI
How did my startup, YesPlz AI, build an AI-powered product filter from a set of scribbles?
With hard work, long hours, and a lot of prototypes–but if you’re reading this article, you probably know that. Our Style Filter wasn’t built overnight, and along the way, I learned valuable insights that I believe are unique to founders looking to build something unique.
The first musings of the YesPlz
And I mean totally unique–not a carbon copy of a popular app. Because when we built the Style Filter, we were (and still are) building a totally new piece of technology, combining elements of eCommerce visual search, product filtering, and artificial intelligence. Arguably, many of the lessons for founders are useful, but not applicable to those creating a completely new path for their fields.
The Style Filter is a new, innovative way for customers to search for clothes intuitively. By selecting their preferences for fit, style, and type of clothing, customers can search in a way that mirrors the in-store experience. We blend visual search and artificial intelligence to create the Style Filter.
But, with any new piece of technology, while exciting in theory, we encountered distinct challenges along the way.
Lesson 1: No matter how great your idea is, if your target user can’t figure out your app, you need to rethink the design.
User experience is at the core of what we do at YesPlz AI. While we love cool, new, and complex technologies, the end users of an eCommerce tool is a shopper–and that shopper can have varying degrees of knowledge in search terminology.
“User experience is everything. It always has been, but it’s undervalued and underinvested in. If you don’t know user-centered design, study it. Hire people who know it. Obsess over it. Live and breathe it. Get your whole company on board.”van Williams, Co-Founder, Twitter
When we first set out to build the Style Filter, we tested the design on users to get feedback. And, unsurprisingly, users were excited about the idea of using a new tool to find clothing, but were also confused.
There was a 1-minute learning curve to use the Style Filter, which of course, was not acceptable by any circumstances. We continued to gather feedback from users and test, and eventually we reduced the learning curve to less than 2 seconds.
Our current goal is to have a learning curve of 1 second or less.
The evolution of the Style Filter
As Evan Williams notes, we all should be obsessing over user experience. When an illustration looks just a little bit strange, it’s easy to dismiss it, or put time into other valuable development areas. While other advisors may tell you to make a barebones product, in my experience, a MVP still needs to include user-friendly design elements (or at least plans to make the product more usable).
We shouldn’t expect users to take on the burden of learning our products–and sometimes a “gut feeling” can produce more valuable insight than the insights of an expert in technology.
Lesson 2: Cross-industry knowledge is transferable, if you’re willing to validate the product with industry experts
I came from a technology background (a product manager at Samsung SmartTV, among other tech industry experience) and developed the idea for YesPlz AI from a simple user insight from my previous business as a Shopify merchant.
Users liked to judge whether or not they liked a product based on an image.
While obvious at the surface, this insight had many implications:
- users are far more visually-inclined to make decisions based on products
- text-based search currently offered on eCommerce websites was not in line with the way we know how users search.
With that insight, I began to build a prototype for the Style Filter.
But, as you may remember, I don’t have a background in fashion. Technology, yes. Conducting thousands of user interviews as a product manager–also yes. But, fashion is a specific industry with very specific terminology. I was afraid my lack of in-depth experience in fashion would keep the Style Filter from gaining traction.
And so, I knocked on doors to make sure I truly understood the problem. I spoke to 17 fashion retailers to validate the problems and challenges I had previously identified. But, without those conversations, the Style Filter would be a tool built by an industry-outsider.
Cross-industry knowledge is applicable, but as a founder, you need to have the humility to realize that you aren’t an expert in a new field–and there are individuals who have spent their entire careers in the new field that you’re building a product for. So, rather than convince industry experts of your credibility post-launch, why not start building relationships and validating products pre-launch?
Lesson 3: Remember all of the tasks you assigned to your intern? They’re your tasks now, in addition to building a product.
The glamour that comes with founding a startup is quickly lost when you realize that the small, menial tasks you used to be able to assign to your intern are your responsibility. And that’s why I believe that founding a startup isn’t for everyone–even if you have an amazing idea–because the level of “grunt” work necessary.
Remember those user interviews I mentioned above? I drove 3 hours to interview potential users in-person. And, I had to be “that” person at conferences, asking anyone relevant that I met if they’d be open to an interview.
The other memory I have is labeling products to train the artificial intelligence. I ended up with carpal tunnel syndrome from weeks and weeks of manually labeling data.
This isn’t meant to complain or discourage others from founding a startup–but just a word of advice to those who are ready to dive in: be ready to get your hands dirty!