Comparion's utilizes Natural Language Processing (NLP) for comparing product reviews. Two key techniques are involved: Topic modeling and Sentiment classification.

For example, I chose to utilize a screener survey during the discovery phase to be efficient - achieve two tasks at one go. This approach would validate the hypothesis that shoppers abandon shopping for spec-driven, similarly appearing products due to an overwhelming number of choices, and it would also screen for potential interviewees who could provide deeper insights into their shopping pain points.

Another example is I opted to begin the design process with mid-fidelity designs created using a UI kit, as this approach enabled usability testers to better envision the online shopping experience and provide more relevant and helpful feedback rather than the hypothetical feedback that might come out of testing sketches or wireframes.

Then, they were unable to answer an important question used to validate the hypothesis about overwhelming product choices leading to shopping abandonment.

In retrospect, I should have broadened the criteria to include any online shopping experience with spec-heavy, similar-looking products for my hypothesis validation, while still screening for stand mixer online shoppers for the interview phase.

Despite my error in the survey structure, "Overwhelmed by the range of products" emerged as the second most common reason for not adding items to the cart and abandoning online shopping. The top reason, "Waiting for bigger discount/sale," is more of a pricing and promotion issue rather than a UX issue. Although the sample size for this question was smaller than desired, due to time constraints and struggles in finding more survey respondents, I proceeded with the project by selecting interview participants who identified overwhelm as a reason for abandoning shopping.

I created user flows to help me gain a clear understanding of the steps and interactions Comparion users would go through when using the application. It allows me to visualize the entire user journey, from the initial interaction to the final goal.

Additionally, considering the time constraint, this exercise served as validation to ensure the coherence and effectiveness of the steps incorporated in addressing the identified "How Might We" challenges.

Leveraging social networks, Comparion enables users to connect with friends who are also using the app. This allows users to discover products compared by their friends and access their reviews, addressing trust issues with branded ecommerce websites' comparisons and reviews.

With research insights and user flows serving as a solid foundation, I began designing a solution for our stand mixer online shoppers, using a free UI kit found online to design a mid-fidelity prototype fast and realistic enough for testing. I then tested with 5 different online shoppers who had experience with stand mixers.

A major issue discovered in testing that helped me iterate from mid- to high-fidelity was the visual disconnect between the compared products and their compared info.

In the second round of testing with another 5 stand mixer online shoppers, a key learning emerged: shoppers are quite privacy conscious and hesitate to give social connections data to apps. Many testers did not see the concept of social proof through trusted friends on the app as that much of a value-add.

Friends’ product comparison activities and more were shown on Explore screen, as well as friends’ product review snippets on the Compare screen.

I should have focused on a narrower set of features and prioritized their execution to ensure a higher level of polish and smoother user flow. Next time, I would opt for designing fewer elements but with greater attention to detail.

Instead of limiting survey participants to those who specifically purchased stand mixers, I should have opened it to anyone who has shopped for home appliances online. This would have alleviated difficulties in acquiring survey respondents and potential interviewees.