swast

Fashion AI chatbot

helping users curate and finalise outfit ideas based on their preferences and saved data. aiming to accurately match what users were looking for, enhancing their decision-making process eventually saving more number of ideas.

solution:

why did we solve this?

drop offs at idea feed
drop offs without saving results
low number of follow up questions
driving initial adoption

what other problems arise?

navigation between questions

as the infinite feed is included - jumping from question to question was a challenging issue

solving for latency
generating answer with visual feed, the cost of accurate results was latency from the tech side - we decided to get few seconds using some design decisions

example problem shown here:

possible solutions tested with users:

data and insights collected from testing:

scroll is the first user instinct to navigate
a button for quick jumping between questions assisted with navigation better - this scoping also aligned with the tech team.

impact:

37.6% more saving of inspirations
users wanted written description of outfit inspirations, positive response on the infinite feed shown upfront.

78.96% users interacted with updating body type for accurate results
visually engaging, users loved seeing results according to their body type hence increasing the number of saves

23.7% increase in asking follow-up questions
users claimed satisfaction with the results recieved after the first query.

User Intent

Improving accuracy in visual results generated, while making the user make the full use of NLP and getting data for finding PMF.

Implement a data collection system (body shape, age, gender) to deliver personalized fashion recommendations, enhancing user trust and retention.

solution:

why did we solve for this?

giving realistic fashion inspiration
help users use NLP
solving for retention
solving for finding PMF

how did we solve this

used this interaction as a form for prompts
this design worked as a form in the backend filling in the blanks for the prompts. this not only scoped out the results but also gave idea about our target group and locating specific use cases.

A/B testing pop-up when logged in & widget to create body profile
small testing was included to see what made users interact more, try out different body types and explore with NLP.

visual cues for avoiding conflicting body shapes
visual experiments were done with the choices of body shapes to avoid confusion between different body types.

impact:

37.6% more saving of inspirations
users wanted written description of outfit inspirations, positive response on the infinite feed shown upfront.

78.96% users interacted with updating body type for accurate results
visually engaging, users loved seeing results according to their body type hence increasing the number of saves

23.7% increase in asking follow-up questions
users claimed satisfaction with the results recieved after the first query.