Feb, 2021 - May, 2022

Fleur

Fleur was an algorithmic art matching service that used a quiz to match users to artists who fit their personal preferences.

My Role

Strategic Planning, Product Management, Programming, Networking & Partnerships, Financial Management, Marketing

Platforms & Tools

Mailchimp, Google Cloud, Instagram, Facebook, Typeform, Python, Html, CSS, Figma, Excel

Outcomes

Fleur allowed us to test art discovery systems and gather valuable data. We pivoted to spring.art after we determined that the platform did not have the stickiness necessary to keep people engaging long enough to generate the large amount of data needed for improved algorithms.

Project Overview

Fleur was the first pivot in our startup that pushed us farther into software development and away from the traditional art business models. In the process of building Fleur's art matching platform I gained a better understanding of market segmentation, user personas, product development, and learned to code to build the front and back end for our web app.

Problem

Most people are underserved when it comes to discovering and buying art. Traditional art businesses have high friction, which keeps them small, local, and traditional, as a result they fail to connect to most buyers or serve most artists.

Solution

At Fleur we were working to build an algorithmic art matching system that could understand a user's interest and connect them directly with artists that fit their personal taste. We believed that by more effectively networking the supply and demand sides of the marketplace we could avoid the need for high-touch relationships and enable art transactions to occur naturally on a much larger scale.

Results

I built an MVP version of our web application which used a psychographic quiz to gather data about new users, we had manually assigned a number of tags with correlated values to the answers of each question, we then ran it through a simple arithmetic algorithm that I had developed to compare against a similar set of tags and values attached to each artist in our database. Users would then receive a weekly match to a new artist with our goal being ongoing matching once we had a full product and database of artists.

Though rudimentary this matching system generated positive feedback from our initial user group of roughly 200 people. More than 40% of users engaged with their new match each week and our user satisfaction scores hovered between 7 and 8 out of 10 throughout the course of the experiment.

Fleur validated our underlying assumption about the potential of algorithmic matching. However,

Reily Haag