Learn how Flügger increased sales by 13% in less than a year leveraging AI to describe their colors.
Preface
By definition, in a technical world, a color is a collection of mathematical values that represent a color in a certain way. However, it's not how we humans experience colors. To people, colors are what add life to the world. They evoke emotions, feelings, and stories.
During the lifetime of Malgodt.dk, the team behind it (Rasmus, Lars & Kresten) did an amazing job bringing out a lot of "those numbers" into easy-to-discover color pages. It may have been one of the reasons the boutique paint shop was so successful. The idea was transferred to Flügger, and when ChatGPT took the world by storm, it was time to see what could be achieved.
Problem
Flügger Group A/S has a range of colors from different color charts to collaborations and themes. With the release of new webshops, it was time to help their beloved painters and future customers discover the emotions delivered through colors that they actually want to own.
Solution
This is where I saw huge potential to leverage large language models (LLMs) to rewrite the mathematical properties into natural language descriptions for some of our most desired and purchased colors. Together with Kresten Kjær Larsen and Lene Hessilt Kjaer, we created a versatile prompt that produced detailed, multilingual descriptions for each color. This small piece of text would turn into a valuable addition next to our color images.
How was it done?
We took confidential and private data from Flügger Labs. This data included information about how bright each color is, the number of pigments used, its coverage on walls, its way of tinting, etc. We then created natural language-based mappings to this color tinting data, effectively describing the colors without overtly stating these technical details.
To further protect our data, we had a zero-data retention agreement with Azure to query GPT-4 models.
This data, paired with the customized AI prompt, was sent to the most powerful model available at the time of the project, GPT-4. The model to this day proves to be one of the best out there.
Results
13.29% increase in sales revenue just after as little as 11 months.
Context
- The first color pages were introduced end of Spring 2023. It gave extra generous quarter for them to become widely discoverable on SERPs.
- The statistics are taken from September 2023 to July 2024 (11 months). The statistics may contain in accuracies due the nature of limited tracking in certain cases to comply with GDPR. It has been normalized to lower ranges.
- The data is based on the net revenue for the webshops in flugger.dk, flugger.se, flugger.no and flugger.pl. Adjusted respectively to the local currencies.
Wins
- It's clear to see that the color pages have had a positive impact on revenue.
- For the company, it's the return on investment that counts and allows for data-driven decisions.
- For the customers, it's the extra layer of information that helps them discover colors, especially in the challenging online search where a lot of spammy results tend to appear.
Comment by the expert
The distance from the technical to the emotional narrative of a colour is an artform in itself, which the usage of LLM is a good start of bridging.
Peter Møllebjerg Andersen
CDO, CIO & CMO at Flügger Group A/S
What does it mean for you?
Every company, whether a small startup like Malgodt.dk or a well-established player like Flügger, can leverage the power of AI to enhance their business.
There's often a lot of valuable data locked away behind difficult-to-interpret values. Personally, this is where I find myself extra passionate about the power of AI. With the correct engineering, patience in testing, and small enough tasks, we can allow ourselves to gain entirely new perspectives. That was the great promise of the internet back in 1993, and today, we have the technology to make it even better.
Credits
It was a great pleasure and honor working with the team at Flügger. Should you have any questions to my former colleagues, here are the contact details:
At the time of development, I was employed by Flügger. However, I was not sponsored to write this article nor did I receive any compensation for it. It is analyzed and written to the best of my ability and in good faith as an independent developer with a passion for AI applied software.