Visa Climate Hackathon

In June 2024 I participated in Visa’s climate hackathon where I worked with a 500K+ row dataset containing anonymised UK card transactional data to come up with a sustainability focused analysis. For its ease of documentation and considering the project didn’t need to be production ready, I used a Jupyter Notebook to do this.

My approach to the hackathon was to identify fast fashion purchasing trends across clothing transactions across the UK which could then be used to inform targeted campaigns to combat against high emission actions.

Downloading the dataset revealed several challenges, particularly the significant amount of mismatched data in each column which wasn’t reflective of the other fields (e.g., amazon.co.uk under the city column). I cleaned this column by downloading an ONS list of UK cities to validate any entries. I also converted all entries to title case to ensure there would be a complete match between the two datasets. This process removed city tags from around 6800 rows of data but was necessary to ensure the analyses were reliable.

Now that the data had been cleaned, I worked to categorise clothing merchants into either fast fashion or non-fast fashion companies. Combining this with another new column for carbon footprints utilising gov.uk resources which map spending to merchant types (e.g., fast fashion, non-fast fashion, jewellery, automative, etc…). Whilst climate emissions naturally vary between companies as some have more efficient supply chains/sourcing options than others, utilising these coefficients as approximate estimators of carbon emissions would provide a good enough indicator of high emission consumers.

Integrating these carbon emissions coefficients to spending across each transaction produced a CO2e estimate figure for each transaction, highlighted in the screenshots at the bottom of the page.

Working with the data, I generated a few different visualisations. Firstly, I looked at the percentage of total fashion spend going to wards non-fast and fast fashion purchases. This identified that 41.6% of purchases in this category were fast fashion (including merchants like Shein, Primark, Boohoo, Missguided, and PrettyLittleThing).

From here, I utilised a pre-existing field provided by Visa which determined the ‘shopper type’ of each transaction, grouping consumers into four categories: urban spenders, affluent segment, frugal family, and sensible spenders. This offered an insight into how trends differed across these categories. The output of this can be viewed below.

Lastly, I wanted to explore how these purchasing rates varied across UK regions. I mapped this to a horizontal bar chart with gradient colouring to make it easier to interpret.

Together, these outputs provided clear insights suggesting that consumers based outside of London from the sample generally spent more on fast fashion than those spending in the city. Furthermore, areas like Liverpool and Birmingham saw the highest rates of fast fashion spending. Looking at shopper types, urban spenders saw the highest percentage of their fashion spend going towards fast-fashion, with the affluent segment predictably spending less than half of this rate on the merchant category. Frugal families were the second highest spending rate on fast fashion items which might align with the relatively cheaper cost of these goods compared to longer-lasting but more expensive alternatives.