🐟 Fishing for Correlation

The Problem

One of our clients, Blue Lobster, work with fishermen who use sustainable methods to catch fish. As they expanded the other side of their marketplace – restaurants who wanted to invest in a more sustainable fishing industry – they saw a need to grow their supply of fishermen.

Fishing sustainably means you are at the mercy of the elements - if the weather is bad, with strong wind and high wave heights, it's impossible for small vessels to go out. To mitigate this risk, the Blue Lobster team wanted to understand where in Denmark they should onboard new fishermen, to complement their existing partners, in order to optimise for continuous supply.

The Solution

We limited our search to the seas of Denmark, in order to reduce the travel time of transporting fish to restaurants in Copenhagen. Harbours in Denmark can be considered to have immediate access to 9 different seas, as shown below.

We extracted data for fish caught in the different seas from the fishing authorities. We filtered the data to only include small vessels, accurately modelling the types of vessel that Blue Lobster would consider working with.

Grouping catch volume by sea and by landing month, we plotted a correlation matrix of seas against each other.

We noticed an overall skew towards positive correlation coefficients - the green shades on the grid. This can be attributed to several things: self-correlation (always 1); seas being adjacent to each other; and the use of the landing month dimension, together with the seasonal nature of fishing.

Because the existing concentration of Blue Lobster fishermen were based in Sjælland, fishing in Øresund, examining the grid we identified the most complementary sea would be Nordlige Nordsø. This recommendation was reinforced by the concentration of relevant fishermen in those harbours, overall catch volume in the region, as well as possible future expansion to deliver fish to restaurants in nearby Aarhus.

Underlying factors in the amount of fish caught include season, weather conditions, fishing quotas, and when fishermen go on holiday! In future analyses we have delved deeper to start modelling the expected catch volume of a fisherman for a given week using a simple regression model.

We've loved working helping Blue Lobster - because of the impactful nature of their mission, the richness of data related to fishing and gastronomy, and the data-driven instincts of the founders.