Visualization goals

I am excited about the potential of using map and network visualizations to support organizers and workers to uncover new patterns and organizing strategies. Throughout the course, I have worked with different datasets to explore visualizing employment violations, concentrations of work, key employers, and supply chain relationships. For this final project, I wanted to expand an existing Tableau dashboard and bring together several visualizations related to migrant work in Ontario.

My goal was to map concentrations of migrant workers working in specific regions and for specific employers in Ontario, and the supply chain relationships between production and retail nodes for several top employers. In doing so, I am seeking to make visible the labor that underpins Ontario’s agricultural industry as well as potential leverage and power to support worker organizing.

Mapping migrant work

Migrant farmworkers and allies in Ontario have been organizing for many years to expose conditions under Canada’s seasonal migrant worker program and to win immigration status and legal protections. With farms scattered across rural areas, it can be challenging to locate and reach workers. While organizers have built up geographic knowledge through years of outreach, I wanted to create a map drawn from an official dataset of all farm locations. 

I used a Government of Canada dataset of employers with approved migrant worker positions (LMIAs) that includes name of employer, city, postal code, program type, industry code and number of positions granted. I cleaned up the dataset in Open Refine by separating the city, province and postal code into 3 columns, fixing faulty postal codes, and eliminating duplicates. 

Through a series of iterations throughout the class, I created a dashboard with 2 maps showing the concentrations of farmworkers by region, and by employer name in the agricultural stream. I also included charts showing the top regions and employers which can also be used as a tool to filter specific cities or employers. 

Following feedback from two organizers, a non-organizer, and Professor Sula, I made a series of design changes to improve legibility, reduce occlusion, and provide more context and guidance for interacting with the tool. I increased the map size and the size of the circles, and eliminated some non-important classifications of industry code. Ultimately, it wasn’t possible to totally prevent the occlusion that occurred with overlapping circles, but mapping by city instead of postal code and adjusting the circle size and opacity did make an improvement. I designed a “scrolly-telling” format with added context in the introduction and specific instructions on how to interact with the maps to guide the user.

Mapping networks of power

To further build out the dashboard, I wanted to layer in network visualizations for the top employers hiring migrant workers in this time period, building on an assignment I completed in our network lab.  For this dashboard, I wanted to create a simple network graphic which would show the major retailers sourcing from specific employers. While some of these employers have brands associated with their name (for example Highline Mushrooms), others do not. The degree to which the employer is associated with major retailers and in turn, the extent to which their products are prevalent in grocery aisles across North America can therefore be difficult to see. In this way, possible leverage workers may have to target retailers at the top of the supply chain is also opaque. 

I created network visualizations in Gephi for 3 employers based on internet research for each company. For simplicity, I wanted to only use nodes for suppliers and retailers. Through my research I found some retailers did have more importance in the supply chain and assigned them a higher weight (for example, Arterra Wines owns an entire retail chain where their wine products are sold). After uploading the node and edge tables, I used the Fruchterman Reingold layout, adjusting gravity and area to create a network graphic that could be clearly displayed in a larger dashboard. I sized edges by weight to show the higher importance of certain retailer nodes. 

Integrating multiple visualizations

I wanted my map, chart, and network visualizations to integrate within the dashboard so that clicking on one region or employer would filter information in each visualization. This would bring in some degree of interactivity for the user which was suggested by Professor Sula as feedback for my initial network design. I also wanted to create visual snapshots of the three employers with supply chain graphics which would appear when that employer was selected anywhere in the dashboard. Professor Sula shared how to create a custom shape in Tableau which could be filtered to another custom shape.

I was inspired by the Canners NY map and wanted to try and integrate some simple illustrations to add a design element to make the overall dashboard feel warmer. In a feedback session in class, another student suggested using an illustration of the primary product grown for each employer, an idea that I really liked. I created some simple illustrations with ink and watercolor and then refined them digitally. These illustrations became my custom shapes 1 and the network visualizations became my custom shapes 2. 

I had a lot of difficulty having two different shapes map onto the same mark (employer name). Tableau would revert to the first custom shape and erase the second one. After researching a fix through ChatGPT, I created calculated fields to ensure Tableau would not override either custom shape folder. I then had further challenges with the sizing of the custom shapes on the dashboard and ensuring they were full-sized (or invisible) when not selected. I had to create a work-around of making one sheet for each of my illustrations which stayed permanently on the dashboard, and having the custom shape 2 remain a small size until the relevant employer name was selected.

Finally, I had to try multiple solutions to keep the filters for the 3 employer names from resetting when the employer name was de-selected on other charts or maps. Once again, I used ChatGPT for assistance in creating another calculated field to override this glitch. I was able to get to a place where each visualization is integrated, but not with the final flow I initially had in mind.

Final reflections

I believe I achieved my goal of creating an interactive dashboard that brings together several types of visualizations to support organizers in exploring patterns and opportunities for organizing. Even the initial steps of identifying top employers and the retailers where the products flow has revealed possible focuses for targeted outreach, potential campaign targets, and where further supply chain research could be strategic. Additional testing by organizers will show how the functionality can be improved and if new patterns emerge.

Overall, I like the minimal design format of the dashboard. While I like the idea of including illustrations, they are a design element that is likely helpful more for a secondary wider public audience and not for my primary audience of organizers. I could improve their quality in a future round (and potentially integrate other illustrations in the dashboard). However, I also don’t like the way the font gets blurred when the custom shape image is selected. I had tried using a title for each sheet instead of embedded text in the illustration, but I didn’t want the title text to linger when the image was deselected. A future iteration might omit the illustrations and just have one custom shape image of the supply chain network. 

See final dashboard here