Team Thomas & Friends
The Tool
Our team has developed a generalised geographic information system (GIS) that processes a range of geographic data and synthesises them into a single scoring system, covering the user’s area of interest.
A key guiding principle behind our design is that the user understands what could be important to them. Our tool is highly customisable, allowing users to both input their own data, and test how using the data changes outcomes in different levels of importance.
As a proof of concept, we applied our tool to the Roundabout challenge sponsored by Quantium. Roundabout is a charity which understands what could be important to them. This tool will allow them to test how robust their decision making is for different scenarios.
However, we believe that it is general enough to generate useful, actionable insights on electric vehicles (Appendix 2), as well as inform broader decision-making about future trends such as the ongoing cost-of-living crisis (Appendix 1).
Roundabout
To help Roundabout uncover where potential future donors are likely to live, we used fine-grained (SA2) ABS datasets across SES decile, fertility rate, and population density. We also have the capability to integrate web-scraped data on relevant sales trends from an Australian e-marketplace, but are awaiting permission to use it. Our choices of data and weighting are justified in the data story below.
With this data, we developed a heatmap focusing on NSW (although our datasets can all be trivially scaled nationally). Our purpose was to provide Roundabout with areas of high potential for any expansion into NSW, while leaving the fine-grained decisions around e.g., specific donation drop-off points to their expertise.
With our best-input weightings, the most demographically promising areas in NSW for Roundabout to expand to are:
This data is granular down to the community level, representing populations in the high 1000’s to low 10,000’s.
We are, however, acutely aware of our subject-matter limitations. As such, we invite Roundabout and other charities to use our portal (homepage, listed below) to adjust which factors they weigh, and their relative importance, to best inform their work. Future data sets can also be incorporated into the model as well quickly, enabling more precise decision-making.
In particular, one promising workflow we foresee is “running the system backwards”. Charities like Roundabout often already have data on the area they operate. By adjusting relative importance of factors until our heatmap represents their real-world data, Roundabout can analyse the relationship between demographics and donations. Subsequently, they can extrapolate that to determine currently-unserved areas, accurately predicting supply.
Our model is also very general, and can be reapplied to other use-cases such as the electric vehicle challenge. In addition, a second model we have developed can also be added to our tool in future to account for existing charity locations and competitors.
Appendix 1: Cost of Living
With an ongoing cost-of-living crisis hitting families across Australia, charities like Roundabout will be impacted with regard to both supply (with fewer families financially able to donate) and demand (more families needing support). Our model as it currently exists focuses on supply, but we believe that with slightly more time it could easily be extended to make predictions about demand, such as areas likely to be in increased need of donations.
On the supply side, our tool can be used to analyse the likely impact of the cost-of-living.
Our model focuses on the impact of the crisis on middle-income families - many of whom have previously been able to donate, but are being hit hard financially. This focus comes from a behavioural change framework: unlike low-income households (who already had limited financial ability to donate items rather than keep or sell them) and high-income households (likely to remain in a position to donate), middle-income households are disproportionately likely to shift from “able to donate” to “unable to donate”.
We modelled this by changing the “shape” of how SES deciles impacted donation, to treat middle-income households closer to low-income households in their donation patterns, and studying the change this made to the final scores.
The results of this analysis suggested that the largest decreases in donation ability across NSW are likely to occur in:
Appendix 2: Electric Vehicles
While we didn’t get time to work on electric vehicles, given the datasets available, we are able to provide a solution to the problem of where to install electric vehicle chargers. Factors for considering where electric vehicle chargers should be installed would include:
Locations of high road traffic
Locations of common parking places
Population density & growth forecasts
Zoning (e.g., residential v commercial)
Socio-economic status (SES) data
These two factors could have been collected through open-source street maps and speed camera data. In addition, we could also account for the location of other electric vehicle chargers using the competitive model we have developed. The socio-economic status data we used could also be used to obtain the best locations for converting petrol car-users into electric cars users.
Subsequently, using real-life examples of electric vehicle take-up, we can adjust our model to most accurately represent real-life needs. Thus, our model can very easily be applied to electric vehicles and other challenges as well, though we have only had time to demonstrate the child donor use-case.
We have used ABS Statistical Area Level 2 indices of birth rates in 2019, total residents in 2021 and the decile of index of relative socioeconomic status to determine the fertility, wealth and population factors that we have put into our model.
We also assessed the possibility of integrating publicly available (scraped) data from Gumtree to determine the commercial activity associated with baby products; this was put on hold while we await legal and ethical clarification around web-scraping the Gumtree site.
We normalised this data between 0 and 1, and “shaped” it according to the likely impact on donation patterns. Population density and fertility rate were assumed to correlate linearly with donations, as they imply more children in a given area ageing out of their clothes, toys, and other goods in any given year. The relationship between socio-economic status and donation was modelled as approximately a bell curve, as a result of consultation with subject-matter experts throughout the hackathon.
We then combined this data to obtain a score associated with each location, then outputted the score onto a map showing the relative strength of each SA2 location for future charity outreach.
Description of Datasets
ABS data
Population
Births
Digital boundary files | Australian Bureau of Statistics
Socio-Economic Indexes for Areas (SEIFA), Australia, 2021 | Australian Bureau of Statistics
(Potential but unused): scraped data from Gumtree - baby-related items for sale for <$20, sorted geographically.
Other
(Potential but unused): scraped data from Gumtree - baby-related items for sale for <$20, sorted geographically.
Description of Use Used to output coordinates for display in the final heatmap. Also used in a competitor model to account for existing charity locations.
Description of Use Values of birth rates in 2019, total residents in 2021, decile of index of relative socioeconomic status were normalized to 1 through min-max normalization. These were then passed on into the algorithm.
Go to Challenge | 10 teams have entered this challenge.
Go to Challenge | 11 teams have entered this challenge.