Overview
AdTrek is an application that was developed on Melbourne Datathon 2018 by Mathemagicians. The application aims to provide better data prediction visualisation for public transport advertising. The team came up with the idea by using the design sprint method. Turns out, it gave us time for further discussion regarding the MVP. And it came out as the winner of Data2App category for all-comers team.
Here is a demo of AdTrek:
Melbourne Datathon 2018
The datathon was conducted by Data Science Melbourne. The event was sponsored by several big companies such as ANZ, KPMG, iSelect, etc. There are two types of competition, the Insight Competition and the Data2App Competition (create an application that derived from the data). Both types based on the public transportation data that exclusively provided by PTV.  
The Matemagicians
Aditia L Ernawan
aditia.lukman@gmail.com
Rohan Nadagouda
rohan.nadagouda@gmail.com
Ronen Becker
ronenbecker@gmail.com
Septi Rito Tombe
ritotombe@gmail.com
Vlad Fridkin (Lead/Manager)
vlad@datastars.com.au

We were comprised of mixed professionals and students in data science. However, my main part was the web designer and app development. I contributed in designing the UI of AdTrek and in the stages of the ideation. 
Ideation
For the first team meeting, Vlad and I discussed which approach we will take to develop the data2app application. Eventually, we used the design sprint by guidance from Google with several considerations:
- We were comprised of diverse background, from professional to students, various domains such as health, marketing, and government. 
- Short development period with little time commitments from most of the members, therefore, we set initial ideas, and vote on them to set a goal.
- Previous experience from the team members.
Here is the draft of our first meeting. 

There were three ideas presented in the meeting.
- CityZen (Application to find best place to live based on the public transport data)
- An application to provide accurate advertising for public transportation ads companies.
- An application to assist disabilities.
source: Vlad's blog
We eventually chose the second option as it had the highest vote. However, we also discussed the impact of the idea socially and economically.
Prototypes
To put all team members on the same page, I immediately created two prototypes in one night to be voted and tested in the next meeting whether it is feasible given the time and technical constraints. The prototypes were created using Adobe XD.
I started from our common understanding from the first meeting:
- Target User: Advertising company (eg. Adshel)
- Competitive Advantage: 5 years of public transportation data directly from PTV, therefore, we believe we can provide more accurate estimation.
- Value Proposition: Information of public transportation advertising given the demography and audience estimation.
Recommedation Tool
The first option was a recommendation tool based on natural language processing. The idea was simple, type-in a search term (eg. "student fast food at CBD"), it will automatically search fast-food restaurants at the CBD with the highest student estimation at public transportation stops around it. It will also show the peak hours, estimation of cost and total audience.
Mapping Tool
The second option was a mapping tool. The idea was also simple; it shows the estimation of the audience given the filter provided by the user. 
Eventually, we agreed on using the second option because the first option will take more time to develop and we were unsure whether it is technically feasible given the data we have. Again, we had time and technical constraints. Therefore, I redesigned the tool with the recommendation tool as a "nice to have" feature. (In this stage we were starting to develop the tool)
Now, the recommendation tool is at the filter pane, and the filter pane is moved to the left side as it contains a lot of information (ref: Gutenberg Diagram). But even so, we need something that is feasible to be built in a short period with the highest value. We decided that the recommendations can be left off and replace it with demographic filters because we it will be more valuable to the user. 
This is the final prototype we created. However, turns out the filter pane became so busy, it can cause action paralysis (rule of thumb, max 5-7 elements in a group). Therefore, in the highlight 1, I highlighted the main filters background with soft blue. This way, the user can be more focused in that area. In the highlight 2, I realised that the button should be less contrast because the user could misunderstand the button as a call to action, it should be a mere simple text "Reset Filter". 
After several iteration of design and development, we came up with the final result as shown in the overview.
Result
The tool won the first place for the Data2App category for all-comers team. Well done Mathemagicians! Vlad wrote the detail about the awarding night in his blog.
Back to Top