I was honored to represent Escalent this March at the 2019 Wayne State University Big Data & Business Analytics Summit, sharing a case study about the development of our DataDialogue™|Park concept.
Some of the best and brightest analytics minds from metro Detroit, and beyond, came together for two days of sharing and learning around Big Data at the WSU Summit. It’s amazing to see how the conversation around Big Data has transformed in even just the few short years I’ve been attending this event. We’ve come so far from asking: “Is Big Data the hope or hype?”
Our breakout session took the DataDialogue|Park content from our award-winning Geotab Data Challenge entry – a concept to help commercial truck drivers easily find safe and convenient parking – and applied a twist. I presented the development process through the lens of a business framework for data-driven problem solving.
We all go through this process daily, whether we realize it or not. We identify and define problems – whether at work, in our personal lives, or watching and reading the news. We all know there’s plenty of problems in the news…but I digress. The key is to focus narrowly on a business problem, and understand if the problem is worth investing time and money in:
It didn’t require much digging to recognize the importance of the commercial truck parking problem, with statistics everywhere. In July 2018, I moved my mother back from Arizona to Michigan in a U-Haul. I lost track of how many tractor trailers were parked along entrance and exit ramps, or even alongside the interstates! Perhaps most enlightening for me is government involvement. In 2012, the federal government enacted Jason’s Law in honor of a driver who was shot and killed in 2009, after parking in an unsafe location. The law mandates the U.S. Department of Transportation to regularly review truck parking.
The value of addressing this challenge is extensive. Not only is time spent looking for parking a non-revenue activity, it could actually cost the company with potential hours of service (HoS) violations and fines. Furthermore, a solution can help increase driver retention – which is a valuable ROI in an age where the driver shortage is publicized regularly.
To address this challenge requires a concept developed through the lens of providing the driver with actionable information to find safe and convenient parking, whether by search function or automated notification. We did this by creating the Parking Favorability Score, which presents the driver with a rank-order of recommended parking locations.
Statistics vary, however, the majority of time solving data problems is spent in data discovery. The term discovery is a bit of a misnomer, as it encompasses so much more than just finding the data to use. Finding the data is just the tip of the iceberg. Anyone who has worked with data knows there is no such thing as that “perfect” dataset.
For us, this part was easy, relatively-speaking. We were provided access to publicly available datasets as part of the 2018 Geotab Data Challenge. The concept was a one-time use case, so we downloaded the data we needed to fit our concept. Then, we went to work transforming and aggregating the raw data into our Parking Favorability Score.
In most cases, the analyst has to scour many different sources, some even behind pay walls. That said, sources are proliferating by the day with indexed contents like this one on GitHub, or searchable sets like data.gov or Google. Once the source is identified, one can simply download the data as we did, or opt for a more sustainable approach with eyes to productization. Doing so requires coding an application programming interface (API) – in essence, a pipeline between desired endpoints that signals to pull selected data from the source to your ecosystem on a defined cadence.
Lastly, the answer is yes. Yes, in nearly all cases, you will need to transform your data in some fashion or another to make it useful. In this space, I will “plead the fifth” on the specifics of what and how we took raw data to create the Parking Favorability Score. There’s no short path to explaining it. Email me if you would like to discuss or learn more.
If you have made it this far in the problem-solving process, you are further along than most data initiatives. Pat yourself on the back for that. But your work is not done. Now, you need to determine what analytical approaches are needed, as well as how to deliver the analytics to solve the problem. At least for us data-and-analytics-nerds, this part is fun. As a best practice, you’ll want to embrace an iterative, scalable development process for the analytics deliverable. Think: “Good. Better. Best.”
The biggest challenge here is embracing, and ultimately staying true to, an iterative development mindset. This is essentially a distilled version of design thinking: creation of a concept MVP guided by understanding target user pain points, test with users, iterate and refine, further test with users, iterate and refine, etc. This approach helps lead you to a customer-centric solution.
For the MVP, we assumed that when a search function around a ZIP code is engaged by a driver, they are intent on finding parking. We assumed a limited search radius (ex: 10 miles) and provided scored parking in that area. This works similar to Google Maps when your GPS is turned off. For “good,” we went a step further, integrating GPS-based data so we know where the driver is and provide parking in a radius around that location. This is like Google Maps when your phone location is turned on. A “better” solution was to layer on HoS data – knowing what time a driver needs to park by allows us to conservatively estimate the location and recommend parking for that area. Finally, the “best” approach is to integrate GPS, HoS, and fleet dispatch data – knowing the final destination for the driver, we can predict if they reach their destination before or after their HoS clock runs out, and recommend parking appropriately for the situation.
There are always opportunities to improve concepts as more data becomes available to the system, or new approaches are envisioned. For example, there is incremental data we could consider like new parking locations, updated parking infrastructure data, or even driver feedback on parking areas. Additionally, there is expanded functionality that can be enabled, such as the ability to customize the analysis radius preference or, going a step further, providing push notifications as opposed to only search-based results.
If you’d like to know more about our DataDialogue|Park Parking Favorability Score concept or want to talk about this data-driven problem-solving framework, email me now.