WM - Internal Tools Suite
Employee-facing tools for route planners, dispatchers, truck drivers, and corporate employees utilizing our Internal Tools design system
Mission: Modernize WM’s suite of internal tools and integrate new machine-learning technology to improve efficiency
Impact: 88% approval in employee software ratings, $3.2 million in cost reduction via improved route efficiency
Introduction
WM’s Internal Tools product portfolio is large and varied; it includes products aimed at employees ranging from corporate employees accessing career tools to truck drivers mid-route checking on the status of a pickup. Designing for so many different end users via so many different interfaces while maintaining a consistent design language and accessibility has been no small feat. From complex and data-rich tables used by dispatchers to the very Design System itself used across all of these products, I had a great time figuring out every way to improve efficiency, bridge functionality, and learn from every of the many types of users that my work would make its way to in production.
Though I began on a team of 3, I became the sole product designer responsible for WM’s Internal Tools suite by the beginning of 2022. Below I’ll outline some of my most impactful work, including the dispatchers’ dedicated tool for planning and routing drivers (Resource Center), the active tool used to track routes already in progress (Gantt for Dispatchers), and the tablet application used by truck drivers themselves (WM Driver Assist),
RESOURCE CENTER
Starting from (near) zero
WM’s Resource Center application is an integral part of the company’s internal automation suite of software. This software takes in data from multiple APIs with regards to the status of drivers, trucks, and routes and how they all comingle and can be best arranged for efficiency.
Resource Center was born at a time when WM did not have a robust team of designers. As such, the application was originated by combining out-of-the-box software blocks (primarily from Microsoft) that could route the correct APIs into data tables and display those tables to the route dispatchers.
Below you can see a screenshot of what that original application looked like, and why it so desperately needed an upgrade. The application was out of date, slow, had an exceptionally high learning curve, and lacked any real consideration of intuitive layout or construction based on the needs of the dispatchers. The Resource Center project was created to replace this application with a more modern and usable version that would help dispatchers improve their own efficiency in the act of improving efficiency on routes. How about that?
I came onto the Resource Center project as one of my first undertakings at WM around halfway through it’s redesign and helped see it over the finish line. It is now in production and I am making weekly iterations based on feedback from the dispatchers.
Here’s a screenshot of the old version:
The user and their needs
Resource Center is an application that was created specifically for vehicle and driver dispatchers early on in the days of modernizing WM’s internal systems. This tool, through a system of complex tables pulling live data from multiple internal bases, aims to project the current status of each driver, vehicle, and route (as well as their various attributes) onto one central system, where the dispatcher can then match an available driver to an available vehicle and then send that pair to a route needing service. For a seemingly simple task, this system is quite complex.
Because dispatchers are already trained in the needed technical terminology, my goal was not always to make this system more accessible to a layman user. Instead, the main feedback I received from them in testing revolved around improving proximity of related elements and tightening up the speed of completing frequent flows. I met with the dispatcher team monthly to understand where the need was greatest in these areas.
The Resource Center application also makes use of the Internal Tools design system, one which I co-created and then maintained on my own for the duration of my time working on the project. Another challenge was maximizing the efficiency of components to serve the needs of the dispatchers in their daily tasks.
Below you can see some of the feedback from focus sessions that we synthesized into proposed updates, which became the ‘Unassigned Assets’ overlay, new left nav, and Sites Bar respectively:
The challenge of modernizing the Resource Center centered around making complex data tables feel clean and simple without sacrificing power and functionality.
RC through our design system
In order to improve efficiency and usability in our new version of Resource Center, we built it around our new design system’s Foundations and Atoms elements to create new navigational elements, data tables, and button clusters. You can visit my Design Systems project page for a closer look at how some of these design system elements came together to create a new, modern Resource Center design that improved readability, accessibility, and usability. Below, you can see an excerpt of some of those new Molecule level elements that match the functionality of the previous version, all while using newer, more accessible atomic components and utilizing a more efficient layout based on focus sessions with the dispatchers.
Simplified Prototype
During my time working on the Resource Center I updated or created over 500 screens. Wow! Below I’ve created a simplified flow that showcases all of the major pages from the navigation bar, as well as a few core functions. Please feel free to ask me about other applications of the system, I’d be happy to geek out, er, elaborate on them.
Impact of Resource Center 2.0
After implementing the new version of Resource Center with out internal dev team, dispatchers reported needing a bit of time to adjust to the new systems, but overall found the program to be much easier to navigate. Key points I took away from the relayed research feedback included that daily tasks happened much faster, resources were much easier to find, and the linkages between different data sets became much clearer with the new simplified UI treatment.
After the planning team got the hang of the software, it improved efficiency so much that the business team was able to introduce a new initiative to optimize the efficiency of routes a full day in advance, something which was previously quite challenging because of the inaccuracy and slow sync rates of data tables in the prior software version. This resulted in the creation of the Next-Day Optimization (NDO) project, a new subset of our internal software focused on future route planning which I went on to begin design work for and integrated into this new version of Resource Center as well.
In subsequent updates that I’ve helped design for, we have integrated new functionality for auxiliary data like broker drivers, temporary site reassignments, and inactive driver logging. Because I’ve been able to integrate these features into the new core version of Resource Center, they can function much more efficiently for route planners and don’t require additional software or extensive retraining.
GANTT TOOL
We’ll do it live!
Where Resource Center is used by route planners at WM to lay out how future routes will take place, our companion software (referred to simply as our GANTT Tool) is used by dispatchers to track active routes as they occur, monitoring for delays, errors, and total successful tickets in a day’s planned service.
There are a few more complicated factors that happen with designing for a program that tracks live data. I worked very closely with the development team to understand where we had opportunities to lean out the program and improve latency, keep data fresher, and simplify the visual depiction of a route’s status. This was a different beast than Resource Center, which did need to keep data up-to-date but at longer intervals and with lower stakes around stale data.
The benefit of the GANTT Tool is it’s simplicity of purpose. Any changes that needed to be made to routes could happen in the Resource Center, so for this tool the goal was simply to provide the best possible tracking experience.
When I came onto the GANTT project a few months into planning and initial designs, GANTT felt like my most daunting project. Unlike RC, GANTT did not have equivalent predecessor software that we could build off of. However, using our new and improved design system, we were able to turn around a new version that matched the new Resource Center and gave the dispatchers all the data they needed at a glance.
A still frame from GANTT’s main live view dashboard
The new GANTT Tool needed to seamlessly show live-updating data about driver and truck locations, ticket status, and estimated efficiency scores on a daily basis.
Making the building blocks
Equipped with our new design system, we began to lay out what a live version of the data rows we used in the Resource Center would look like. In terms of the type of data displayed, GANTT was not too different from RC. However, and especially per the requests of dispatchers, we needed a better visual way to show tickets and their time slots that didn’t just use start- and end-timestamps.
This led to the creation of the timeline within each GANTT row. This timeline shows all of the driver’s tickets for the day with icon labels and color coding, and uses a lollipop-icon slider to track where exactly within the context of each ticket block the driver is at any time. We referred to many different sources when thinking about this setup, including ride sharing apps and food delivery services where live tracking is integral to the service experience.
After integrating the timeline for each driver into the standardized row, we used a similar approach to Resource Center to build out the remainder of the table background with the appropriate column headers and time markers for the ticket timelines.
Filling in the gaps
In order to fit several rows of data together, we began to consolidate text wherever possible and try to move to more simple icon representations to save space and simplify each of the rows. We realized that, in order to serve the larger range of functional elements represented with icons, we would need to design several net new ones exclusively for GANTT. This led to an icon audit to discover which existing icons could be repurposed and which would need to be added.
After compiling all the icons together, we created a reference document at the request of the dispatchers to help inform new team members about the meaning behind icons. You can see that document below:
GANTT in action
Because a prototype does not illustrate the many states of GANTT based on the nature of it’s live tracking, I put together a slideshow below of the various states of the application that you can review.
Standard view of GANTT for active day routes
GANTT view with time toggled to night
Loading page that displays as rows populate
Filtering by driver with menu deployed
Site scores modal deployed to track changes in efficiency quotients
Detailed ticket view that shows granular data on the separate pieces of each ticket in the timelines
Impact of the new and improved GANTT tool
After its update, GANTT has proven to be a superpower tool for WM’s route dispatchers as they track the status of resources live. As routes occur, they can see where drivers are spending extra time, where inefficiencies arise, and even where special cases happen that require maintenance or backup dispatching, events which produce data that can then relay back to the Resource Center where assets or workers need to be allocated to complete routes as planned or expected. This improvement in efficiency yielded in an over $3.2 million in cost reduction in it’s pilot year.
Since its implementation, the new focus of the team has been to improve the latency of data sent to the program and also to find ways to relay more granular data to dispatchers so they can be even more informed about where inefficiencies rise and how they can be accounted for. One offshoot of those initiatives can be seen above in the ‘ticket breakdown’ modal, which splits each individual ticket in a route timeline down by the minute in terms of time spent traveling, with customers, returning to sites, emptying containers, and more. This new modal also shows the expected versus actual timing for completed tickets, giving dispatchers insight into where planners may be over- or under-estimating the time needed for each part of the task.
I have not been privy to the actual quantifiable breakdown of improvements for these projects as they are handled by the research team and not shared due to privacy concerns, but I have heard that in tandem, the new Resource Center and GANTT tools have accounted for a large improvement in usage efficiency across the board, and are exciting bases for expanded functionality that will continue to push our software ahead.
DRIVER ASSIST TOOL
The tool for WM’s truck fleet
Driver Assist was one of the most exciting projects that I worked on at WM because of the direct impact it had on the company’s most basic and essential line of business; residential and commercial waste and recycling collection. The Driver Assist tool was loaded onto Android tablets fitted to the dashboard of every WM truck in their collection fleet, and was created for several purposes that previously had to be done manually.
First, the tablets connect to the trucks’ OBD ports to automatically run mechanical tests as part of every truck’s pre-route maintenance check. These tests required revving the truck’s engine, checking various air pressure and evaporation sensors, testing the functional of mechanical arms, and more. Driver Assist ran a software that would perform all of these tests for the driver and automatically report back data and any causes of concern or additional inspection, replacing what was previously an entirely manual process.
Second, the tablets were made to run custom navigation software to guide the driver through each scheduled pickup and ticket, including directing them to access points, showing visuals of the container type and location, providing information about the business or residence and who to contact, and recording the status of each pickup as it occurs. This navigation software also helps drivers adhere to the most efficienct collection routes to reduce truck carbon emissions where possible.
Last, but certainly not least, the tablet serves as a crucial communication point between the driver and the dispatchers, allowing the latter to push any updates about the route, tickets, or pickups to be made, and allowing the former to communicate any errors such as missing containers, obstructions, or overfilled containers that cut routes short and require additional landfill visits.
DA screen showing ticket details and container types for current pickup
Modernizing DA for the 2020s
Unlike other internal tools I worked on, the Driver Assist tool was not starting from scratch or from a previous out-of-the-box version. Our task was to overhaul it by bringing it into our internal design system and adding in needed functionality that had become available via recent Smart Truck advancements to the company’s vehicle fleet and internal data processing systems.
Our goal with Driver Assist was to simplify the UI with new design system elements, improve usability by improving process organization and proximity of related tasks, and integrate new functionality.
Driver Assist was overdue for an overhaul, with added functionality and an improved, design-system powered UI that made complex driver functions feel intuitive.
Supercharging with our updated design system
Following the successful applications of the design system in Resource Center and the GANTT Tool, Driver Assist was the final internal tool to benefit from our new atomic-based design system to bring a consistent visual feel to the full spectrum of WM’s portfolio.
We faced some challenges working with developers on integrating certain components because of hardware and software limitations, however. The version of Android run on the tablets, as well as the processing capabilities and memory of the tablets themselves, created a goalpost for leaning out the software and trying to work around native derivatives where possible. We made some compromises in this regard, but overall were very happy with how things turned out.
We followed similar feedback session protocols as the prior two tools to make iterations until we felt the application was up to proper operating order for the truck drivers. This new version has not yet been shipped, but is slated for January 2024 deployment after making its way through the dev team’s November sprint. You can look at a simplified prototype below to see how many of the screens turned out:
TAKEAWAYS
1. Design Systems super-power product suites.
2. The end user's needs are not universal, but will always be crucial to product success.
3. Design Operations and strategy-at-scale can make even the largest projects efficient.
IMPACT
$3.2 million in cost savings by improving route efficiency
machine learning integration flagged over 8,000 sources of inefficiency in route planning standards of the previous version of Resource Center
the cost savings came from reduced route lengths, avoiding missed pickups, less fuel used, and fewer instances of drivers going over their planned hours of service
Board confidence in the future of Driver Assist
After several years without updates, the improvements designed for the Driver Assist application and implications for service efficiency gave renewed confidence in meetings
Faster, more scalable building of components for the future of Internal Applications
The new system of component building and updated atoms + molecules libraries allow for building new pages not only faster, but with improved consistency and guaranteed component function
The apps that employees at WM use can evolve faster and more effectively as new technology becomes available to modernize services
88% Employee Approval Rating for Internal Apps on VOE Survey
After first year of deployment, employees rated internal software at 88% positive for ease of use and comprehensive functionality (up from 59% the previous year)