Global Mile is a foundational team within Amazon that helps to build, manage, and operate its global transportation network for orders that cross international borders. Global Mile Exports Visibility was set up in 2019 to architect and own the team’s data management, strategy, and underlying platform.
At Amazon, the customer promise (great products delivered exactly when they are scheduled to) is its unique selling proposition—and we’re a huge part of making that happen, so getting this right is a priority for us. Fast, scalable, global architecture that delivers value—that’s our focus.
In addition to overseeing the worldwide Global Mile (GM) exports organization and its data engineering and business intelligence (BI), we also focus on:
- Using insights to support the wider business and executive decision-making
- Developing supply chain control mechanisms to preemptively detect issues and mitigate risk
- Use data science for planning, optimization, and automating decision-making
In this post, we discuss how Global Mile migrated their BI dashboards and reports to Amazon QuickSight to enhance their operations.
Increasing data volume and smarter reporting requirements: The time to change
We decided to switch to QuickSight due to the sheer amount (currently at 0.24 petabytes) of mission-critical data that we deal with. We needed to make it more accessible and shareable—and we needed to be able to pull powerful insights from it. We work alongside a group of international carriers, all of whom rely on us for coordination, so that data is vital to identify and resolve supply chain issues before they can impact customers.
We report across a wide global network. This required a secure tool that could deliver more efficient, integrated reporting and be able to share customized dashboards to both external partners and internal stakeholders without compromising data protections.
Metrics are core for us; from Delivery Estimate Accuracy to Scan Compliance Rates and Defect per Million Opportunities, we had to find a better way to report and share data from a central source of truth. By democratizing data through dashboards, we’d be able to reduce manual tasks, drive operational efficiency, and verify that everyone involved had clear oversight of metrics and goals.
We are also able to improve our BI further by integrating additional machine learning (ML) capabilities to deliver a higher level of insight and predictive analytics modeling. That’s why QuickSight, with its built-in ML capabilities and powerful AI, felt like a strong choice.
Technical challenges needed a modern solution
We also wanted to use QuickSight to tackle some of the technical and architectural challenges that were stopping us from delivering even better insight. These challenges included data availability, external reporting systems, custom data queries, and tech stack integration obstacles.
Core to everything that we do is the freshness and availability of our data. In our current system, we were coming up against data capacity issues, and had scheduling restrictions on data refresh rates. Our aim was to get as close to real-time reporting as we could—and we couldn’t deliver that in the way we were working. QuickSight’s powerful data pipelines, SPICE (Super-fast, Parallel, In-memory Calculation Engine), and cloud-native model meant we could circumvent any stoppages or slowdowns—without the need for additional operations budget.
We also struggled to share our data with external partners (each of whom had differing needs, focuses, and geographic regions) in an interactive, tailored, and meaningful way. Even internally, reports were typically generated as static PDF documents—meaning they lost relevance fast. QuickSight allowed us to create tailored, interactive dashboards that we could share with our partners securely.
Additionally, our stakeholders often ask for particular datasets or have specific questions they need answered—something that would usually require a lot of manual approvals and report building. Moving to QuickSight gave us the ability to add in a level of self-reporting and querying for those stakeholders, vastly speeding up the process of information acquisition. We’re in the process of rolling out Amazon QuickSight Q natural language querying, which will add an even greater level of intuitiveness to this process.
A final challenge was a common one: trying to get rival systems to work smoothly together without a host of manual data engineering. QuickSight was an ideal choice for us, because we had plans to integrate more of AWS’s ML tools for greater insight into our data—and the vast array of APIs and integration aspects available through QuickSight has made that even easier.
Taking the leap
Considering the scale of Global Mile and Amazon, we knew changing our BI solution would be a significant undertaking.
But in reality, from the point we made the decision, it only took us 7 months to migrate all of Global Miles Exports Visibility’s processes into QuickSight. We also managed to avoid the need for downtime, so day-to-day operations were unaffected.
The following is a broad list of the phases of our transition to QuickSight:
- Deprecation of low-usage dashboards based on utilization statistics
- Migration of peak season dashboards and other high-usage critical dashboards
- Migration of Weekly Business Review, Operations Business Review, Monthly Business Reviews, and Quarterly Business Reviews dashboards
- Migration of remaining non-critical data for peak and deep dive dashboards
- Deprecation of legacy licenses
Our data systems are complex because they span a large number of external partners, so the data lake that forms the basis for our BI is mainly based in Amazon Redshift. The following diagram illustrates our data warehouse architecture.
It was impressive that such a comprehensive data network was integrated onto the new architecture in such a short time period—something that was, again, helped by the flexible URL and API abilities of QuickSight.
Operational insights at your fingertips
The migrated QuickSight dashboards are now being used across our organization to carry out day-to-day operations. We have near-real-time dashboards such as the linehaul performance dashboard, which is a one-stop shop for linehaul visibility into on-time performance of approximately 750 linehauls per week, along with tracking the number of packages impacted due to daily linehaul delays.
We have several deep dive dashboards such as the DEA/EAD deep dive dashboard, where we have set up dynamic selection parameters and URL actions to enable stakeholders to directly navigate to the details of a certain shipment in tools such as Outbound Lookup Tool, ATROPS, and Eagle Eye.
Our Weekly Business Review (WBR) meetings are conducted using our WBR dashboard, which is a central dashboard for all WBR metrics. This dashboard contains a landing page that enables the stakeholder to navigate to the desired tab as needed instead of having to scroll through the page bar on top.
We have also built an NASC capacity utilization dashboard and have set up alerts through the visuals, which notifies the operations team whenever a particular sort center is at the risk of capping out. This was a very critical tool for peak seasons.
We have also scheduled customized email reports for our carrier cap slam dashboards to be sent to the respective carrier managers, which help them proactively reach out to the carrier for day-to-day planning of volume.
Alongside the migration, we embedded QuickSight including QuickSight Q natural language querying into Pharos, our Export Control Tower system. Pharos, in its new iteration, works as a one-stop shop for partners within our supply chains to access all export Delivery Estimate Accuracy (DEA) risks in one place. This gave us the ability to tailor datasets and UIs for each partner, breaking down issues by destinations and origin countries, and providing visibility across the supply chain in a much more granular fashion.
Delivering insightful analytics for Global Mile
The results that we’ve seen after the migration have been nothing short of revolutionary.
The ability to set up notifications on near-real-time critical dashboards has led to our network operations team being able to work much more proactively to address supply chain gaps. We estimate that we’ve saved 6–8 hours a week (which would otherwise be spent on manual audits and data analysis) thanks to the paginated reports and interactive visuals QuickSight delivers.
Additionally, the parameters and bookmarks for various dashboards used by stakeholders are far more efficient. Each user can get relevant data for their region, marketplace, and role for more efficient insight—and they have the ability to change this and look at different views and datasets when necessary.
URL integration has also opened up the wider AWS world for us. Various important internal tools—such as Outbound Lookup Tool, FMC, ATROPS, and Eagle Eye—are all linked via QuickSight, which means we save even more time by not needing to switch between apps to access data.
QuickSight has helped increase productivity across the board. For example, our WBR deck was previously generated by running a script that downloads PDF snapshots of different Tableau dashboards for each region and stitches them together. This end-to-end manual process took approximately 3 days per week. QuickSight helped us build a cleaner and robust dashboard solution, increasing the agility to drive the WBR meetings and provide our leadership with easier access to navigate through our dashboards when required.
Due to the end-to-end cloud solution and it not needing to work from licensed desktop programs, development time is estimated to have been reduced by 15–20%. And that’s not to mention the reporting capability we now have for weekly business reviews. Where before we had a 3-day process, which resulted in PDF reports for stakeholders to work through, we’re now seeing both a huge reduction in time and a better standard of report with the interactive, paginated reporting system in QuickSight.
Although the time-saving and data intelligence aspects are most important, we’ve also seen a substantial impact on our bottom line. With the current size of our fast-growing team, being able to leave behind the hundreds of Tableau licenses has lowered our costs tremendously. And, as the team expands, so too will those savings.
Enabling continued innovation
Crucial to our successful migration to QuickSight was the openness and flexibility of the team itself.
To make the change to a fundamental piece of Amazon’s business model, we held weekly training sessions to upskill our team as we went. Plus, each time we had feedback, the QuickSight team took these on board and developed new functions and systems to better the product. With the end-to-end serverless model, there weren’t updates to tackle when these were deployed; we’d open up the system and it would have automatically updated its capabilities.
We’re very excited to explore QuickSight moving forward and continue working with the team as the system evolves. In particular, our rollout of QuickSight Q looks set to be a game changer—an intuitive, natural way for those using the system to query, adjust, and examine data without relying on approvals, report building, or dashboard creation.
To learn more about how QuickSight can help your business with dashboards, reporting, and more, visit Amazon QuickSight.
About the Author
Aishwarya Busi is a Business Intelligence Engineer on the Global Mile Exports Visibility team at Amazon. Prior to Amazon, she worked as an Analytics and Business Intelligence Lead for a kidney care startup and focused on improving and impacting patient health outcomes through building reporting infrastructure that helped the care teams drive patient engagement and daily operations.
This is a companion discussion topic for the original entry at https://aws.amazon.com/blogs/business-intelligence/global-mile-exports-uses-amazon-quicksight-to-orchestrate-a-global-transportation-and-logistics-network/