In 2021, I joined Amazon as a Business Analyst for a four-month period, working within the relatively young DSP (Delivery Service Partner) program under Amazon Logistics. The DSP program is a franchising initiative that empowers entrepreneurs to establish delivery businesses with Amazon’s support, enabling them to deliver packages to customers. My role was specifically within the EU DSP Communications team, which is responsible for all communications directed to the hundreds of DSP businesses across Europe.
The goals and targets that the team had were clear, so were the metrics that tried to capture how well the team was doing their job and how satisfied DSP owners were with the communication they received. What was lacking was a clear way of reporting on the metrics that the team set itself. Reporting was a manual, inconsistent task carried out by the manager of the team every week. The relevant data was scattered across various platforms, including Redshift clusters, S3 buckets, and tools like Mailchimp. Accessing, parsing, and unifying this data into Excel-based reports was time-consuming, particularly in Amazon’s fast-paced environment.
This was especially critical because of the fast cycle-times a company like Amazon has. Always moving and shipping fast, changes to the DSP program were rolled out very frequently. Keeping DSP partners in the loop and informed, and staying ahead of seasonal peak events such as Prime Day or the christmas season required communication to be opened and read quickly. Information on open times was locked in internal tools, and could not be easily cross-referenced against the campaigns the team was sending.
Within my 4-month stay, I architected and implemented a python-based reporting tool that automated all of this work. While I first considered to tackle the root cause of siloed data, concerns around data privacy, technological legacy burdens and more made clear that this was not a viable solution at this point. Instead, the reporting tool I engineered interfaced with the different data sources via their APIs or Query Languages to gather the data it needed, and took care of parsing and calculating metrics itself — ultimately outputting a formatted PDF file with generated tables, charts and plots.
A significant concern for the team was maintainability. Since they did not have access to a permanent Business Analyst, it was essential that the reporting system could be customized without requiring in-depth software engineering knowledge. To address this, I developed a separate customization module that allowed the team to adjust settings and KPIs through a single configuration file.
The reporting squashed the time needed for the team to generate their reporting, while improving on consistency and accuracy. Overall, I loved a lot of things about "the Amazon way" of doing things — small and nimble teams, working away relentlessly at the goals they set themselves. All the while moving fast and meticulously, especially through the elimination of unnecessary clutter such as powerpoint slides, which is something I try to stick to to this day.
