Welcome to Data Guasu! In this blog I will be sharing opinions, ideas, experiences and details of projects I have worked on, and my journey to becoming a data passionist (due to the impostor syndrome I’m having trouble calling myself a data scientist).

As the name implies most posts will be related to data, covering a variety of topics, techniques and technologies as tools to answering (real life) questions.

Now, where does Guasu come from? It comes from Guarani (one of the official languages in Paraguay, in addition to Spanish) and it means big. So there it is, the name Data Guasu might refer to big data (I could not find a suitable Guarani word for diverse data) but the scope of this blog is much guasu than that.

Traditional Chipa Guasu dish from Paraguay. (Source)

My Journey

Data Warehousing 1.0

My interest in data began around 2009, when I started working as a Data Warehouse Consultant for a telecommunications company in Accra, Ghana. At that time we had a Data Warehouse running in SQL Server 2008 with multiple ETL processes in place to load CDRs from call, SMS and data transactions. I got to learn a lot about SQL Server technology stack (SSIS, SSRS and SSAS) and dimensional data modeling based on Kimball techniques.

Wrong Data = No Data

After being part of the DWH team for a year and a half, I moved to a different role focusing on building dashboards and reports for the operational team so they could track their sales and monitor their commissions.

It was a difficult time to join the team because there were a lot of claims on how commissions were paid. I remember receiving easily a few hundred emails per day from freelancers that wanted to know why they were paid less than expected. Going over the process and doing analysis on the data we identified multiple inconsistencies that were generated because freelancers were usually sharing mobile devices (and thus, phone numbers). The numbers were used to match the sales to their names so one freelancer that did not sell a line was receiving a commission for the one that did.

To solve this, we worked out a new registration system via USSD asking each freelancer to send their details and we assigned a unique ID for each. So, if they wanted to use a different phone number they could just use their ID to ensure their sale was going to be assigned to the correct person. This solution provided consistency in the data, improved commission payment process and also increased sales because freelancers were motivated!

(Business) Value then Technology

When we got over the data issues, we felt that we needed to have a better and faster way to keep track of the sales. From DWH we could only get data the next day but the team was eager to look at how they were doing during the day to take immediate action.

I started doing some research on possible reporting/dashboard tools and found one called Dundas Dashbard. The integration with SQL Server was pretty straightforward so I decided to build a set of specific dashboards to share the benefits with the business stakeholders.

It wasn’t anything fancy but it attracted enough attention because of its potential value. These simple dashboards were providing data that was not previously available in that format and with (only) an hour delay. Soon after, my request to buy more licenses was approved and the dashboards were shown not only in offices but also sent by email to territory managers. The project then became regional and was replicated to other countries in Africa. This was a great learning of the importance of using use cases that add real value to the business, even when prototyping.

Data Warehousing 2.0: Learning to Fail (or Why Big Bang approaches don’t work)

Once the Business Operation Center was set up, a new opportunity came up to lead a regional Data Warehouse project. The idea was to implement new DWHs from scratch in all of the African operations based on a consolidated set of requirements from these countries. After going over the bidding process we decided over a particular provider.

The scope of the project was to first understand business requirements and then translate these into actual reports, OLAP cubes and dashboards that would be delivered along with an Enterprise Data Warehouse and a standard data model to be shared across. Everything looked great the moment.

The advantages we saw on this approach was that the cost was contained, the scope was fixed and there would be no surprises. The problem was that there were surprises. And bad ones.

• We did not realize of the (poor) quality of the deliverables until late in the project.
• Many of the reports and dashboards that were asked to be developed were not longer relevant.

Change in Direction

We had to put a hold on the project and decide on a new way forward. We realized that we needed to work with a different approach with shorter end-to-end iterations and show value at each stage. Reducing scope allowed us to reduce timelines and focus on what is important. We also decided to work on deliverables that could answer future business questions and not static canned reports that were to become useless later in time. Examples of these deliverables were faster ETL processes with an auditing system to track data lineage, standard data models to perform ad-hoc queries and OLAP cubes for data analysts.

This new approach was successful and we were able to gradually deliver better Data Warehouses across the region.

Back Home: New Perspective

In 2015 my family and I took the decision to go back home (Paraguay). We had been living abroad for many years and after my second daughter was born in London, felt that we needed to be closer to the family.

The company where I had been working had a local operation in Paraguay so my first choice was to contact them. They were looking for someone to lead the Loyalty and Retention team and offered me this position. First, I hesitated because it was not what I was hoping for. My idea, naturally, was to continue working on DWH/BI projects but there was no vacancy on that area. After some careful thought, I decided to take the position and use my data passion and experience in the field to support initiatives with a different perspective.

It’s been three years since we arrived and I’m happy to have accepted the challenge. Last year, after working on great projects to increase loyalty and reduce churn, I was promoted to Strategic Offer Manager in the Prepaid segment. Here I get to use my data skills to evaluate new potential offers, work on pricing models, segmented offers and better understanding customer behavior.

Back to (online) School

I wanted to pursue a Masters degree right after finishing my Bachelor in 2007. I’m glad I didn’t at that time because I was not really sure what I wanted to work on. Almost ten years later, in 2016, I applied to the online Masters in Data Science from the University of Indiana and got accepted. It was a tough decision because I had to manage my time between my family and a full time (pretty intensive) job. Even so, I knew this is what I wanted to do and went for it!

Almost two years later I’m in my last semester and if everything works as planned, I will be graduating in December.

What’s Next?

I did not expect for this first post to be this long but I felt I had to share my journey before I go deeper into the subject.

In the next post, I will be analyzing a dataset of ATP tennis competitions (forgot to mention I’m a big tennis fan) using Python matplotlib and seaborn visualization libraries.