EXPEDIA GROUP TECHNOLOGY — DATA
A Summer Like No Other: My 10 Weeks With Expedia Group
Starting my career with a virtual bang!
The day I received my offer to join Expedia Group™️ (EG) as a summer Analytics Intern was one of the most exciting days in the past couple of years. I applied knowing full well that I would have a steep learning curve for the first few weeks. I study Economics at University, so it was safe to say that coding and analytics wasn’t my forte. However, I was so excited to join one of the biggest tech travel companies that my nerves were overshadowed by my excitement.
The Early Days
Before I joined, my manager, Ellie Su, and my buddy, Jenny Boylin, contacted me to introduce themselves, and I knew it then — I was about to join an amazing team and have the summer of my life.
Onboarding
During my first week, the Early Careers team had organised a number of onboarding/introduction sessions. They were very helpful, as they explained what EG does, how things work, and how the company makes money. Then, we were given resources to learn more role-specific skills.
During my first day, I had an intro meeting with my manager and team, who were super nice and welcoming. In the following days, I discussed my goals and aspirations for the internship with my manager, and she introduced me to the projects I was going to work on, more on that later. For one of my projects, I needed some coding skills, SQL and R in particular.
Learning SQL and R
I started my internship with no coding skills, so my manager provided me with the resources (online courses) to learn at my own pace. In the first few weeks, I spent around half of my time working on my research-based project, and the rest learning SQL and R.
It was a very steep learning curve — I began with SQL, which I needed to query data for my analysis project. I followed a beginner’s course, then a slightly more advanced course. I knew the way to learn wasn’t by watching videos, so I started trying out queries by myself.
After the SQL course, I started an R course. This was more challenging, and it took me a bit longer to get my head around it. However, I started practicing on some sample data, and things were going well.
Around week 4, I was ready to get my hands on the real project.
Intern Sessions
Throughout the internship, we had two types of sessions — the Leadership Series, where leaders of EG spoke to us about their journeys, and a series of sessions on topics such as Mental Health, Personal Branding, and Allyship. I thoroughly enjoyed both, some highlights including:
· The talk with EG’s CEO Peter Kern, where he gave us interns some very helpful advice for our future careers, making it very clear that EG values the interns and our fresh perspectives, and that we were encouraged to share our ideas, challenge the status quo, make mistakes, and try again.
· The talk with Ariane Gorin, President of Expedia® for Business, where she shared her career journey and gave very inspirational advice.
· The Mental Health session, especially all the different techniques to improve your mental health.
· The Personal Branding session, especially the personality test and reflection.
My Team
When I was interviewing for the role, at the end of every interview, I always asked: What is the best thing about working in EG? I always ask this question, as the answers of those interviewing me really showcase the kind of company I’m trying to get a job at. I was almost convinced that everyone who interviewed me had a scripted answer to that question that they were told to say — they all said the people and the culture. I found out for myself, pretty quickly, that it was not, in fact, scripted.
Everyone I have interacted with, from the leaders, to the staff in the London office, were so nice. The people I interacted the most with, my team, were amazing beyond description.
I was placed in the Product Insights and Measurement Strategy (PIMS) team. My manager, Ellie, went above and beyond to ensure I was comfortable and able to be my best self throughout. Early on, we established her expectations of me, as well as my expectations of her. We agreed on a style of communication and feedback that worked well for both of us. I felt fully supported throughout, with weekly catch-ups and ad hoc calls, but never micro-managed.
My buddy, Jenny, was also amazing. She dealt with my million daily questions and tech issues. We also had weekly catchup, which always overran, as we were talking about our work, as well as catching up on our lives and the fun things we did during the weekend.
Another person on my team, Haochen Song, was equally great and very helpful. In addition to our weekly catchups, and office chats, she very gladly helped me with one of my projects by working with me to query some data for a mock-up.
I had the chance to go out with Jenny and Haochen, along with a few other people from the wider team, which was great fun.
My projects
Cohort Analysis
One of my projects was to deliver a fact pack and a set of recommendations on Cohort Analysis (CA) and its uses in EG.
CA is a type of behavioural analysis, where users are divided into groups (cohorts) based on shared behaviours/characteristics, and the cohorts are tracked over time. There are two types of CA — Acquisition CA, where the groups are split by acquisition date, and Behavioural CA, where the groups are split by certain behaviours they exhibit.
After laying out the background of CA, I set out to put together a list of metrics to recommend for use in CA in EG. I landed on a list of 15 initial metrics, separated into 4 buckets — Financial Metrics, User Behaviour Metrics, Property Metrics, and Trip Characteristic Metrics.
In addition, I also:
· created a Suggested Framework on how to perform CA in EG, which can be used by anyone new to CA to understand how it works and what the steps are.
· put together a timeline of highlights of work done on CA within EG, which will be used as a reference in the future.
· explored a new way to measure churn using CA.
· created a list with possible example analyses, including what metrics to use in them, and what insights we can gain from them.
I rounded off the deck with possible issues, such as CA being very computationally expensive. Although CA is very good for spotting trends and patterns in the behaviour of our customers, which can show us why they are or aren’t coming back, it doesn’t deliver any actionable results by itself.
My project concluded that EG should use CA in the future, as there is a lot of potential to unlock using it.
I ended up creating a mock-up CA, for which I needed to obtain historical data in the right format, which proved difficult to do. I asked my teammate Haochen to help me query the data, which I was then able to use to create the CA tables, which illustrated one of my examples.
I presented this project in the Intern Showcase, as well as to an internal expert, Li Guo, who had begun doing practical work using CA during my internship. Li validated my work and we discussed and agreed on the next steps and how my work will integrate with hers.
H4P Sorting Algorithm Widget A/B Test
For my other project, the aim was to analyse an A/B test on the sorting algorithm of an H4P (Hotels.com for Partners) widget. The test ran earlier this year, and its results were going to form the base for the next steps for that algorithm and widget.
I used my newly acquired SQL skills to query the data from the test, which consisted of Control and Variant Impressions (the number of times the widget was seen, Control as shown by the normal algorithm, Variant as shown by the algorithm that was tested), and Control and Variant Clicks (the number of times customers clicked on the widget). After querying this data, I adapted some old R code used for a similar test and ran it to check if there was a significant difference between the control and the variant CTR (Click Through Rate — the number of clicks divided by the number of impressions).
After obtaining the overall result, which was negative (the variant algorithm performed worse, as the CTR was lower for the variant than the control), I dug into the data further, to see if there are any learnings we can gain. I split the data by partner, and for all partners with enough data for robust analysis, the impact of the variant algorithm was negative. I then queried more granular data to investigate the effect of hotel star rating, guest rating, and location (country).
The most notable findings were to do with the property location. I looked into the countries which had enough data for robust analysis. All of them had a negative lift, except for one of them, which had a very high uplift. I looked into this outlier by querying the cities where the properties were based. This highlighted interesting insights around the algorithm’s mapping. To show these insights, I created an interactive map showing the cities from the dataset, along with the country’s airports.
I presented my deck to the project’s stakeholders, who were pleased with my work. There was a discussion on the next steps, where I was asked to replicate my analysis for the high uplift country for the rest of them.
Challenges
The main challenges I faced were getting up to speed with the EG and industry-specific jargon/terms/acronyms and learning the required technical skills. This took me a few weeks to overcome, but after that, it was mostly smooth sailing. I also encountered problems with writing queries, but I always had someone to ask for help.
Outcomes
My last day was very emotional, as I was quite sad to be leaving. I started the internship knowing next to nothing about analytics, coding, and working with data, but by the end of it, I was much more confident in myself and my technical skills than I thought I could be. I learned a lot (that’s an understatement) and created some great relationships with my team. My summer with Expedia Group was nothing short of amazing, and I really felt like I found my place, doing work I thoroughly enjoyed, in an amazing team I felt like I belonged in.
