Experience with Adobe Analytics
I’ll be narrating the first phase. If you want to read about the second phase of the competition, here’s the link! https://bit.ly/gdsc-vjti-sahethi
Initiative
Trying something new every now and then has become a constant factor that we as engineers live by. Hence, getting our hands dirty with entering this world of Data Analysis and Research was extremely rewarding I would say! The season for bagging an internship was in full swing and certainly took a toll on everyone’s normal schedule. I desired to seek some activity that would help me step out of my comfort zone. An activity that would be out of the ordinary pertaining to what we regularly study.
One day, a friend of mine sent me the promotional banner of Adobe’s annual contest — Adobe Analytics Challenge 2021. I gave it a quick glance. The competition was about analysing an enormous database and discovering hidden KPIs (Key Performance Indicators) that served as a detriment to the company’s financial business. Having no knowledge about the said field was not the primary concern, for now, I had to search for a teammate!
The Team
After Sahethi agreed to be my teammate, the next task looked much more daunting. According to the rules, each team should have a mentor and that respective person should be one of your professors! We guessed such a rule must be enforced to protect the huge data contributed by the brand to not get into malicious hands. After a long day of discussion, both of us sent an email to the most felicitous professor, Prof. Shraddha S. Suratkar, who had a history of being extremely amiable as well as having experience working with different kinds of data. Fortunately, she agreed to be our mentor and the next day, all 3 of us registered as a team for this competition! Team ‘Goofy Data’ is ready!
The past editions of the contest looked quite formidable. We had no experience or knowledge about the research sector, hence our first approach was to gain as much knowledge as possible while also tackling our normal schedules. The competition spanned 3 weeks. The first week was just about registration while the next two weeks were to analyse the brand’s data and create a presentation to showcase your research. In such scenarios, maintaining a good network is advantageous as I found a couple of data analysing specialists who were ready to guide us and answer our puerile questions.
The competition commences
The contest started with a long tutorial about how to use Adobe Analytics. I would say it was pretty neat. And then the data contributor’s name was revealed- Disney! Sahethi and I were thrilled to be honest. Working and analysing the enormous data of such a reputed and historical brand was a marvellous opportunity. The questions and goals for the contest were provided after that. The first few days were perplexing, we were just experimenting with Analytics. Both of us had jot down notes regarding business workflows, retention boosters, KPI, customer paths and many other things.
Adobe Analytics had 4 main components:
- Dimensions: Reference for dimensions usable in Adobe Analytics.
- Metrics: Reference for metrics usable in Adobe Analytics.
- Segmentation: Focus on a subset of your data.
- Date Range: The time period to focus on
The competition was all about playing with these components to get the correct output that clearly reflected your idea or perspective. At first, both of us made a number of visualisations without thinking about what outcome was desired. At times, we were not able to understand what purpose a metric or dimension served, it was extremely puzzling.
The data was huge, the dimensions and metrics encompassed numerous aspects pertaining to the user’s history, visits and even countless paltry details that we never seem to pay attention to! Here’s a glimpse of most of the dimensions, metrics and other components:
The dimensions basically characterise the data in the rows. The metrics act as columns that filter the data in all the rows according to what they represent. And segments are the tools that provide the ability to group metrics and dimensions based on a certain logic.
These tables exhibit a lot of data but it might get difficult for other teams and personnel to digest or understand the science of reason behind it. Over here, visualisations come to play. We can augment the visual presentation of the data that can collectively convey the entire logic or proof. In schools, we all used to study about pie charts or histograms. We saw such graphical models as an easy way to earn marks in exams swiftly, haha! However, in the world of data analysis and research, they act as the essential tools to aid in presenting the varied outcomes.
Examining a huge dataset is an intimidating task. It’s the duty of the data analyst to not leave any stone unturned. The analysis conducted by such researchers predicts and structures the future campaigns and modifications of the brand’s business and many other factors. A tiny mistake can lead to a huge loss while discovering a key point can help the company soar in profits. Global as well as local factors too are taken into consideration in the research. A common task known as “finding the trend” is analysing a dimension filtered under a metric year wise. A lot of other researchers use Tableau, Python programming, R programming and many other tools.
After a week of achieving little to nothing, we almost gave up. Both of us observed that we had the correct views in mind but we failed to execute it every time. Why? Because we were not familiar with the naming conventions of the data or the visualisation tools. Where could we find help for that? The documentation!
Slides from our presentation
Walt Disney explained his success this way: I dream, I test my dreams against my beliefs, I dare to take risks, and I execute my vision to make those dreams come true. Dream, Believe, Dare, Do: These words reverberate across the decades of Disney achievement. We tried to reflect these 4 principles in our analytical work, hence we abide by the 4 principles of
- Visualise: Create valid visualisations that aid the brand
- Analyse: Examine the data thoroughly
- Dare: Propose concrete evidence backing our research and our discovered KPIs
- Do: Propose solutions and relevant actions that improve the growth of the brand and satisfy the end-users
We have taken a 3R approach for our presentation, i.e. Rundown, Remedies and Reinforcement.
In this slide, we have focused on our first business goal.
•Through the graph, the behaviour indicates that when a new customer becomes a return customer they are bound to spend less time on the website/app and perform higher cart additions than their first visit.
•The area highlighted on the map shows the countries with the most profitable activities. And you can see it ranked, first being the USA, then Canada and so on.
•Below, the revenue — discount ratio has been calculated indicating the more loyal your customer becomes the more discount they expect. So Revenue Discount Ratio is inversely proportional to customer loyalty.
This slide focuses on our 2nd business goal where we are aiming at driving the average value of the cart up
SUMO here states single unit order vs multiple unit order
Over here you can see that we have analysed yearly how people are buying more items per order rather than a single item. There are only a few who buy a single item per order. Our goal here is to maximise the multiple unit orders so that the ratio of single item order is decreased and multi unit order value is increased.
According to a survey we conducted with 100+ candidates, the pie chart represents the feature of “frequently bought together” and how many people think that it will allow them to make a MUO. And here in the image you can see that the Disney Shop website doesn’t have this feature but having this can lead to an increase in average order value.
We estimate a growth rate of 31.95% by 2023 on implementing the solutions provided by our 3R approach. (Y-axis — revenue in dollars). I used a number of revenue-predicting techniques that revolve around normal mathematical operations. As the data is confidential, I am unable to reveal the intense calculations.
The documentation of Adobe Analytics was a hidden treasure that we discovered extremely late. It explained many of the key metrics and dimensions, various ways to create your own segment & date range and many other things. I basically finished reading the entire documentation in almost 2 days and immediately notified Sahethi about it. The next few days were rigorous.
Read about the next phase narrated by Sahethi over here! https://bit.ly/gdsc-vjti-sahethi
Team Members
- Neel Dandiwala, connect with him via LinkedIn.
- Sahethi DG, connect with her via LinkedIn.
