cathvos/ih-final-project

Final project

Ironhack Final Project

Cocktail/Mocktail flavor trends

by Roger Serret & Cath Vos, June 2022

Project Brief

Scenario:

Bacardi, the company:

• Sell +200 brands
• Present in +150 countries worldwide
• Main product categories: sparkling wine, rum, vodka, gin, vermouth, cognac, whisky, bourbon, tequila, cachaça, liqueurs
• Sustainability as a key branding element:
  • Positive environmental impact (environmental-friendly production)
  • Reduce alcohol-related harm (low/non-alcohol drinks)
  • Be inclusive (none-sexist ads)
  • Responsible sourcing partnerships (“local” partners)

Bacardi had several project requests:

• Forecast sales (KPI: Volume, 3 next years, detail expected accuracy)
• Suggest next spirit hit (By Country, Age Range, Wealth Range…)
• Suggest next preferred flavor for customers (By Country, Age Range, Wealth Range…)

They provided us with 3 internal data sets to use and we were free to use whatever external data source we wanted. When using external data sources, they asked us to explain:

• Which External data you have used (in the model)
• Explain Why you selected that data
• Explain How you mixed it

We decided to focus on the flavor and next spirit hit part of their request and connect it to one of bacardi´s sustainability goals: Reduce alcohol-related harm (low/non-alcohol drinks). As this was our focus, we did not use any of Barcadi´s provided data, but created our own data set that met our needs.

Problem:

Can we suggest the next preferred flavor for low/non alcohol drinks?

Challenge:

Create a data set of minimal 1000 cocktail recipes with a timestamp to find out what are the preferred flavors across seasons and years.

Data

Leveraging on cocktail recipe data we scraped from the web, retrieved through API´s from social media (Instagram and Pinterest) and found on Kaggle, we put together a Data frame in Python with almost 2000 recipes. The original Data frame was very simple and only contained 3 columns, Date, Ingredients, and Source Tag. The Ingredients column contains the full recipe(post) text, so we created a function in Python to extract the information on the ingredients mentioned in each of the recipe (post) texts. Our approach here was to create a function that would search through the text and create a Boolean value if the ingredient on our list was found in the text or not. This resulted in a huge Data frame of 72 columns. Once the Data frame was complete, we used Tableau's data visualisation tools to analyse the data.

Process & Tools

Process

• Github: set up our Github repo to collaborate on.
  
• Trello: set up our Trello board to help us plan.
  
• Python: web scraping, process API retrieved data, EDA, assessment of dataframe, data cleaning and wrangling
  
• Tableau: data analysis, creating visualizations
  

Deliverables

• Data: csv´s created with Python - Link to Data folder
• Notebooks: Jupyter Notebook - Link to Notebooks folder

Visualizations

• Plots & Charts: Tabeau worksheet - Link to Tableau
• Presentation: PowerPoint - Link to Presentation folder
  

Key Take Aways

1. We have validated that nolo drinks are a trend.
2. The most popular ingredients:

• Fruits: Orange, Grapefruit and Cranberry
• Condiments: Ginger, Syrups and Bitters
• Spirits: Gin, Rum and Vodka

Thank you for reading!
If you have any questions, please reach out to us.

Roger Serret & Cath Vos