The aim of this project is to keep skills for dashboards and deployment.
The data set comes from the papar Antonio, Almeida and Nunes, 2019. The data files can be obtained at tidytuesday repository.
A hotel PMS (Property Management System) is a system that manages information about hotel room reservations: Who reserved a room of type X for the period T1 to T2, the status of the reservation, and so on.
This dashboard reproduces the PMS dashboard.
What kind of dashboards can we build for the company/hotel by utilizing the PMS data?
- Hotel usage: Number of occupied rooms. Number of guests.
- Sales: Revenue. Revenue per occupied room (RevPOR).
- Marketing: Sales/Number of guests by country. Number of families. Marketing segments and distribution channels.
- Cancellation: Cancellation rate (by country), survival rate, number of no-shows.
Our ETL pipeline follows so-called medallion architecture.
- Bronze:
hotels.parquet. This is just a parquet version of the original data (CSV). - Silver:
bookings.parquet. An analysis ready data. Several fields are added, so that we can easily analyze the data. A simple example isarrival_date. The original data has three columns for the dates: year, month, day.arrival_dateis a compilation of the three columns. - Gold:
actions.parquet. This data shows the flows of reservations:arrival→stay→departure.
The data pipeline and the data assets are managed by DVC. The following command executes the data pipeline and reproduces data assets.
poetry run dvc reproDVC is enabled in this repository. Because the remote repository is private, you have to replace it with your repository if you want to reproduce this project.
- Codebase: GitHub
- Dashboard platform: streamlit.io
- The platform retrieves automatically the source code from GitHub and deploys the newest version.
- CI Pipeline: AWS CodeBuild
- If I push a commit, CodeBuild pulls the latest codebase and executes unit tests
- Data Storage: S3
- Dashboard, CI pipeline and local development environment pull the data from S3 via DVC.
