Project: #1_Tourists in Public Space
Development of frequency measurements and personas for marketing and management/politics.
The Solution is a management tool named TWO VIEW. This tool was created during the Hackdays on 28/04/2021 and 29/04/2021. TWO VIEW is a simple solution for the management of a destination. One dashboard represents the target in numbers and facts. The second dashboard shows the development of the KPIs of the destination (like revenue, mapping of personas, weather correlation to revenue and seasonality). This could also be predicted in the future.
Utilize data to analyze and visualize frequencies and dependencies
Outlook for further analysis / data requirements
Data from different sources is available but not combined to get the full picture. Below we present the useful data sources which could have a great impact - on the management tool TWO VIEW.
Projection should be taken with a grain of salt, as it does not represent reality. But there is information with added value - such as the arrival of guests, evaluation of daily and multi-day guests and number of stays (duration and number of days).
Rich data source to understand visitor segmentation.However, must be broken down to the tag, article to generate value. Revenue (Restaurant)
Through the Facebook and Instagram can be determined the guest structure, as well as generate reach through paid advertising.
Google Analytics should be implemented correctly. Here, an analysis of the users on the website would be beneficial. Also, advertising could be placed via Google Ads.
GWD Data (WIFI)
Collect wifi data to measure # of people passing by (currently just 2 months).
Diemtigtal: SKIDATA, Restaurant Turnover, Social Media, Snow Conditions, Value Study Solothurn: SKIDATA, WLAN, Social Media, Google Analytics Intervista: POI (Parking and cable railways)
Though there is plenty of data, it was hard to combine the different data sources. Some data formats were hard to clean and visualize. In example, there were not unique ids and inconsistent entries for some data columns. In the future, a stable process should be a goal. For that all exceptions have to get caught. The development machine learning models was hard to implement with limited time. The data needs some feature engineering to apply i.e. time series analysis. We did not want to publish preliminary results. We would recommend DeepAR (https://arxiv.org/abs/1704.04110) for this problem. Additionally, we wanted to integrate the weather data, but our API key failed to work and we ran out of time.
We have compared the web frequencies with the estimates from Intervista. The differences are quite large, but they balance each other out. The estimates of the Intervistata data are one time twice as high and one time half as high as the rail frequencies of the Wirihoren railroads. I could not compare the figures in Solothurn.
As part of the project, we were able to participate in a hackday event for the first time and gain experience. We were able to work with real data from various data sources during this time. The project partners were very motivated and helpful. Since there were several project partners involved in this project, communication and scheduling was not easy, but after a few meetings we were able to develop a flow that worked for all of us.
- Setup software pipeline to automate visualization process
- Recommendations for data quality: unique id, ‘Kartenbezeichnung’ unique
- Identify visitor segments missing at certain times (evaluation of daily and multi-day visitors, main means of travel arrival, number of stays (duration and number of days))
- Look for impact marketing events and weather conditions on frequency (correct implementation Google Analytics)
- Data would allow for Machine Learning models
TWO VIEW ROCKS! :)
Jan Alexander Zak
OUR CHALLENGE OWNER ARE: