Tatry Mountain Resorts

Datamining

Mark O

Reporting

10 years

of collecting a database that we had to evaluate

52,8 million rows

contained in the input data file

1 clear report

that gives the client an overview of the profitability of resorts and products

At the start of your business, you wanted to manage your business with data. So you set out to collect it. And you did it for 10 years. But accumulating information is just the beginning. The moment you want to evaluate the data, you can get stuck. You might be able to process such a huge database, but you're already dreading the time it will take. This was exactly the situation of our client Tatry Mountain Resorts (TMR), who approached us with a clear requirement: to find answers to questions key to their business by analysing customer data.

We were working for a key player in the tourism industry in Central and Eastern Europe

In the ten years of its existence, Tatry Mountain Resorts has become the largest investor in the High and Low Tatras region. It manages 20 hotels, 10 ski resorts, over 40 bars and discos as well as water parks, restaurants, ski schools, sports shops and much more in Slovakia, the Czech Republic and Poland.

And what did we solve for TMR? We were looking for questions and answers

The client wanted to mine the database to not only get to know their customers better, but also to find out which resorts were bringing in the most profits.

He needed to find out which products are the most profitable, how the smart season pass is doing in sales, or what the demographic distribution of customers is in each resort.

Therefore, as part of the data mining process, we designed the data processing and editing for the segmentation questions we helped the client compile. After that, it was time to clean and analyse the data.

We enlisted the help of the Indians

It must be clear from the outset that a database collected over 10 years doesn't quite fit on a floppy disk. So we couldn't use traditional processing methods. Instead, we "called to arms" a tool specifically designed to process such a large volume of data - Apache Spark.

The biggest challenge was getting a proper grip on the data and understanding dependencies that were not clear at first glance. Subsequently, we repeated all the analyses several times, as it was often possible to come up with a more reliable or accurate method when cleaning and analysing the data. This process therefore required judicious query setup, testing methods on smaller datasets with subsequent deployment on real data.

And what was the result of our efforts? We calculated the necessary metrics and tables and visualised the graphs into a presentation in our Mark O. What did it look like?

Datamining: number of new purchases F and repeat purchases R per year in absolute numbers (all purchases)

The number of purchases in general has been declining since 2019 due to pandemic measures. In contrast, the repeat purchase rate has held steady at around 70% (we work with filtered offline purchases in our analysis).

Míra opakovaných nákupů má pozitivní trend a obecně se neztrácejí vracející se zákazníci.

Repeat purchase rates are trending positively and generally returning customers are not being lost. Datamining.

Almost half (49.3%) shop for 2 or 3 people. 22.7% shop for just one person, but that doesn't mean they don't go to the mountains in a group,

The Tatra Mountains are looking good... for better data times.

A co by mohlo náslAnd what could follow? Automating defined segments and tracking them in real time - perhaps deploying different strategies to influence them. Wherever our joint steps with TMR will go, we're already looking forward to them!