Case Study: When Do People Buy Snow Tires?

Posted on: September 4th, 2022 | Posted by: Landan G




If you live anywhere that gets cold or has snowfall, then you've probably either bought or wanted winter tires to help with traction and grip in those harsher conditions. Most people typically are not shopping for winter tires in July or August. It usually becomes something to consider when you get the first snowfall of the year and the initial panic of winter driving sets in.

With that being said, we can use Google search data to explore when people are typically searching for winter tires, and even predict the searches for the future, hence this case study.



Snow Tire Search Interest


The data we're using in this case study contains monthly Google search data from the Canadian province of Alberta between the dates of January 2004 and August 2022. Alberta can get very cold and a good amount of snow so it should reflect the shopping habits of a large group of people living in a similar geographical area.


As you can see, there is a very clear repeating trend in the data of a steep rise in the fall, the peak in the winter, and then a steep drop come spring and summer. You can also see a slight upwards trend, meaning you have higher valleys and peaks indicating an upward trend in searches over time. Despite this data being repeating and very similar year after year, we can smooth it out a tad using a moving average to make the data less prone to outliers.


After applying the moving average we can clearly see the pattern. The data has a clear slight upward trend and repeating seasonal peaks.

Now that we can see the data and it's smoothed out, we can begin to forecast future search interest for winter tires. While we can use some extremely sophisticated tools for forecasting, we can also keep things simple because of how clearly cyclical the data is. In this case, we're going to use Facebook's Prophet library.


If you haven't seen a Prophet visualization before, the thick blue line is the predictions, the lighter blue shading is the upper and lower margin of error, and the black dots are the actual data points. Obviously we want to reduce the impact of outliers, and outliers will always appear in the real world of data, so it's okay if the blue line doesn't exactly match the black dots as we're providing accurate predictions, instead of overfitting on historical data.

In this example, we're forecasting out two years (24 months) into 2024 so if we wanted to we could look at a specific month in the future and know roughly what the interest will be for winter tires.

If we want to, we can break the data down even further to see the trend and seasonality of the data.


Looking at the top chart of the graph, you can see that from 2004 to 2022 there is am obvious upwards trend in the data. Meaning that each year more and more people are searching for winter tires. This could be due to something simple like more people shopping online (using Google) for winter tires or something more specific such as an increased awareness of the importance of using winter tires.

The bottom chart of the graph shows the swings of search interest between the months averaged out over all years in question. We can see that there is a big upswing in searches around September and October. This makes sense when you realize that in Alberta (where this data was taken from) we typically get the first few snowfalls in October or November.

We can confirm the above "theory" by looking at average snowfalls (in inches) per month to see if it roughly aligns to the above graph.


Looking at the above graph showing average snowfall by month in Alberta, you can see that Alberta tends to get the most new snow from October to November which tracks quite accurately to the data above.



So what now?


While this case study may have been obvious (people tend to want snowtires in October and November), it is important to verify these thoughts with cold hard data and even take it a step forward by using artificial intelligence and machine learning to predict future demand.

We can use the Previse engine created by Bizatta to dig down even deeper and include a number of data points so we can provide the most advanced and accurate predictions on the market, and not just for snow tires.

If you'd like to learn more about Bizatta or our Previse forecasting and analytics tool, please shoot us a message at contact@bizatta.com.