Seasonality in 3 minutes

Hey! One of the things we have noticed since we started LokadTV is how confusing some of the fundamentals of supply chain science can be. There’s loads of resources out there but none of them get straight to the point and explain things simply and concisely. As such we have decided to have a go ourselves and explain everything you need to know about supply chains in about the amount of time it takes to have a coffee break……so let’s get started with our first topic, Seasonality!

So seasonality is simple right? Everyone knows that for a toy manufacturer sales are likely to spike around Christmas. But what people don’t consider is the fact that seasonality actually impacts the vast majority of human activities (text overlay). Everything from traffic to electricity consumption to even death can be proven to be seasonal to some extent (insert pictures)…..For this reason seasonality is one of the most powerful statistical patterns in a Supply Chain Scientists arsenal and it can be exploited to make sure that you’re fully prepared for every eventuality.

So let’s have a look at some examples starting with one of my favourite subjects….ice cream!

  • If we look at this graph of ice cream sales - much like our Christmas example we can quite obviously see sales spike during the summer months
  • But how about if we zoom in on one month of that busy summer period - now we can see that sales are not consistent throughout the week. This could be due to fluctuations in the weather…….but there are also obvious spikes in sales around the weekend. This means that as well as the time of year, the day of the week also has a seasonality which we have to take into account too!
  • The key lesson to learn here is that a product exhibits seasonality when the underlying time-series undergoes a predictable cyclic variation (text overlay) - so what we have observed so far are both yearly and weekly seasonalities.
  • And if we zoom in even further, perhaps on the day we can see there is a lot of variability. This is why our graph is not very smooth and as a supply chain practitioner it can be very confusing.

This variability is known as noise and is one of the reasons why computers are better than humans at predicting the future. Often when companies are trying to predict how much stock they are going to need they focus on these statistical outliers, ensuring that they are prepared for every eventuality resulting in a lot of money wasted on unneeded stock and valuable warehouse space. However machines don’t have the same worries. If there is a stock out they probably don’t need to worry about losing their job! That’s why they are so good at averaging out this noise and focusing on the trends that are actually important.

What we have seen so far is that seasonality can lead to uncertainty in amplitude - that’s the size of those spikes in sales we saw. Another complication is that seasonality can also lead to uncertainty in frequency. Some seasons start early, some much later…..so our spike in sales can occur at different times depending on a number of exterior factors. Some of these can be easy to predict.

For example the last Saturday before Christmas is normally one of the busiest shopping days of the year. But what happens if the Saturday fell on the 23rd December. This might be a little late for many of the more nervous shoppers which would push that big spike in sales back to the next Saturday before Christmas leaving the last Saturday to those risky last minute shoppers.

That might be fairly easy to predict however some patterns can be harder to predict, a good example of this is with the weather. This year we have already seen a freakishly early warm spell of weather in February. This could result in a huge boost in sales of flip flops and sunglasses with shoppers getting ready early for the summer whilst leaving store owners with piles of unsold hats and scarves that would have been snapped up when the ‘Beast from the East’ hit last year. Unfortunately nobody can predict the future exactly.

That’s why instead of making one prediction for your stock requirements it’s much better to hedge your bets and take a probabilistic approach - this is the only way you can be sure you’re protecting yourself against every eventuality.

So what have we learnt today?

  • Seasonality is a repetitive cyclical trend that can be shown to have an impact on almost every part of daily life.
  • This can impact on both the amplitude and frequency of sales which can get far too confusing if you try to keep track of it yourself……even with the help of excel
  • By leaving the hard work to computers and by taking a probabilistic approach you can ensure you cover all possible futures…leaving you to focus on the important things in life, like ice cream!

Find out more about seasonality