Time Series Forecasting

Time series forecasting uses historical activity to predict future results. Let’s walk through how this type of forecasting works, the data needed, and an example to put it into practice.

What is Time Series Forecasting?

Time series forecasting uses historical activity to predict future results. To build the most accurate model possible, this method looks at the following four trends in a data set:

• Secular trend, which describes the movement across the time period
• Seasonal variations, which represents seasonal changes
• Cyclical fluctuations, which corresponds to periodical but not seasonal variations
• Irregular variations, which are other nonrandom sources of variations

This forecasting method can be used on any variable that changes over time. There is no requirement for the length of time to evaluate. That said, larger data sets can increase accuracy.

Generating the Forecast

While this space is quickly becoming the domain of machine learning, you can still build a great model using Microsoft Excel and regression analysis.

Historical Data – You will need historical data for the variable that you want to predict. This should go as far into the past as you can reasonably, accurately collect to better predict trends. This is the primary driver for the “secular trend.”

Seasonal Impacts – You will need to understand how seasonality impacts your business. For example, are you selling consumer products? Your revenue is likely highest before Thanksgiving and Christmas. Running a hotel near the beach? Revenue would peak near the 4th of July. This is the primary driver for “seasonal variations.”

Economic Impacts – Is there anything in the broader economy that could impact your business beyond the items above? For example, items like Cryptocurrency, the movement towards digital, and recessions may impact your business. This is the driver for “cyclical fluctuations.”

Example:

In this example, we will evaluate a hotel near the beach that has a high degree of seasonality. For purposes of this example, we will assume no changes in the underlying drivers such as room rate or rooms available.

Step 1 – Gather historical data

Step 2 – Analyze Trends

We will analyze both the secular trend and seasonality and the slope of the line. With a high degree of high seasonality, it is necessary to calculate the secular trend using regression analysis.

Step 3 – Calculate the Forecast

Keep in mind that seasonality can affect the slope of the line. I recommend taking two approaches to the forecast to care for the impact of seasonality. If you have significant changes between the two methods, further analysis may be needed.

In the first method, we take the full-year growth of $262,236 and add it to the full year of history. Then, we use the seasonality % to spread the full-year number. In the second method, we add the growth to the prior year’s number. Reviewing the two methods, they are close to each other and both methods can be relied upon.

Have any questions on Time Series Forecasting? Are there other topics you would like us to cover? Leave a comment below and let us know! Make sure to subscribe to our Newsletter to receive exclusive financial news right to your inbox.