Get startedGet started for free

Building sensitive forecast models

1. Building sensitive forecast models and common forecast assumptions

Welcome to the final chapter of the course. In this chapter, we are going to look at how to make our forecast "smarter" by taking into account interlinked variables and have a look at the assumptions that drive our forecast.

2. Considerations when forecasting

Let's have a look at what we should consider when forecasting. Forecasting is not as simple as just plugging in numbers. We need to correctly gather and interpret data in order to make a good forecast. This includes handling changes in data, especially those that have a knock-on effect on other variables in the forecast. These are known as interlinked variables, which have dependencies meaning that the change in one variable will impact another variable in the forecast, usually with a direct relationship. There are also sensitivities, which means that a variable could change due to other factors. Usually, this is not a direct relationship but can also cause significant changes to forecast variables. We will put aside the sensitivities and dependencies for now, and focus on assumptions, how to set them, and what they mean.

3. Assumptions

An assumption is a variable that is a best guess based on the data available at the time. It is usually set at the beginning of a forecast process and used to drive the forecast process forward. For example, if a company currently employs 100 workers, an assumption is that this level should not change during the year, based on historical data. This can be something directly under a company's control, for example, it could be stated that the company will not hire more than 100 workers in the next year. Some assumptions are based on factors outside a company's control, one such example that is common in forecasting is basing your sales forecast on market indexes or other intangible aspects like market sentiment that can change based on factors outside a company's direct control. Of course, assumptions, in the end, are just a best guess. It is impossible to plan for everything. In the financial world, a "black swan" is the name given to a totally random event - one that cannot be predicted and can make any assumptions made in the forecast process invalid.

4. Different types of Assumptions

Often in forecasting, we need to understand probability. We will not cover the statistical methods of probability in this course but will have a look at something called weighted probability, which is simply calculated by multiplying each outcome by its probability percentage and adding the total result, thereby obtaining a weighted average probability for a specific calculated outcome. Another assumption that can be used is market sentiment. Market sentiment is the feeling or tone of a market, or its crowd psychology, as revealed through the activity and price movement of the securities traded in that market. This drives, for example, expected sales growth and is usually based on external market indices. There is also demand and supply assumptions that can be set by a company, which is common in industries with longer lead times for products where customers can place orders in advance or in industries with reliable sales patterns.

5. Working with pairs in Python

Sometimes, we have to build lists that are a little more complex. There is an easy way to do this in Python, which we will explore now. Take for example this table. Instead of taking two lists, the outcomes and the probability, we can create a combined list by converting the pair values into a string separated by a pipe character. Have a look at the resulting outcome, the outcome_probability list where you can clearly see the combined values. This technique is useful when needing to multiply the pairs by each other, as is done with weighted probability.

6. Define a Python Function

We will explore a basic Python function, where we can assign the dependency formulas to a function, which will become very useful later on and prevent us from retyping a function many times. So, if we look at this example, we simply define our formula for an assumption on market sentiment to the assumption1() function.

7. Let's practice!

Now let's try some examples.

Create Your Free Account

or

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.