#Post 8 - Demand forecasting - Tools to get your assortment right
In the previous post on assortment planning, I mentioned that assortment planning is a crucial part of category planning.
If you running an established category, that means you will already have transaction data / sales data, that will help you to figure out what next. Although you will have a demand planning team if it is a big retail chain to help out with the forecasting method, it’s good to know the fundamentals, factors that might affect your model. If you are working for an e-commerce startup or setting up your own store, it’s good to know how much to order.
So first of all, what is demand forecasting. Demand forecast is the estimate value of the future demand.
For the sake of simplicity, let’s just say in the month of Jan, Feb, March, April you sold 18, 16, 18, 15 running shoes respectively. Assume there is not much of a seasonal variation in the purchase (people like to run every month more so in the month of Jan when there are some promises to be made to yourself in the form of new year resolution), it’s safer to assume that you can keep ~16 shoes in your inventory. (range 15-20)
It’s not as simple as above calculation but it gives you the flavor of forecasting. There are safety stock to be considered to avoid opportunity cost. There are seasonality. There is weather, promotions, economic condition to be factored into as well. But one day at a time :)
So in nutshell, demand forecasting is the basic input in your decision-making process for assortment planning.
There are multiple method to do forecasting, but we will focus on two techniques mainly.
Associative forecasting technique
In this technique, you identify related variables that impact the variable in question.
Let’s come back to the shoes again. Sales of running shoes may be related to the weather condition (people might not go running when it’s raining or it’s hot summer), endorsement deals you just signed with one famous athlete, or the fact that your competitors have recently launched a new product in the market.
Once you have your independent variables in place, all you gotta do is to develop an equation. And, of course you need data as a raw material for your forecasting model. You can use regression (tools - R, excel) to run the analysis.
Shall we put theory into the game?
We will take only one independent variables, and one dependent variables (Sale of Shoes) for this example. You have noticed that the sales of shoes (in volume) depends on the amount you have spent on promotion activities in dollars. Let’s estimate the no. of shoes you will be selling when you are spending $800 this month.
Here is your data.
When you run regression using excel, you will get this output.
You can go through this page - https://www.ablebits.com/office-addins-blog/2018/08/01/linear-regression-analysis-excel/ to know more how to run regression on excel and about correlation coefficient, R square and all that, because that would be important to know how good your model is.
For this post, we will stick to the work in hand.
Here is your equation
y = 7.32+ x (0.0418)
When you 800 into this equation, you will get ~41. So here you go, order 41 shoes.
As you keep adding other variables, the complexity of model will increase. But for now, we know the basics.
Time series forecasts
Basically, this method is about project future based on past experiences. You identify the pattern in the past and extrapolate for the future.
Above example where we have seen monthly data for shoes and then extrapolated it for the next month, that’s your time series.
Moving average
Weighted moving average
exponential smoothing
trend-adjusted exponential smoothing
using seasonal relatives
The selection of technique will depend on cost, available time, available data, and category for which you are forecasting.
At this point, I should also tell you about the concept of Margin of Safety by Benjamin Graham that I found true for anything in life including the stock level you will be maintaining.
he said that,
The purpose of the margin of safety is to render the forecast unnecessary.
Accuracy in forecasting is difficult to achieve so leave the room for errors. Consider forecasting as a range rather than a fixed number.
In upcoming posts, we will do other methods as well. But in short, you don’t need to learn all. You just need to know that you will be working with demand planning team to know the numbers, accuracy level etc. and work with that.
Do leave a message if you come across any challenging problems related to demand planning for your category.