1. Demand curve: meals and drinks
In this lesson, we'll explore another objective, supply vs. demand. You'll focus on exploring the demand to then predict the supply needed.
2. Understanding demand, helps you ensure supply
Supply and demand are two key aspects of data-driven operations. Data and data science methods underpin both but it is up to you as the decision maker on how to incorporate and adjust your operation. Once you have a grasp of the demand, you can see how much supply to produce. That supply can be a physical material like wheat for bread or it can be cashiers for a checkout line.
3. Time series arrival patterns
When you manage a system with a cyclical demand, it's important to know the peaks and valleys.
For example, you may want to schedule enough people for the busiest times within a 30 minute interval and on a specific day. Additionally, employee breaks, training and system upgrades should take place in non-peak times to minimize the effect on customers.
Remember it's not just call centers, or check out lanes, arrival patterns impact traffic, website visitors, streaming services and many other modern applications.
To understand a time series pattern, it's a good idea to visualize it. A time series chart has time as the x-axis. The y is any measured amount like phone calls, customers coming to a drive-through, or online viewers. In the xy plot the number is illustrated as a line.
4. A time series chart
Here is an example of a time series chart. The days and specifically 30-minute intervals are represented in the x-axis. The number of customers at a business are shown as connected points of a line. You can see a repeating pattern, though not perfectly repeating for each of the days because there is a clear peak happening within similar time frames.
5. Supply and demand
This pattern, shown in a time series chart, represents the demand on a system like a call center. Based on this demand, operational professionals must create a corresponding supply of agents to meet the demand. The arrival pattern is captured in an interval or specific time period, often 30 minutes.
For example, when you buy groceries and not enough checkout lanes are open, the system demand, or number of shoppers for that particular time frame, may not be forecasted properly.
System demand on many of these operations such as website traffic, call centers, or even elevators can be understood in an arrival pattern.
Here is an arrival pattern for a call center where incoming calls are measured every thirty minutes. These calls are the demand on the system and managers must schedule agents to answer the phones against this arrival pattern.
6. Use common sense
Sometimes, new data scientists and forecasters will use the arrival pattern to predict an application's future demand. However, if the data scientist is too far removed from the practical implementation then mistakes can occur. It's always good to apply sound judgment, not merely accept the output of a mathematical formula.
Common pitfalls include an arrival pattern that trends the wrong direction, has floating point numbers when the operation needs integers or even contains negative numbers.
For example, you can't have a fraction of a customer at a checkout, and similarly if the trend of the arrival pattern is going down, you can't have negative customers either.
7. A questionable forecast
Here is an example of a time series chart requiring you to ask questions. This chart shows a forecast in red that is at odds with the upward trend shown in black. While this trend reversal may be justified, it is worth asking the forecasters why this is expected versus immediately accepting it. Additionally, the forecasted repeating peak to valley seems OK but it does seem to be exaggerated in the red section compared to the historical pattern in black. Again, there is no harm in asking why this could be when you have to make an operational decision.
8. Let's practice!
In the next exercises, you'll have to demonstrate understanding of the demand side of a supply and demand used in a real-life system.
In the next lesson, we'll transition to the supply side so you can make data-driven staffing decisions. Let's get to it!