Often, managers use average delivery times to make future predictions. Even though this method is intuitive and widely spread, making delivery commitments based on averages is not reliable. Let’s explore the challenges that come with this approach. 

What Are Delivery Times Averages?

When we talk about using averages to make delivery predictions there are three main values we refer to – the mean, the median and the mode.

The Mode is the easiest average to calculate – this is the number that appears most often. Since that’s the most commonly occurring delivery time, if you ask a team how much time they usually need to complete a task, this would be the answer.

The Median shows the middle number of a data set. It denotes that half of the tasks completed so far have taken less than the median value to be finished. However, the other half have taken longer to complete.

The Mean is the average calculation that you are most likely to be familiar with. This involves adding up all of the values and dividing them by the number of instances in the data set.

Most often, the average we use to make a delivery prediction is either the mode (the most common value) or the mean (the arithmetic average). 

Why Using Averages to Make Delivery Predictions Will Land You in Hot Water

Making predictions based on an average is highly likely to land you in hot water. Delivery forecasts based on averages only make sense if you know something about the shape of the underlying distribution of your delivery times.

Let’s explore a couple of examples.

Using averages to make delivery predictions - fat-tailed distribution

This is a Cycle Time Histogram exposing a high-variability delivery workflow. The Cycle Time Histogram shows the frequency distribution of the completion times of the tasks in your workflow. The vertical axis displays a frequency and the horizontal axis shows your cycle times.

A Cycle Time Histogram with a big hump on the left and a very long tail to the right indicates that your cycle times vary significantly. This means that your process is inconsistent and you’re maintaining an unpredictable system.

Here, the mode points to 1 day, the median is 9 days, and the mean –  21 days. The most popular delivery time is 1 day, and the tail extends all the way up to 130 days. If you were managing this team and you were about to commit to the average value (the mean), you may end up with a delivery time that is 6 times higher than what you’ve promised to achieve.

If you don’t know the distribution of your delivery times, there is no way that you can give a probability of where the average falls. If you don’t know the probability, then you cannot make a reliable delivery prediction. There could be a 20% or 50% or 80% chance of meeting your commitment.

Would you commit to a delivery date if you know that there is only a 20% chance to keep that promise? Probably not, right? It certainly wouldn’t be something that we’d recommend doing.

Using the Frequency Distribution of Your Delivery Times Effectively

Now, assume you know the frequency distribution of your delivery times. Let’s analyze the following example.

Using averages to make delivery predictions - thin-tailed distribution

Here is the cycle time frequency distribution of a mature team that maintains a stable system. 

A key point to note here! The more stable your system is, the more predictable it becomes. And predictable systems produce more accurate delivery predictions. If you’re interested in adopting the practices and the proven strategies to establish a stable system using a step-by-step roadmap, we couldn’t be more thrilled to welcome you to our Sustainable Predictability program.

In the example above, all the averages are very close to each other – the mean is 7.20 days, the median (50% of the cases) points to 7 days and the mode is 9 days.

Using their frequency distribution, this team can say that there is a 50% chance of finishing any work item in less than 7 days.

By committing to that delivery prediction, what they say is there’s an equal likelihood that they’ll either make it on time or not. There is a 50/50 chance of making that happen. The risk is considerable.

In order to provide a reliable commitment, what you need to do is to come up with a set of delivery times and the probabilities that come with each of them.

Using the example above, what this team should do is to deliver a probability forecast that looks like this:

Using averages to make delivery predictions - probabilistic forecast

Now, it’s up to your customers to decide the level of risk that they are willing to take, and which probability they feel most comfortable with. When you’re maintaining a stable system, that decision will be quite an easy one, as the values that come with each probability will be very close to each other.

It’s important to remember that your forecast is a living thing – it will change as you discover new information, so it’s crucial to reevaluate it on a regular basis. You will need to adjust your course accordingly to be able to hit your goals.

Switch to Probabilistic Forecasting As Soon As You Can

Let’s make this clear, what we are saying is that using averages to make delivery predictions is not a reliable approach, not that it’s not applicable at all. If you’ve been relying on estimating and guesswork so far, this method is something you can transition to as a starting point. However, strive to switch to probabilistic forecasting as soon as you can.

Using averages to estimate your work can land you in hot water, as this method communicates one single certain commitment. To produce a reliable forecast, you need to provide a range of delivery times and the probabilities that come with meeting each of them. That’s the most reliable way to establish realistic expectations and deliver on time!

Do you find this article valuable?
Rating: 5 stars (5 readers voted)