blog | 4 min read

How to Make "Average Order Value" a Valuable Metric

January 1, 2020

Teal datawave pattern
Editor's Note: Enjoy this blog from the Custora archive, acquired by Amperity in November 2019.

Companies often want to track things like the size of first orders and the size of repeat orders. For example, you might put a plan in place to increase how much your customers spend per order — that’s one way to grow revenue.

As a result, you might monitor the average order value of repeat orders (often described as AOV). Also, you might inspect the AOV of different types of customers to learn how different customer segments interact with your business.

But, as simple as AOV sounds, there is a common mistake that people sometimes make. There are a few different ways to calculate AOV, and, as usual, it can be dangerous for companies to just throw everything together and find the average. The important thing to keep in mind here is that there’s a subtle (but important!) difference between "average repeat order size" and "the average of each customer's repeat order size."

Here’s a simplified example to illustrate the distinction. Imagine you have two customers:

Customer A makes 2 repeat orders, each for $10.Customer B makes 10 repeat orders, each for $20.

If we want to determine the average repeat order size across all repeat orders, regardless of customer, we'd divide $220 (the total spent on repeat orders between the two customers) by 12 (the total number of repeat orders) and get $18.33.

But what if what we want to know is the average repeat order size for a given customer? That's a different story. In the example above, our answer would be $15 (One customer averages $10 and the other averages $20).

So why does it matter which method you use? On the surface, the difference may seem inconsequential but consider another example.

Let’s say the marketing team at Socktown is working to increase revenue with their existing customers. To get started, they want to understand how their customers are currently behaving.

Socktown knows that customers spend $40, on average, on their first order. They follow the first approach above – calculating the AOV across all repeat orders, regardless of the customer – and learn the AOV on repeat orders is $50. This looks promising, and the team might conclude that customers are spending a healthy amount on their repeat visits.

However, when they normalize things by taking each user's average and then average that value, they see the AOV per customer is $30.

This paints a very different picture!

So why would this be the case? This is where things get interesting.

One possibility is that Socktown customers who order often tend to have increased repeat purchase sizes — just as we saw in the first example above. These two factors (frequent repeat orders and increased size per order) actually “weigh up” the pool of repeat orders, which gives us the first, higher value.

However, that higher number obscures the important fact that majority of Socktown customers are in fact spending less on their repeat orders.

By focusing on per customer statistics instead of overall average statistics, the Socktown team learns some interesting and perhaps counterintuitive insights about their customers.

First, customers who purchase frequently also purchase more per order. A sensible next step would be to drill in deeper with analysis on these customers. Which acquisition sources attract these customers, and what types of products do they buy?

Second, a large percentage of the population has a relatively low average purchase size. The team might begin to brainstorm strategies to encourage these low-AOV customers to increase their shopping cart basket size. More importantly, the team also can run marketing experiments targeted towards these low-AOV customers to learn which strategies resonate.

If your retail marketing team is interested in these forms of customer-level analytics and targeted marketing experiments, you can start by taking a product tour or requesting a demo.