The Many Flavors of Price Variance
One of the most insightful steps of any pricing project is identifying and investigating price variances. Price variances are price differences that exist between customers for the same offering. There are two essential price variance questions to answer:
- Do price variances exist in the dataset, and
- Which factor (or factors) is driving the variance?
Generally, the answer to question 1 will be a resounding yes. As good pricing professionals, we know that we can increase profitability by better aligning our prices with the willingness to pay of our customers. And since that willingness to pay can differ dramatically between customer segments, we use price segmentation to price offerings differently for different customer segments. If we have successfully executed our price segmentation plan, then we will definitely have price variances. However, even if there is no formal price segmentation plan, price variance will most likely appear as a result of promotions, coupons, discounts, rebates, sales negotiations, and much more. For most companies, price variances between customers is the rule.
To answer question 2, I am going to create some data visualizations to ascertain relationships between variables in the data. Visualizing these relationships will help to determine drivers of the price variance.
There’s more than one type of price variance
However, there are multiple ways to look at variance. Essentially, you are looking at how much actual prices vary from expected prices. But you can measure variance from list price, variance from average price, or even variance from target price. Each of these measures of price variance will tell you something slightly different.
For example, if you find a segment of your customers that routinely purchases at or very near list price, then you may have an opportunity to find out what makes your customers less sensitive to price. Do these purchases always involve the same customers? Do they happen at the same time of the month or quarter? Are they only for certain products? Answering these questions can help you uncover opportunities to improve profitability. Or, if you find that a sizeable percentage of your customers are paying list or near list price, then maybe your list price is simply not set high enough.
For the following visualizations, I generated some transactional data for a single product. I assumed that it had a list price of $100. I also assumed that all customers received a 10% standard distributor discount. Additionally, this company has an annual revenue rebate that offers a percentage off list price based on customer annual revenue, per the schedule in Table 1. (I also generated annual revenue for each customer and used that to calculate their target price, which is list price minus standard distributor discount minus the annual revenue rebate. Because the annual revenue rebate varies from customer to customer, the target price will vary from customer to customer.)
|Minimum Annual Revenue||Maximum Annual Revenue||Annual Revenue Rebate|
(% Off List)
In addition to looking at the 3 types of price variance identified above (from list price, average price, and target price), it is important to compare them to some standard marketing variables: deal revenue (for the individual transaction), customer annual revenue, and salesperson.
Let’s start with looking at percent deviation from list price versus deal revenue. (Because all my charts include customers with revenue across several orders of magnitude, I used a logarithmic scale on the x-axes to better see the detail.)
In this chart, you can see that transactions with small deal revenue generally get higher prices and that transactions with large deal revenue generally get lower prices. However, there is a very wide band of variation for transactions with deal revenue between $10K and $100K. This band has both the highest and lowest per unit price.
Let’s look at % deviation from list price versus customer annual revenue.
Chart 2 tells a story very similar to Chart 1: customers under $100K are generally getting a smaller discount, customers above $1M are getting a larger discount, and prices for customers in between are all over the place.
These charts demand investigation. There is a cluster of customers with almost $100K in revenue that is getting a 20% discount from list price. Should their discount be reduced to between 10% and 15% (like the similarly sized customers directly above them)?
What about the customers with just under $1M in revenue that are getting better prices than the customers with well over $1M in revenue? Or the customers with around $700K in revenue that are easily paying more than any other customer? Is this price variance sustainable?
Price variance from the average price
We’re going to switch gears here and look at the price variance from the average price. Let’s stick with comparing it to customer annual revenue.
You’ll note right away that the y-axis has changed from all negative values to a mixture of positive and negative values. For Charts 1 and 2, 0% on the y-axis indicated that the item was sold at list price. For Chart 3, 0% on the y-axis indicates that the item was sold at the average price (or $81.09).
Although the data points in Chart 2 and 3 are oriented the same, changing our y-axis allows us to very easily determine which transactions occurred above the average price and which occurred below the average price.
All transactions for customers with more than $1M in revenue occurred below the average price. Except for the small cluster of customers with almost $100K in revenue, all transactions for customers with less than $100K in revenue occurred above the average price. And the customers in the middle have prices both above and below the average price.
Price variance from the target price
The final type of variance that I am going to investigate is price variance from the target price. As mentioned above, the target price changes from customer to customer based on their annual revenue. Larger customers are expected to get a lower price due to the higher rebate percentage, so their target price will be lower than for a smaller customer.
Let’s continue with plotting the target price variance against customer annual revenue.
As before, changing the price variance changed our y-axis. Now all transactions above the 0% line exceeded the target price (for that specific customer), while everything below the 0% line failed to meet the target price.
In some ways, looking at the variation from target price is the most sophisticated of these three measures of variance. The variation from list and from average price are based on static signposts: there is a single list price and a single average price. The result of every transaction is compared to the exact same number.
However, we know that not every customer should be treated the same. If our business model encourages us to offer a larger discount to a customer once they reach a certain size, then our target prices should be adjusted to reflect that strategic decision. Of course, one problem is that not every company has assigned target prices for each of its products.
Let’s look at how the representation of the data has changed. In Chart 3, the highest price was from a customer that was paying 17% above the average price, and the lowest price was from a customer that was paying 13% below the average price. In Chart 4, the highest price was from a customer that was paying 11% above their target price, and the lowest price was from a customer that was paying 17% below their target price.
The prices have not changed. Only the calculation of variance has changed. A major advantage of looking at the variance from target price over the variance from average price is that the focus is not on what customers are paying on average. Rather, the focus is on performance versus the target price that management sets. It’s the difference between what happened and what management intended to happen. The conversation can turn from questioning why the lowest price is 13% below average to why the lowest price is 17% below target.
The final marketing variable that I will investigate is the salesperson. I will stick with variance from target price. Because salesperson is a category, I will switch to using a boxplot (also called a box-and-whisker plot). A boxplot allows you to quickly visualize the distribution of different groups of data.
A boxplot breaks your data up into quartiles. Because I created a revenue-weighted boxplot, each quartile corresponds to one quarter of the salesperson’s total revenue. The top whisker is your first 25% of revenue. The orange box is the revenue between the 25th and 50th percentile. The gray box is the revenue between the 50th and 75th percentile. The bottom whisker is your last 25% of revenue.
There are a few things of note. First, the line separating the orange and gray boxes is the median. Comparing the median for two salespeople allows you to compare the centerpoint of their data. I usually start by looking here. Second, the orange box and the gray box combined give you the middle 50% of the salesperson’s revenue (because it is from the 25th to 75th percentile). It’s like the centerpoint, but with additional detail. Finally, the height of the boxes and whiskers let you know how dispersed the revenue is. A short box or whisker indicates very concentrated pricing, while a long whisker or box indicates very dispersed pricing.
As you can see, Amanda has larger negative deviation from the target price than the other salespeople. Her median deviation is -7%. In fact, her median lines up with the bottom whiskers of the other salespeople. This means that her 50th percentile would be in the bottom 25% of the other salespeople. Additionally, her bottom whisker extends to -17%, which is far lower than any other salesperson.
This finding would require further investigation, of course, but it appears as though at least some of the price variance from the target price is due to variances between salespeople. I expect that this is driving the wide price variation for customers with $100K to $1M in revenue in the scatterplots above. Additional products would need to be checked to see if the price variance between salespeople is consistent.
I hope that this review of price variances has revealed just how effective asking the right questions about your dataset can be. After all, when crafting a corporate policy, it is important to translate a company’s reality into pricing rules that can be easily understood and applied across the board. Looking at price variances in a few different ways can help you better understand that reality. And understanding it will better prepare you to move the pricing discussion forward.