Researching What Customer's Value

A typical challenge many companies face is to figure out what price they should charge for something, especially if the product or service is unique or different.

I have written about this before, in these two blogs:

The Van Westendorp Price Sensitivity Meter
https://www.davidabbottspeaker.com/blog/2021/08/10/2022-12-10-pricing-your-new-product-or-service

The Gabor-Granger Pricing Model
https://www.davidabbottspeaker.com/blog/2023/04/28/2023-4-23-the-gabor-granger-pricing-model

I want to introduce you to another model, the Becker-DeGroot-Marschak method. This might be my favourite approach.

But before we dive into that, I want to look at some of the limitations of traditional research where we just ask customers what they’ll pay. And in fact, I want to extend the discussion of research into understanding value too.

Why? Well, I’m a huge fan of value-based pricing. Even if you are going to use one of the models to establish willingness to pay, it helps to have a good feel for what a fair price based on value should be. It’s a great starting point.

This whole topic is quite long, so I’ll focus on research this month, and the Becker-DeGroot-Marschak method next month.

Typical market research asks customers either what they’ve done in the past or what they think they’ll do in the future.

For example, given a new product or service, the research might ask the simple question ‘what would you consider a fair price for this?’. Or, if the research is trying to understand what customers value, it might ask the customer to either rank features in order of importance to them, or ask them to assign a score out of 10 to each feature.

There are three problems with this type of research.

First, when asked a question the responder engages their conscious logical brain to think about the answer. But that’s not how we act in normal life. When we encounter a product or service our subconscious makes the decision and hands that over to our conscious mind, which largely then rationalises why it was the right decision all along.

Second, we imagine a perfect version of ourselves. ‘I’d never do such-and-such!’ or ‘I would always do this particular thing’. But when we’re in the moment, we act differently. For example, prior to the Oasis reunion tickets going on sale, fans had very specific ideas about the maximum they would pay for a ticket; but after they had queued for hours online and finally got a chance to buy a dynamically-priced ticket which was much higher than planned they still bought it anyway. The excitement was too much, or they’d queued so long it seemed a shame to come away with nothing.

Third, we might want to influence the outcome. If being asked what a fair price is, we might deliberately quote a low figure to sway the final price.

So just asking a customer ‘what’s important to you?’ or ‘how do these features compare in terms of value?’ is a challenge.

If we want to understand what features customers value, and how much they value each of them, then there are ways to get around this.

One approach is to ask either/or questions. Which is more important to you, option A or option B? If you have 9 different features to compare, then there are 36 possible combinations of ‘option A or option B’. For the mathematically minded, the formula is [n(n-1) / 2]. You then count how many times an option is selected as the preference. The one selected most often is weighted 10, and all others are scaled to that weighting.

For example, let’s say you just have 4 options. That’s 6 either/or questions (AvB, AvC, AvD, BvC, BvD, CvD). You ask 100 people to do the survey. The results are:
Option A - chosen 184 times
Option B - chosen 43 times
Option C - chosen 276 times
Option D - chosen 97 times

Then option C is weighted 10; option A is weighted 7 (184/276 = 6.66); option D is weighted 4 (97/276 = 3.51); and option B is weighted 2 (43/276 = 1.56)

A similar but slightly different approach is to offer a choice of four things at a time, and to ask the respondent to choose the most and least favourite item.

To illustrate this, imagine you have 8 features, options A to H. There are 70 ways to choose 4 different combinations of options (again, for the maths inclined, the formula is [n! / k!(n-k)!], where n is the total number of options, and k is how many we want them to choose from).

Let’s say one of the choices is between options A, C, D and G. The respondent likes D the most and A the least. D then scores +1, and A scores -1. At the end of the 70 questions, you will have a range of scores that might look like this:
Option A: (-17)
Option B: (-9)
Option C: +14
Option D: +42
Option E: (-61)
Option F: +3
Option G: +12
Option H: +16

I have put the negative numbers in brackets to make them easier to see.

Now we want to turn this into a weighting from 10/10 to 0/10. The scale runs from -61 to +42, a range of 103. We want to know where everything in that range sits, so we add the largest negative number to each number: we add 61 to -17, we add 61 to -9, add 61 to +14, etc.

Our results now look like this:
Option A - 44
Option B - 52
Option C - 75
Option D - 103
Option E - 0
Option F - 64
Option G - 73
Option H - 77

If 103 is 10 out of 10, then the rest of the scores are:
Option A - 4
Option B - 5
Option C - 7
Option D - 10
Option E - 0
Option F - 6
Option G - 7
Option H - 7

Another rigorous alternative is discrete-choice or conjoint analysis. By presenting respondents with realistic bundles that mimic a real shopping situation – complete with trade-offs on price, performance and brand – you collect revealed-preference data that can be modelled to predict demand curves, simulate ‘what-if’ scenarios and estimate willingness-to-pay without relying on any single, self-reported figure. This is quite complex, and might require a separate blog all on its own.

There are some things you must be careful about when conducting surveys.

It’s critical to ensure that whoever completes your survey is as representative of your real customers as possible, and that you get a big enough sample size.

Next, the way questions are framed can anchor judgements and compress the scale. The order in which options are presented can distort results, so it’s important to randomise the order for each participant. It’s also important to make sure there are no accidental anchors, e.g. mentioning a number which then influences what respondents think they might pay for something.

All of this helps you to understand how customers value different things. That can help you to work out what price they might pay for the value you are delivering.

But before you go to market you want to test the value-based price you have just come up with. Will customers really pay it? You can’t just ask them ‘would you pay this price?’, for the reasons outlined above.

So what do you do? You could try the Becker-DeGroot-Marschak (BDM) pricing method, and we’ll explore that next month.