Thread by @Patticus

The most common question I get asked:

How do you determine customer willingness to pay and pricing?

Here's how 👇

People treat pricing like sex, religion, and politics - "oh no we shouldn't talk about that"

In reality, if you don't have massive traffic (and if you think you do, you probably don't) the only way to determine willingness to pay is to talk to your customers.

2/

You have to ask properly though.

We'll take advantage of how the human brain thinks about value, which is based on a spectrum.

We know that a bottle of water is worth less than a computer. Go to the desert for 3 days without water, and the value differential shifts. 3/

So we'll ask questions like:

- At what point is this way too expensive?
- Getting expensive?
- A good deal?
- Too cheap that you'd question the quality?

These are known as the Van Westendorp questions and the model is highly efficient. 4/

When I plot the answers, I get this output. The four lines correspond to the questions - furthest to the right is "way too expensive" while furthest left is "too cheap".

The intersection would be the price you'd need to be at to get as much of the segment as possible. 5/

You also get an elasticity curve, which allows you to calculate the % of sales won/lost by moving the price up or down.

Is this model perfect? 6/

No, but that's the wrong question. You're never looking for perfection. You seek to minimize risk in a decision.

VW's questions were the innovation. His calculations will get accuracy of +/- 20%.

If you spin your own math with the questions, you can get down to +/- 5%.

7/

Other methods exist though (conjoint, gabor granger, etc), but we've found VW to be the most efficient - you get the most output vs the amount of work/cost.

This is why you should be weary of anyone who throws method names around or says "this is all going to be biased"

8/

News flash: All data has bias.

You reduce that bias by proper questioning, data cleaning, and model building, as well as increasing your sample size. You never eliminate bias.

It's more important to understand the limits of your data.

9/

If I'm making a $10M decision, I'll spend money and time to collect lots of data.

If I'm doing very early research, maybe I just ask a few folks at which price is it way too expensive and when it's a great deal.

If I'm making a $100M decision, I pull out all the stops. 10/

Other fun objections to collecting pricing data:

"omg people won't be honest"

Ranged questions, large sample size, outlier analysis, etc all control for this.

Also, you need to collect data from customers, prospects, and targets who've never heard of you.

Why? 11/

Roughly 80% of you reading this did 0 research and then insecurely priced your products too low, so your current customers are poorly anchored. We have pricing software and did this, so don't feel bad.

This is why @patio11 always tells you to charge more. :)

12/

More objections:

"ugh people hate surveys"

Hate to tell you, but - People hate surveys, because we're bad at sending surveys. You email 40 question surveys to me where the first question is "what's your email?"

Don't hate the game, hate the playa.

What's a good survey? 13/

If you're doing a non-compensated survey (raffles don't count as compensation), your survey needs to be 30-60 seconds long. Compensated can be 15 minutes long.

We've sent tens of millions of these things at this point. They're *super* effective. 14/

In sum:

1. Talk to your customers about pricing, just ask in the right manner.

2. Segment data as much as possible.

3. Understand the limits of your data

4. Know you're still going to have to make a decision, but pricing data (and other data) will inform the decision

15/

Ok, back to my analyst hovel:

If you reached this far, retweet the parent tweet. Want to see who got through the whole thing.

Also, let me know where you want me to go deeper. fin/