For Starters #56: How I Used ChatGPT in a Recent Pricing Project
A Fancy Calculator for Both Words and Numbers
FYI: Mark Stiving and I recently talked JobsToBeDone, value, and pricing on his Impact Pricing podcast
I recently wrapped up a pricing project for a 10-product suite of acquired products, all targeting 3 intertwined customer segments in the residential real estate market.
Each product had its own pricing tier somewhat compatible with the other pricing tiers. The goal of the project was to update and normalize the tiers across the suite and determine the pricing for each product at each tier. The end result was nearly 200 distinct price points spanning 1-15,000 users.
ChatGPT was my non-exclusive thought partner1 in both getting a handle on the market mechanics and in crunching the numbers.
Scrolling through my chat history, my actual prompts for this project covered both high-level market research and data analysis. On the high-level market research side:
“How many active real estate agents in the US & Canada in each of the last 5 years”
“How are the real estate agent commissions calculated?
“Describe the size of the typical US real estate brokerage”
“Detail the financial structure of the typical US real estate brokerage”
“What's the average home price in 2024 in the US and Canada?”
“How many real estate transactions does the average US real estate agent complete each year?
All of these questions helped me quickly understand the market overall and the atomic unit of value for both the brokerage and the agent.
On the data side, I had three significant datasets to work with;
a list of the top 500 real estate brokerages including; number of annual transactions, number of agents, average price per transaction, and overall sales volume
anonymized sales data for all the products over the past 4 years including; price, package add-ons, location, and number of users.
the current pricing models for all the products
By merging these datasets, we can easily get at my preferred metric of value - % of annual revenues - for every product and every customer and the actual price paid. We can also determine the break points for the pricing tiers based on the entire market. All of this is key for dialing in quantifying value and designing a comfortable pricing architecture around it.
The data analysis where I leveraged ChatGPT most heavily (yes, I would further anonymize the data prior to analysis).
“Divide the following series into quartiles/subquartiles/octiles/deciles”
“What’s the mean & median of this dataset?”
Sure, just 2 prompts, but I used them dozens of times, to really understand these datasets from every angle.
By iterating on how the dataset was divided up, I could hone in on these natural break points in both brokerage size and sales volume. The key insights from all this analysis; smaller brokerages typically have more sales per agent than larger ones. As you can imagine, this has significant implications for customer segmentation, marketing strategy, pricing models, and pricing tiers. With ChatGPT’s help, I could quickly create and compare multiple pricing tiers based on different break points and different units of value, all within the overall market. ChatGPT substantially accelerated the prototyping of multiple options to the same degree of resolution.
The final pricing architecture, delivered as an 18-sheet Excel file, is stronger and more reflective of the real world market because of it.
My overall sentiment of tools like ChatGPT remained the same through this project. I think they’re really good for helping you figure out which direction to go. Then, once you have a direction, better tools exist - Excel for example, or an actual experienced data analyst.
There were multiple times on this project imagined having a human data analyst by my side. Leveraging ChatGPT on this project helped me understand how I’d like them to help me and the size of a project too big for me and ChatGPT, et al.
I also used both Claude and Gemini here and there, I found their output equivalent to ChatGPT for my questions. So, just use your favorite until it isn’t any more. Switching is easy enough.