Synthetic Data in Pricing: Trust It, Test It, or Ignore It? with Steven Forth

Steven Forth is a pricing strategist and AI innovator with decades of experience building value-based pricing models.

As the founder of Value IQ, he blends rigorous pricing theory with emerging AI applications—often pushing the boundaries of how pricing professionals think about data, modeling, and buyer behavior.

In this episode, Mark and Steven step into another live debate aka 'intellectual challenge' about AI-generated synthetic data with real pushback, not polite agreement.

They challenge whether synthetic data is a breakthrough for pricing or just smarter-looking "fake data" that distances us from buyers.

What unfolds is an unscripted stress test of the idea itself, and it ends with a surprisingly human conclusion you should definitely listen to.

What You'll Learn in This Episode:

• What synthetic data actually is—and how it differs from simply "making up numbers."
• Where synthetic data becomes dangerous, especially when assumptions about buyer behavior go untested.
• Why even the most advanced AI modeling cannot replace direct conversations with buyers.

"Go out and talk to buyers and understand their buying process."

– Steven Forth

Topics Covered:

00:00 – Why synthetic data is suddenly a pricing topic. Steven introduces Value IQ and the idea behind AI-generated pricing intelligence. The setup: why synthetic data is gaining attention—and why Mark is skeptical from the start.

03:45 – What is synthetic data (without the buzzwords)? A plain-language definition of synthetic data and how it differs from CRM or ERP history. Why backward-looking data limits pricing strategy.

06:30 – The "fake data" objection. Mark challenges the idea head-on: Isn't this just inventing numbers? A sharp exchange on statistical misuse, p-values, and the danger of generating data that simply confirms what you want to see.

09:30 – Interpolation vs. extrapolation in pricing models. Why most pricing data isn't normally distributed. Discussion of fat tails, clustering, segmentation signals, and what synthetic data might distort—or reveal.

12:30 – The three types of synthetic data. Steven outlines three practical applications. (1) AI-generated buyer simulations. (2) Stress-testing value and pricing models. (3) Modeling competitive and economic scenarios. This is where the conversation moves from theory to use cases.

16:30 – Can AI predict buyer behavior? Mark pushes the core issue: pricing changes behavior. So how can synthetic data anticipate it? A discussion about assumptions, validation, and ground truth.

20:00 – A practical example: AI-driven Van Westendorp studies. A concrete scenario: simulate 100 real buyers, test pricing sensitivity, validate with actual survey data, and refine the model. A tangible way to experiment responsibly.

23:30 – The risk: Are we moving further from real buyers? The philosophical tension of the episode. Does synthetic data create insight—or another buffer between pricing teams and customers?

26:30 – The surprisingly human conclusion. After 25 minutes of AI debate, Steven's final advice is simple and grounded: talk to buyers and understand their buying process.

29:00 – Closing thoughts and where to connect. How to reach Steven and Mark—and a final reminder that AI is a tool, not a substitute for customer insight.

Key Takeaway:

"Synthetic data is data that is generated for you by your AI." – Steven Forth

"With synthetic data, you can explore scenarios that do not yet exist or parts of the market you do not yet touch." – Steven Forth

People and Resources Mentioned:

• Craig Zawada – Former McKinsey partner, co-creator of the pocket price waterfall; now Chief Strategy Officer at PROS
• Benoit Mandelbrot – Referenced in the discussion about fat-tailed distributions and why pricing data is often not normally distributed.
• Pocket Price Waterfall – A pricing analytics framework originally developed at McKinsey.
• Van Westendorp Price Sensitivity Meter – Used as a practical example of how synthetic data could simulate buyer responses.
• Conjoint Analysis – Discussed as a potential future application for synthetic respondents.
• Bayesian Updating / Bayesian Statistics – Mentioned as a way to iteratively improve models by aligning synthetic data with real-world results.
• Interpolation vs. Extrapolation – Statistical concepts debated in the context of synthetic modeling.
• Normal vs. Fat-Tailed Distributions – Discussion on why pricing data often violates normal distribution assumptions.

Connect with Steven Forth:

• LinkedIn: https://www.linkedin.com/in/stevenforth/
• Email: steven@valueiq.ai
• Subscribe to Steven's Substack: Synthetic data in pricing: https://pricinginnovation.substack.com/p/synthetic-data-in-pricing

Connect with Mark Stiving:

• LinkedIn: https://www.linkedin.com/in/stiving/
• Email: mark@impactpricing.com