THE SLIP: "What If?" causal models are a fast way to de-risk your 2025 plan. And it's never been easier, thanks to OpenAI and Proof Analytics

A B2B CMO saw a podcast where I discussed counterfactual models, which are a specific modeling technique used to explore "what if?" scenarios and create "Blue Ocean" solutions. He asked if I'd share a detailed outline of how to do it. Given that part of the process involves ChatGPT4 and later editions, I thought I'd use it to first outline the general process and then ask it to give a basic -- but specific -- example.

Here's the high-level: as a result of the combination of ChatGPT4 and Proof Analytics, you can generate a spread of counterfactual models in a matter of hours.

Uses cases range from a pre-rev company looking to firm up their pro forma financials for investors to a GTM team wanting to explore and test their plans before they spend any cash on them. It's also a terrific way to de-risk your 2025 plans while creating a de facto template for converting the synthetic data to actual data.

A few comments about the synthetic data generated in ChatGPT. After well over a hundred models myself, I can say that the representative quality of well-prompted synthetic data is very high. You can specify as many rows and columns of data in any category of measurement that you might require.

A nuance here that adds a lot of value is to ask ChatGPT to help you identify any significant marketplace variables that impact your business performance, and then assemble representative data for a range of volatility over the course of the scenario. I'll be writing about that soon, and also about how to model the impact of Brand and Reputation on your business -- those too have been hotly requested!

This is detailed work that you need to practice, but it is not hard, and it teaches us so much about how to think.

The following is verbatim from ChatGPT:

Step-by-Step Outline for Creating a Counterfactual Model Using ChatGPT-4 and Proof Analytics

Counterfactual modeling involves creating a hypothetical scenario to examine a alternative scenario and / or compare what actually happened with what could have happened under different conditions. Integrating ChatGPT-4 and Proof Analytics enables a powerful combination of AI-driven insights and analytical modeling for decision-making. Here's how to proceed:

Phase 1: Define the Objective

1. Identify the Purpose:
2. Clarify the Scope:

Phase 2: Gather and Prepare Data

1. Collect Relevant Data:
2. Ensure Data Quality:
3. Integrate Data into Proof Analytics:

Phase 3: Create the Counterfactual Framework

1. Define the Counterfactual Scenario:
2. Model the Baseline:

Phase 4: Use ChatGPT-4 for Hypothesis Generation

1. Brainstorm Counterfactuals:
2. Generate Insights:

Phase 5: Build and Run the Model in Proof Analytics

1. Set Up the Counterfactual Model:
2. Run Causal Analysis:
3. Validate the Model:

Phase 6: Interpret and Visualize Results

1. Analyze Key Insights:
2. Generate Recommendations:

Phase 7: Communicate Findings

1. Prepare a Comprehensive Report:
2. Present to Stakeholders:

Phase 8: Implement and Iterate

1. Apply Learnings:
2. Iterate on the Model:
3. Replace Synthetic Data with Natural Data:

Tools and Resources

• Proof Analytics: For causal modeling, analytics visualization, and scenario analysis.
• ChatGPT-4: For hypothesis generation, insight extraction, and narrative creation.
• Data Sources: CRM systems, financial data, external market reports.

By combining ChatGPT-4’s generative capabilities with Proof Analytics’ causal modeling, organizations can create robust counterfactual analyses to drive strategic decision-making.

Here’s a specific but simple counterfactual model following the outlined steps. There are no limits to how much more detailed a model could be made to be. In the end, the objective is to evaluate whether reallocation of resources across marketing, sales, product, and customer success metrics could help achieve an additional $2M ARR within 3 years from a current base of $650,000 ARR.

Phase 1: Define the Objective

1. Objective: Determine whether reallocation of budget and focus across key business functions (marketing, sales, product, and customer success) can drive $2M ARR growth within three years.
2. Scope: Baseline Revenue: $650,000 ARR. Target Revenue: $2,650,000 ARR. Hypothetical Change: Adjust marketing spend, sales team size, product pricing, and customer retention initiatives.

Phase 3: Create the Counterfactual Framework

1. Baseline Metrics:
2. Counterfactual Hypotheses:

Phase 4: Use ChatGPT-4 for Hypothesis Generation

Example Prompts:

• "What impact would a 20% increase in marketing spend have on ARR growth over three years?"
• "How much ARR growth is achievable if churn is reduced from 15% to 10% over three years?"

Hypotheses Generated:

• A 20% marketing budget increase with improved targeting yields 20 additional customers annually.
• Reducing churn by 5% improves LTV by $2,500/customer.
• Pricing optimization leads to a 5% revenue uplift per customer.

Phase 5: Build and Run the Model in Proof Analytics

1. Build Baseline Model:
2. Input Counterfactual Adjustments:

Phase 6: Interpret and Visualize Results

1. Key Insights:
2. Visualizations:

Phase 7: Communicate Findings

1. Comprehensive Report:
2. Presentation to Stakeholders:

Phase 8: Implement and Iterate

1. Implementation Plan:
2. Monitor and Refine:

By following this structured approach, the company can confidently model and implement strategies to achieve its $2M ARR growth goal within three years.