I recently completed a take-home case for a GTM/strategic finance role at a B2B SaaS company. The task: build a pricing model comparing two options for a new AI add-on, then make a recommendation.
This is a walkthrough of how I approached it. The original deliverable included company-specific details I can’t share, but the thinking process is all here.
The Prompt
A B2B SaaS company was launching an AI Package as an add-on to their existing Business pricing tier. Two pricing models on the table:
- Credit model: $0.25 per AI action
- Flat-rate model: $500/month unlimited
Baseline assumptions provided:
- 1,000 existing Business-tier customers, adding 100 net new per month
- Survey data: ~2,000 AI actions per customer per month
- COGS: $0.16 per AI action
- 80% gross margin on core platform
Build a 2-year monthly model comparing both scenarios. Make a recommendation.
Step 1: Map the Business Scenario to Spreadsheet Logic
First step is always translating the verbal description into something I can actually model. What are the inputs, calculations, outputs?
The core logic:
- Customers: Start with 1,000, add 100/month
- Adoption: % of customers who buy the AI add-on
- Usage: Actions per adopter per month
- Revenue: Adoption × Usage × Price (credit) or Adoption × $500 (flat-rate)
- COGS: Usage × $0.16
- Gross Profit: Revenue − COGS
Once I had that skeleton, I could fill in assumptions.
Step 2: Identify What’s Missing
The prompt gives some numbers but not everything. If I were actually making this decision, what else would I need?
Adoption rates. Not provided, but probably the biggest driver of the model.
My thinking: flat-rate has lower friction — customers know exactly what they’re paying. Credit creates uncertainty (“what if my team uses more than expected?”), which suppresses adoption. I estimated 45% initial adoption for flat-rate vs. 35% for credit. Directionally reasonable, flagged as a key sensitivity.
Usage growth. The survey says 2,000 actions/month, but that’s a snapshot. Usage will likely grow as customers discover more use cases. I modeled 3% monthly growth, capped at 3,000 to prevent unrealistic Year 2 projections.
Adoption growth. Early adopters first, then gradual increases from word-of-mouth and product maturity. Modeled 2% monthly growth, capped at 60–70%.
The goal isn’t to nail the exact numbers — it’s to make reasonable estimates, document them clearly, and structure the model so they’re easy to adjust.
Step 3: Build and Stress-Test
Built a 24-month projection for each scenario. Here’s the output:
Credit wins on both revenue and profit despite lower adoption. The margin math: flat-rate collects $500 whether a customer uses 500 actions or 3,000. As usage scales, margin compresses. Credit maintains 36% regardless of volume.
But which assumptions are actually driving this? I ran sensitivity analysis — adjusted each variable by ±25%:
Most sensitive to adoption rates — also the assumptions with the least real-world validation.
Step 4: Frame for Decision-Making
A spreadsheet isn’t a deliverable. A recommendation is.
I structured the write-up around what would actually help make the decision:
- Recommendation: Credit model
- Why: Better unit economics; advantage widens as usage scales
- Key bet: Adoption and usage assumptions aren’t validated yet
- Risks: If credit adoption falls well below 35%, or usage anxiety causes early churn, the math changes
- De-risk: 60-day pilot with ~50 customers. If adoption drops below 28% or usage below 1,600 actions by week 6, revisit
The model says “credit is better under these conditions.” The recommendation includes how to test those conditions and when to change course.
Reflection
Looking back, a few things stood out:
The assumptions I had to invent (adoption rates, usage growth) ended up mattering more than the numbers I was given. That’s usually how it goes — the ambiguous inputs drive the model more than the concrete ones.
I also spent more time on the sensitivity analysis than I initially planned. But that’s where the real insight came from. The recommendation isn’t “credit wins.” It’s “credit wins if adoption holds, and here’s how much it matters if it doesn’t.”