Most indie hackers I know treat financial modeling like a dentist appointment — uncomfortable, avoidable, and easy to postpone until something actually hurts. The problem: by the time it hurts, you've already burned three months building something that was never going to be profitable.
I've seen this pattern too many times. Someone ships an AI SaaS, gets 50 signups, and then realizes their OpenAI bill eats 40% of revenue at scale. Or they price at $9/month because "that feels right," not because the math works. This article gives you the framework to run the numbers before you write a single line of code.
Why AI SaaS Changes the Math Entirely
Traditional SaaS had a brutally simple unit economics model: infrastructure costs were nearly fixed, so margins improved as you scaled. AI SaaS breaks this assumption.
Every request to GPT-4o, Claude, or Gemini is a variable cost that scales linearly with usage. If your power users hammer your app, your gross margin doesn't improve — it can actually deteriorate.
Here's a concrete example. You're building a content generation tool:
- Subscription price: $29/month
- Average tokens per user/month: 500,000 input + 150,000 output
- Claude Sonnet cost: ~$3 input / $15 output per million tokens
- AI cost per user: (0.5 × $3) + (0.15 × $15) = $1.50 + $2.25 = $3.75/month
That's a 13% AI cost ratio — manageable. But if your power users consume 3× the average, you're looking at users who cost you $11/month against a $29 subscription. Now factor in Stripe fees (2.9% + $0.30), your SaaS tools, and maybe a small VPS, and your gross margin on those users is negative.
Model this before you build. Not after.
The Four Numbers That Actually Matter
Forget vanity metrics. Before launch, you need clarity on exactly four numbers:
1. Your target gross margin
For indie SaaS, aim for 65-75% gross margin minimum. Below 60% and you won't have enough left over to do marketing, support, or product iteration. AI costs are the biggest threat here — model them at P90 usage, not average.
2. LTV (Lifetime Value)
LTV = ARPU / Monthly Churn Rate
If you charge $49/month and lose 8% of customers monthly, your LTV is $49 / 0.08 = $612. That's the ceiling on what you can afford to spend acquiring a customer. Most indie devs never calculate this number and then wonder why growth feels impossible.
3. CAC (Customer Acquisition Cost)
If you're doing content marketing, your CAC is your time cost. If you're running ads, it's ad spend divided by conversions. The rule of thumb: LTV should be at least 3× CAC for a healthy business. If LTV is $612, you can spend up to ~$200 to acquire a customer profitably.
4. Payback period
How many months of subscription revenue does it take to recoup your acquisition cost? Payback = CAC / (ARPU × Gross Margin). Above 12 months and you have a cash flow problem even if the unit economics look good on paper.
Building the Spreadsheet (What Goes In It)
You don't need a 47-tab financial model. You need one sheet with these blocks:
Revenue model
- Monthly starting MRR (your month-1 target)
- Month-over-month growth rate (be conservative — 10-15% is realistic for early stage)
- Churn rate assumption (start with 7-10% monthly for B2C, 3-5% for B2B)
- ARPU by tier if you have multiple plans
Cost model
- AI API costs (model at average AND P90 usage per user)
- Infrastructure (VPS, database, CDN)
- SaaS tools (Stripe, email, error tracking, etc.)
- Your own time cost (yes, this matters — use an hourly rate)
Unit economics output
- Gross margin per user per month
- Break-even users (fixed costs ÷ contribution margin per user)
- LTV at current churn
- Implied CAC ceiling
Scenario planning
Run three scenarios: conservative (5% MoM growth, 10% churn), base (12% growth, 7% churn), optimistic (20% growth, 4% churn). Most indie devs only model the optimistic case. That's why they're surprised.
Pricing Math: Stop Guessing
The single most common pricing mistake is anchoring to competitors without checking whether their pricing works for your cost structure.
Here's a better approach — price from margin backward:
- Decide your target gross margin (say, 70%)
- Calculate your fully-loaded cost per user (AI + infra + tools)
- Work backward:
Price = Cost per user / (1 - target margin)
If your cost per user is $8/month, you need to charge at least $8 / 0.30 = $26.67. So $29/month is your floor, not a starting point for negotiation.
Also model your pricing tiers carefully. A freemium tier that converts at 3% sounds great until you realize you're paying AI costs for 97 users who never subscribe. Either rate-limit free aggressively or calculate the exact conversion rate needed to make free sustainable.
The 30-Day Action Plan Before You Code
Here's what I'd actually do, in order:
Week 1 — Build the model. Set up the spreadsheet framework above. Don't skip the AI cost modeling. Run your usage assumptions by a few potential users to sanity-check them.
Week 2 — Validate pricing tolerance. Talk to 10 potential customers. Don't ask "would you pay for this?" Ask "what are you currently spending to solve this problem?" That anchors real willingness to pay, not hypothetical.
Week 3 — Define your break-even milestone. Calculate exactly how many paying users you need to cover your monthly costs (including your time). Make this your first goal, not "$10K MRR." Small, achievable milestones compound.
Week 4 — Model your acquisition path. How do you get your first 50 customers? Content, communities, cold outreach, Product Hunt? Map the estimated CAC for each channel and check it against your LTV. If the math doesn't work, change the channel or change the price.
Ship the Model First, Then the Product
The discipline of building a financial model before writing code does something important: it forces you to make explicit all the assumptions you're holding implicitly. The market size assumption. The conversion rate assumption. The churn assumption. Every one of these is a hypothesis you'll need to validate anyway.
Running the numbers first doesn't constrain your creativity — it protects your time. You'll spend 3-6 months building this product. That's too long to wait to find out the economics don't work.
The indie devs who consistently build profitable products aren't smarter. They're just more honest about the math before they start.
I compiled everything into a practical guide: AI Build Financial Model: Stop Leaving Money Behind
Top comments (0)