Thank you to our affiliates who help support this newsletter:
Thank you to our sponsors who help support this newsletter:
3,000+ AI tools. 1000+ prompts. All from one newsletter.
Most AI resources are scattered, untested, or buried behind a paywall.
The Shift puts it all in one free newsletter, covering what’s new, and what practically works under 5 minutes a day.
A vault of 3,000+ vetted tools so you’re never guessing. A 1000+ prompt library for every use case you’ll actually hit and free AI courses that make sure nothing stays theoretical.
Plus, right now, 3 subscribers win a free 1-year Claude Pro subscription just for signing up.
AI is completely flipping the unit economics of SaaS. And if you aren’t paying attention, your “growth” is going to kill your company.
I recently reviewed a strategy from a veteran Product Leader. On the surface, they were crushing it. They told me:
“Our usage and engagement metrics are through the roof, yet we're constantly hitting 'resource constraints' and can’t get budget approval for the roadmap items we need to actually scale. It feels like we’re winning on growth, but the business is tightening the screws on our development capacity for no clear reason.”
He bragged about a 5:1 LTV:CAC ratio. The kind that makes boards and investors salivate.
But when I asked, “What percentage of your revenue goes to API and compute costs?” the answer was 52%.
At a 48% gross margin, their true CAC:LTV was 5 × 0.48 = 2.4:1. Below the minimum threshold for sustainable growth. Serving each additional user was actually costing them money. His company wasn’t a high-flying SaaS business. They were a low-margin services business disguised as an app.
AI companies operate with a fundamentally different cost structure, and applying standard LTV-CAC frameworks without adjusting for it produces dangerously misleading numbers.
This has massive consequences on your product strategy, from pricing to product excellence to how you look at your user personas.
Here’s how to fix your product strategy before your compute bill eats your company.
What You’ll Learn Today
Why you need to stop training others to ask “When will this ship?”
How timeline roadmaps create false commitments.
Where real business commitments should live. (Hint: not on your roadmap.)
The "Two-Way" system for connecting product work to business outcomes.
How to build executive trust without committing dates.
How to build a system that proves ROI, not just activity.
The Cost Structure Problem
Standard SaaS benchmarks assume 70-80% gross margins. This is because traditional SaaS delivers the same software to each customer. The marginal cost of serving an additional user is near zero.
With an AI product, every customer interaction costs money. An LLM-powered feature runs inference on every request. A computer vision model processes every image. The more a customer uses your product, the more it costs to serve them.
This changes three things simultaneously:
Variable Costs Are Higher. Traditional SaaS variable costs run 10–20% of revenue. AI products run 30–60%, depending on model size, inference volume, and whether you use third-party APIs or self-hosted infrastructure.
Variable Costs Scale With Usage. A customer sending 10,000 API calls per month costs 10x more to serve than one sending 1,000. They may not pay 10x more, especially on flat-rate plans.
Variable Costs Are Unpredictable. Token costs, GPU pricing, and API rates change. OpenAI has cut pricing 10x in two years. Your COGS line item can shift 30% quarter over quarter.
In AI, every time a user clicks "Generate," you get a bill from OpenAI, Anthropic, or AWS.
If you use the standard SaaS formula:
LTV = (ARPU × Retention) / Churn
You’re lying to yourself. You’re counting revenue that’s already promised to your chip or model provider. If your Revenue is $100 but your API costs are $52, your “real” LTV is cut in half before you even start.
Instead, you need to use Contribution Margin that accounts for variable compute:
Contribution Margin LTV = (ARPU - Variable Costs per Customer) * Average Customer Lifetime
Variable costs per customer include:
API costs (OpenAI, Anthropic, Google, or self-hosted inference)
Cloud compute (GPU instances for self-hosted models)
Per-request infrastructure (load balancers, queues, logging)
Data processing and storage tied to usage
Third-party per-transaction fees
An Example
A document analysis AI charges $200/month. Variable costs per customer:
LLM API costs: $60/month
Cloud processing: $15/month
Storage and bandwidth: $10/month
Total variable: $85/month
Contribution margin per customer: $200 - $85 = $115/month
Average customer lifetime: 18 months
Contribution Margin LTV: $115 × 18 = $2,070
Standard LTV (using 75% gross margin): $200 × 0.75 / (1/18) = $2,700. Overstates by 28%.
At $800 CAC, the standard formula shows 3.4:1 (comfortable).
The Contribution Margin formula shows 2.6:1 (below threshold).
Different number, different strategic decision.
CAC Differences For An AI Product
Demo and POC costs are higher
AI demos often need custom data ingestion or model tuning before prospects can evaluate. This can add $2K - $10K+ per enterprise prospect per demoed workflow to CAC.
Free tier costs are real
A SaaS free tier costs near-zero per user. An AI free tier with LLM access costs $0.50–$5.00 per monthly active user.
At 10,000 free users, that’s $5K - $50K per month in additional compute with no revenue.
If 5% convert (500 paid), that’s $100 per customer in free-tier CAC before any marketing spend.
What Boards & Investors Are Looking For In AI Unit Economics
Based on conversations with senior executives and investors and my own experience working at an AI startup:
Gross margin trajectory
A 40% margin trending toward 60% as you optimize inference is acceptable. A flat 40% for 12 months is not.
Model cost reduction plans
Quantization, smaller models for simpler tasks, caching, provider negotiation. “We’ll switch to open-source” is not a plan.
Usage distribution data
What % of customers hit each usage tier, and the contribution margin at each. A Pareto distribution where 5% of users cause 50% of compute costs is manageable.
Payback under 18 months on contribution margin basis
Tighter than standard SaaS because AI costs are less predictable.
Strategies for Product Leaders
Segment by “Usage Personas”
Instead of looking at users just by their job titles or roles, you need to segment them by how much they “eat.”
The Casuals: Low usage, high margin. They pay the bills.
The Power Users: High usage, potentially low (or negative) margin. They are your biggest risk.
If your roadmap focuses only on making the Power Users happy, you are effectively optimizing for bankruptcy. You need to build features that provide high value with low inference density.
Efficiency as a Product Feature
In AI, product excellence isn’t just about UI/UX or speed. It’s about Inference Efficiency.
Every prompt you write and every model you choose is a financial decision. So, switching from a massive model (like GPT-4) to a smaller, faster one (like GPT-4o-mini or a fine-tuned Llama) is a product win. It directly increases your LTV.
Be careful, though! In high-stakes fields like healthcare (e.g., medical coding), you can’t just swap to a cheap model. If accuracy drops, human review time goes up, and the customer loses value. If you save $1 on API costs but cost the user $10 in manual fixes, you’ll see that reflected in your churn rate.
Your job as a product leader is to find the “Goldilocks” model that balances Cost vs. Accuracy: the cheapest one that doesn’t break the user’s trust.
Choose Your Pricing Wisely
You can’t stick to “Unlimited” plans anymore. You have to pick a strategy that protects your margin.
You have three main choices. None of them are perfect.
Model | Pros | Cons |
|---|---|---|
Flat-Rate Subscription | Simple and familiar for users. Predictable revenue. | You take 100% of the usage risk. |
Usage-Based | Solves the usage problem directly. Charge per call, per document, per minute, per compute. Protects your margins. | Customers dislike unpredictable bills. High “mental tax”; users afraid to explore. |
Tiered Pricing with Usage Caps | “Pro plan: 10,000 analyses per month for $299.” Protects margins while capping exposure. | Power users churn at the cap. |
Hybrid Models | Combine a base fee with usage charges above a threshold. Predictability of revenue + margin protection against usage costs. | Complex for Sales to explain, customers to track, and Finance to bill. Some users may still churn at the cap. |
Have Hard Conversations
You can’t do this alone. As a Product Leader, you need to lead specific “margin-first” conversations. Get in the room with your leadership peers and get real.
With the CEO — the Growth-Margin Paradox:
Ask: “Are we scaling a ‘Star’ product or a ‘Trap’ product?”
A “Star” follows the classic software law of marginal costs trending toward zero. So, as you grow, your efficiency improves. Your gross margins expand over time. You have more cash to reinvest in R&D, making the product even better.
A “Trap” looks like a winner on a bar chart but functions like a weight around your neck. The product has high inference density. Every user action requires a massive, expensive call to a frontier model (like GPT-4o or Claude 3.5 Sonnet). The result is your costs are linearly coupled with your revenue. You grow 100%, your compute bill grows 100% (or more). Because you never get economies of scale, you are perpetually stuck with low margins, leaving no money to reinvest in R&D or innovate.
The Goal: Ensure the CEO understands that 100% YoY growth is dangerous if the Contribution Margin is only 20%. You need to align on whether the current goal is a market land-grab (at any cost) or sustainable unit economics.
With the CTO — the Model vs. Margin Tradeoff:
Ask: “Do we actually need the most expensive model for this specific feature?”
The Goal: Push for a “multi-model strategy.” Use the expensive “frontier” model for complex reasoning and a cheaper, smaller model (or a fine-tuned open-source model) for 80% of routine tasks.
Infrastructure: Discuss caching common queries to avoid redundant inference costs.
With Finance — Real-Time COGS Monitoring:
Ask: “Can we attribute our API/compute bill to specific customer segments or features?”
The Goal: Finance usually sees a lump sum bill from OpenAI or AWS. You need to build “Unit Economic Visibility” into the product so Finance can see exactly which features or customers are eroding the margin.
With Sales & Marketing — Refining the ICP:
Ask: “Are we attracting Efficiency Seekers (high usage, low margin) or Value Seekers (high value, controlled usage)?”
The Goal: Adjust your Ideal Customer Profile (ICP). You want to acquire customers where the AI provides massive value (high willingness to pay) but requires relatively low inference density.
Final Takeaways
LTV is for vanity, Margin LTV is for sanity. Use the right formula. If you don’t subtract your API/compute costs first, your math is wrong.
Re-think and manage your Personas. Identify users who cost more than they pay. Monitor heavy users like a hawk.
Model selection is a product decision. Don’t leave it all to the engineers. Choosing a smaller model or caching responses isn’t a technical problem. It’s a margin play. “Cheap” models can lead to expensive churn. Performance vs. Accuracy vs. Cost is a strategic trade-off.
Stop selling “Unlimited.” It doesn’t exist in AI. Someone is paying for those GPUs. Make sure it isn’t just you.
Price for the upside. If your AI saves a customer $10,000, don’t just charge them $20 a month. Match your price to the value, but floor it at your cost.
That’s all for this week.
Have a joyful week, and, if you can, make it joyful for someone else too.
cheers,
shardul
Here are 4 ways I can help you today:
Executives: Eliminate Decision Drag and Drive Commercial Impact. I help organizations build the product strategy and discipline need to turn technology into a high-margin business. Let’s discuss your next phase of growth. Let’s discuss your next phase of growth.
Product Leaders: Invest in Your Product Operating Model. Stop the “delivery drone” cycle and unlock your team’s true potential as a strategic business function. Schedule a Strategy Call Today.
Product Managers: Get 1:1 Street Smart Career Guidance. From 1:1 coaching to a resume review to a mock interview, get real-world strategic feedback from an executive who has hired, mentored, and promoted at every level, whether you’re breaking into PM or are rising to the leadership ranks. Book a Coaching Session Today.
Aspiring and New PMs: Learn the Unvarnished Truth on What the Job Really Is. It’s one of the most misunderstood roles in tech. It can be a meaningful role for the right people. But only when entered with realistic expectations, self-awareness, and intent. Get the unvarnished truth about the role before you commit your time, money, and entire career. Get Early Access Here Today.
Shardul Mehta
I ❤️ product managers.