When Your Best Customers Are Also Your Most Expensive Ones
In traditional SaaS, your power users were your best customers. They logged in daily, used every feature, and became your biggest advocates. The more they used your product, the stickier they became—and it barely cost you anything extra to serve them.
AI flipped that equation. Now your most engaged customers might be the ones destroying your margins.
The Power User Paradox
Consider two customers on identical $300/month plans. Same acquisition cost. Same contract terms. On the surface, they look the same.
Customer A is a power user—the type every product team celebrates. They log in daily, rely heavily on AI features, and generate about 150 AI requests per day. That's roughly 4,500 requests per month. At $0.04 per request in API and infrastructure costs, that adds up to $180 in variable expenses. Factor in another $40 in support and platform costs, and you're spending $220/month to serve this customer.
On a $300/month plan, that leaves just $80 in contribution margin.
Customer B is lightly engaged. They use the product occasionally—maybe 10 AI requests per day, around 300 requests per month. Their variable costs? About $12. Their contribution margin? $248.
Same plan. Same revenue. Wildly different profitability.
The Math Gets Worse
Here's where it really hurts. If your customer acquisition cost is $3,600, Customer A's payback period stretches to 45 months. Despite being highly engaged—despite being exactly the kind of user your product team optimizes for—they're barely covering their acquisition cost and eroding your margins every month they stay on the platform.
Customer B? With their $248 contribution margin, they pay back in under 15 months.
The customer you're worried about churning is actually three times more profitable than the one you're celebrating in your weekly metrics meeting.
Why Traditional Metrics Lie
This is the trap. If you're only measuring revenue or engagement, you might mistake your most expensive customers for your best ones.
In AI-powered SaaS, the metrics that worked for traditional software actively mislead you:
Daily Active Users? Your highest DAU customers might be your biggest margin destroyers.
Feature adoption? The customers using your AI features most heavily could be costing you money on every interaction.
NPS scores? Power users love your product because they're getting enormous value—value that exceeds what they're paying.
Without understanding contribution margin at the customer level, you're likely scaling a product that loses more with every "successful" user.
The Flat-Rate Time Bomb
Flat-fee pricing with unlimited AI usage may feel user-friendly early on, but it's a time bomb for unit economics. As usage scales, your margins collapse.
The problem compounds with growth. When you acquire more customers, you're not just adding revenue—you're adding usage variance. A handful of power users can burn through compute, spike infrastructure, and eat into your margins quickly. You may still show revenue growth while your unit economics deteriorate beneath the surface.
One user on an AI-powered platform might consume 10x more resources than another—and generate 10x more value. Under flat-rate pricing, they both pay the same.
The Visibility Problem
Most companies don't even know they have this problem.
Finance sees a total AI spend number in their monthly reports. Engineering knows the aggregate API costs. But nobody can answer the fundamental question: which customers are profitable and which ones aren't?
84% of companies report margin erosion from AI costs, but without customer-level visibility, they can't identify the source. They optimize in the dark—cutting features, raising prices across the board, or hoping efficiency improvements will save them.
Meanwhile, the variance between their most and least expensive customers continues to grow.
What Changes Everything
The solution isn't to punish your power users or discourage engagement. It's to understand your cost structure well enough to price appropriately and make informed decisions.
This requires a fundamental shift in how you track costs:
1. Attribute costs to customers, not just totals. Aggregate AI spending tells you nothing about where you're making money. You need to know what each customer costs to serve.
2. Understand the distribution. What's your median cost to serve? What does the 90th percentile look like? How much variance exists between customer segments?
3. Connect usage to revenue. Which pricing tiers are profitable? Which customer segments are underwater? Where do you have headroom, and where are you losing money?
4. Model before you scale. Before launching that unlimited AI feature, map out your variable costs per account and model three scenarios: light, average, and heavy usage. If the high-usage scenario drops your gross margin below 50-60%, you likely can't afford a true flat fee.
The Path Forward
The companies surviving the AI margin squeeze are the ones building enough product depth that they're not just marking up API calls. They understand their unit economics at the customer level. They can identify which accounts need attention and which pricing models actually work.
Clear visibility into your compute, storage, and API costs is the only thing that will prevent you from underpricing your AI product and losing money, overpricing it and driving customers away, or failing to scale profitably.
Your best customers aren't necessarily the ones who use your product most. They're the ones who generate value that exceeds what it costs to serve them. In AI-powered software, you can't tell the difference without measuring both sides of that equation.
The $20 customer costing you $40 isn't a hypothetical. It's probably in your customer base right now. The only question is whether you know which one it is.
Stop guessing which customers are profitable. tknOps gives you customer-level cost attribution for your AI spend—so you can see exactly who's making you money and who's costing you. Because in AI, what you can't measure will cost you.