AI Cost & Margin Analytics

See which customers and AI features are eating your margin

Track AI cost per customer, feature, workflow, model, and plan — without storing prompts or changing your AI stack.

AI cost is not just infrastructure spend. In B2B SaaS, it belongs in the customer account view. SaaS Tracker helps you understand which accounts, features, and workflows create AI cost, how that compares to revenue, and where pricing or packaging may need to change — included in every plan from €79/month.

AI margin in the account view

Provider dashboards show total spend. SaaS Tracker shows which customer accounts, features, and workflows caused it — next to product usage, health, and revenue.

AI cost by customer account

Total spend per organization — with MRR, plan tier, and billing status beside the numbers.

AI cost by feature and workflow

See which product features drive LLM spend — support replies, summarization, agents, and more.

AI cost by model and provider

Break down spend by OpenAI, Anthropic, Azure, or other providers and model versions.

AI cost vs plan / MRR / customer tier

Spot margin-risk accounts where AI cost is high relative to revenue — not just high in absolute terms.

Metadata-only tracking

Tokens, models, features, and cost — without storing prompts, responses, or routing through a gateway.

Built for B2B SaaS AI features

Connect AI economics to the same customer_org_id as product usage, health, and revenue.

When AI ships, finance and product need the same picture

  • “We don’t know what AI costs per customer or per feature.”
  • Trial or small accounts consume a disproportionate share of tokens.
  • Pricing and packaging don’t reflect AI margin.
  • Teams want AI cost + MRR in one place — not another spreadsheet.

Your app stays in control — we only see events

Integration is event-based on the same Ingest API as the rest of SaaS Tracker. Your code calls OpenAI, Anthropic, Azure, or others as today, then sends metadata after each call.

  • No AI gateway or proxy in the path
  • No prompt or response storage
  • Dashboards in Analytics → AI Cost

Data flow

  1. 1 Your app calls the provider API
  2. 2 Provider returns tokens + metadata
  3. 3 Send event data to SaaS Tracker Include your cost in the event, or model + tokens for us to price
  4. 4 View AI Cost dashboards per customer, feature, and model

Two ways to record cost

Use whichever fits your stack — you can rely on one approach or combine them as your integration matures.

Send cost in the event

Include the cost for each AI call in your event data — for example what your provider charged or what you already calculated in application code. SaaS Tracker aggregates spend by customer organization, feature, and model.

Best when you already have authoritative per-call cost from the provider or your own billing logic.

Send model + tokens — we calculate price

Send provider, model, and token counts only. SaaS Tracker applies EUR pricing at ingest from the global model catalog and any negotiated rates you configure in Settings — we track pricing so you do not have to maintain rate tables in code.

Best when you want consistent, catalog-based costing without duplicating provider price lists in your app.

What you see in the product

Analytics → AI Cost — overview, customers, features, value signals, and recent calls.

Overview

AI cost over time, KPIs (cost, calls, avg per org), and top features and models.

Customers

Spend per customer organization, MRR context, billing status, and margin-risk signals.

Features

Cost by feature name — e.g. support_reply_generator — so product teams see what drives spend.

Value signals

Growth plan and above: accepted, rejected, and edited outputs; cost per accepted result.

Recent calls

Call explorer with filters and metadata. No message content — privacy by design.

Screenshot coming soon: sign in at app.saastracker.eu → Analytics → AI Cost → Overview.

Built for B2B SaaS economics

Privacy by design

Metadata and tokens only — never prompts or model responses.

No infrastructure change

Call providers directly, then send an event. No gateway, no proxy.

Your cost or ours

Send cost in the event and we roll it up — or send provider, model, and tokens and we calculate EUR cost from tracked pricing.

Same platform as org health

AI spend links to customer_org_id and imported MRR on one account-level timeline.

Integration in one pipeline

Send events through the Ingest API after each AI interaction. On each completed call, either pass through the cost you already know or pass provider, model, and token counts and let SaaS Tracker derive EUR cost from tracked pricing.

Always include customer_org_id, properties.feature, and ai_call_id where relevant for output events. See the docs for exact property names per mode.

  • ai_call_completed
  • ai_call_failed
  • ai_output_accepted
  • ai_output_rejected
  • ai_output_edited

Details: AI Cost Analytics docs · Event schema

Guides on AI cost per customer

Practical reads on attribution, margin, and metadata-only tracking — without storing prompts.

What Is AI Cost Analytics for SaaS?

AI cost analytics helps SaaS teams track AI feature costs by customer, feature, model, and plan — and connect usage to value, margin, and retention.

Read article →

How to Track AI Costs per Customer in B2B SaaS

Learn how B2B SaaS teams can track AI costs per customer, connect LLM spend to MRR, and see which accounts, features, and models affect margin.

Read article →

AI Cost Analytics Without Storing Prompts

Learn how SaaS teams can track AI usage, tokens, models, features, and customer-level cost using metadata only — without storing prompts or responses.

Read article →

Why Your OpenAI Bill Does Not Tell You Which Customers Are Profitable

Provider invoices show total model cost, but not feature value or revenue per account. Learn how to connect OpenAI spend to SaaS AI unit economics at the organization level.

Read article →

AI Cost & Margin included in every plan

Founder includes basic AI Cost Analytics and margin view. Growth adds full AI Cost & Margin Analytics — workflow-level cost, value signals, and margin risk. See pricing for plan details.

Common questions

Can I send cost myself, or do you calculate it?
Both. Send cost in the event and we aggregate it, or send provider, model, and tokens and we calculate EUR cost at ingest from our pricing catalog and your Settings overrides.
Do you store prompts or chat content?
No — metadata and token counts only.
Is there an AI gateway in the product?
No. Your application keeps calling providers; you send events afterward.
Are AI cost alerts included?
Not yet — alerting is on the roadmap, not part of the current release.
What happens when I change model pricing?
New events use updated catalog or negotiated rates. Historical rows are not automatically recalculated when aliases or prices change.

Ready to see AI margin per customer?

14-day Founder Trial · no credit card · AI cost tracking included in every plan from €79/month. The trial starts when your first data arrives — not when you create your account.