The Bill That Grows Faster Than the Business: Auditing Egress and Per-Seat Costs Before Renewal Season
Photo: Joe Haupt from USA, CC BY-SA 2.0, via Wikimedia Commons
Every US founder can recite their cloud bill from last month. Far fewer can tell you what a single customer actually costs them — and the difference between those two numbers is where margin goes to die.
Where the Bill Stops Tracking the Business
The core compute and storage line items scale with usage, which at least correlates with revenue. The trouble lives in the line items that don't behave that way:
Egress bandwidth. Major providers still charge meaningful rates per gigabyte of data leaving their network. A media-heavy SaaS, a data analytics platform, or anything streaming files to end users can watch egress grow super-linearly as adoption deepens — customers who upload more, share more, and integrate more are your best customers, and they cost the most to serve.
API call tiers. Usage-based API pricing often looks cheap at the demo stage and punitive at scale. Crossing a tier boundary can reprice your entire volume, not just the marginal call. Teams discover this in the invoice, not the pricing page.
Per-seat creep. Collaboration and observability vendors price per user, and your headcount grows with success. Worse, licenses rarely get reclaimed when people leave or change teams — the average US company pays for a meaningful share of seats nobody uses.
Cross-zone and cross-region transfer. The micro-fees inside your architecture — traffic between availability zones, replicated storage, inter-service chatter in a microservices setup — are invisible in any single dashboard but material in the aggregate.
Modeling Cost Per Customer, Honestly
A workable model needs only a spreadsheet and one month of detailed billing data:
- Pull line-item billing at the finest granularity your provider offers, tagged by service and, where possible, by customer-facing workload.
- Separate fixed from variable. Baseline environments, security tooling, and reserved commitments don't scale with customers. Everything else gets attributed to usage drivers: gigabytes out, API calls, active seats.
- Compute marginal cost per customer for your last cohort of signups. If your newest customers cost more to serve than your earliest ones, your architecture has a scaling inversion — and no pricing tweak will fix it.
- Stress-test against growth. Model the bill at 2x and 5x revenue. If gross margin compresses, you've found the exact line item to attack: egress-heavy features to cache or CDN-front, tier boundaries to engineer around, seats to reconcile.
Teams that run this exercise are routinely surprised by one finding: the cheapest architectural change often beats the best negotiation. Moving a data-heavy workflow behind a CDN, or batching sync jobs into off-peak windows, can save more than a discount call ever will.
Walking Into Renewal Season
Renewal negotiations reward preparation. A few tactics that work in the US market:
- Bring your own data. Vendors discount what they fear losing. A documented, per-customer cost model shows exactly where their pricing pinches you — and makes churn credible.
- Negotiate the rate card, not the invoice. Committed-use discounts and custom tier boundaries beat one-time credits, which merely defer the problem to next year.
- Time it to your leverage. Approaching a tier threshold, a funding round, or a competitor's evaluation cycle all shift the conversation.
- Benchmark the per-seat vendors against each other. Seat-based SaaS pricing has real elasticity; procurement teams that request competing quotes typically see meaningful movement.
Unit economics aren't an accounting abstraction — they're the difference between a startup that compounds and one that treads water while revenue climbs. Read the fine print before the fine print reads you.
