The Seven Places Your SaaS Product Spends Money Every Time a User Clicks a Button

Most SaaS founders can tell you their monthly cloud bill. Very few can tell you what happens in the thirty milliseconds between a user clicking a button and seeing a result on their screen.

Your costs are being spent in those milliseconds, about the time it takes you to do a third of an eye blink.

Understanding the journey a single user request takes through your infrastructure is not just an engineering curiosity. It is the foundation of knowing what your product costs to deliver, which customers are driving that cost, and whether your business model holds up as you scale. Every
dollar on your cloud bill traces back to one of seven stops along this journey.

Stop 1: The User’s Device and Internet Connection

Before your infrastructure spends a cent, the user has already spent something. Their computer, their power bill, their internet service provider - these are the user’s costs, not yours. A click generates a request that travels from their device across their ISP network toward your servers.

This stop matters to you for one reason: latency. A slow connection on their end affects their experience, and experience affects retention. But the cost here lands on the user, not your cloud bill.

The journey becomes your financial responsibility as soon as the request reaches your infrastructure.

Stop 2: The CDN Edge Node

The first place your money is spent is the Content Delivery Network, or CDN. Companies like Cloudflare, Fastly, and AWS CloudFront operate servers positioned geographically close to users around the world. When a user in Dallas clicks a button in your SaaS product, their request hits a CDN node in Dallas rather than traveling all the way to your primary data center.

CDNs serve two purposes. They make your product faster by reducing the distance data travel. And they absorb traffic, protecting your core infrastructure from being overwhelmed.

You pay for CDN usage based on the volume of traffic routed through it and the amount of data delivered. For a SaaS product serving static content like images, stylesheets, and cached pages, CDN costs are modest and predictable. For a video platform or a product that delivers large files, the CDN becomes a significant line item.

Stop 3: The Load Balancer and API Gateway

After the CDN, the request reaches your load balancer. Think of this as the traffic cop at the entrance to your infrastructure. It distributes incoming requests across servers so no single server gets overwhelmed.

Alongside the load balancer sits the API gateway, which does a different but equally important job. It checks whether the user is who they say they are, enforces rate limits, routes the request to the right part of your application, and logs the transaction. Every request your users make passes through this layer.

You pay for load balancers and API gateways based on the number of requests processed and the amount of data transferred. The cost per request is small but the volume can be enormous. A SaaS product with a hundred thousand active users might process tens of millions of API requests per day.

This is also the first place where per-customer cost variation begins to appear. A user who clicks frequently generates more API gateway transactions than one who clicks rarely, and those transactions add up differently across your customer base.

Stop 4: The Application Server

This is where your product’s brain lives. The application server receives the request, figures out what the user is asking for, applies your business logic, and determines what needs to happen next.

If a user clicks “run report,” the application server interprets that instruction, decides what data is needed, how it should be filtered, what calculations are required, and how the results should be formatted. It then interacts with other systems needed to fulfill the request.

You pay for application servers based on compute resources: CPU time, memory consumption, and the duration each request takes to process. A simple request that resolves in ten milliseconds costs a fraction of a cent. A complex report that runs for thirty seconds costs meaningfully more.

This is the stop where the nature of what different customers do begins to diverge significantly. A customer running complex analytics all day drives far more cost than one performing routine data entry. Same product, same subscription price, very different compute consumption.

Stop 5: The Supporting Services Layer

This is the most variable stop on the journey, and the one that has changed most dramatically in the AI era.

After the application server decides what needs to happen, it often needs to call on external or supporting services to do the work. Depending on your product, this might include any combination of the following:

The cache. Before hitting the database, the application server checks whether the answer already exists in memory. A cache like Redis or Memcached stores frequently requested data in fast-access memory so it does not need to be recalculated every time. A cache hit is fast and cheap. A cache miss means the request continues to the database, which costs more time and money.

The database. This is where your data lives. Reading a record, writing a new entry, running a query across millions of rows - each of these operations consumes resources. You pay for database compute, storage, and the number of read and write operations your product performs. For most traditional SaaS products, the database is the single largest cost driver after application compute.

External services. Your product probably connects to services you do not own. Stripe for payments. Auth0 for authentication. SendGrid for email. Twilio for SMS. Each of these external connections cost money, either per transaction or as a monthly subscription, and that cost flows into your infrastructure bill.

AI APIs. This is the stop that has changed everything in the last two years. If your product has any AI feature - a chatbot, a summarizer, a code assistant, a recommendation engine - then somewhere in this layer your application server is sending a request to an AI model provider like Anthropic, OpenAI, or Google. That provider charges you by the token, for both what goes in and what comes back.

This is where the cost variability becomes most acute. A user who asks a simple question might trigger an AI call that costs a fraction of a cent. A user who pastes a hundred-page document and asks for a detailed analysis might trigger an AI call that costs several dollars. Both users clicked the same button.

Stop 6: Object Storage

Not every request ends when the application server finishes its work. Many require storing or retrieving files. A user uploads a document, exports a report, generates a certificate, or requests a video - all of these involve object storage.

Services like Amazon S3, Google Cloud Storage, and Azure Blob Storage hold the files your product needs to keep. You pay for how much you store and how often you access it. Storing data is relatively cheap. Retrieving it adds up.

For many SaaS products, object storage is a modest cost. For products built around file management, video, or large data exports, it becomes a primary cost driver. The key variable, as with most stops on this journey, is what your customers do. A customer who generates fifty reports a day creates far more storage activity than one who generates two a month.

Stop 7: Data Egress

This is the stop that surprises most people who have not looked carefully at their cloud bill.

Cloud providers charge you not just to store and process data, but to send it back out across the internet. Every time your product delivers a response to a user - a page load, a file download, an API response, a video stream - data travels from your infrastructure to the user’s device. That outbound data transfer is called egress, and it is billed by the gigabyte.

For products that serve small payloads - a few kilobytes of JSON per API response - egress costs are negligible. For products that stream video, serve large files, or return complex data- heavy responses, egress becomes a meaningful part of the bill.

Consider two customers on the same platform. A small estate planning law firm pulls up client files, reads them on screen, and closes them. The documents stay in the cloud. Minimal egress. A litigation firm downloads enormous discovery packages - perhaps thousands of scanned pages - to share with co-counsel, opposing parties, and the court every week. Same platform, same subscription tier, but the litigation firm is generating egress costs that could be ten times higher.

The same pattern appears across industries. Two regional hospital systems look identical on paper - same bed count, same patient volume. But one is a general medical facility while the other has a major cancer center where every patient generates gigabytes of imaging data that gets downloaded and shared across multiple providers. Two banks with identical assets, but one does complex commercial lending with massive document packages and the other does routine consumer mortgages. The salesperson who priced them identically made a costly assumption.

Without per-customer cost attribution, the SaaS company sees its egress line climbing every quarter and has no idea which customers are responsible.

The Full Picture

Seven stops. Seven places your money goes before a user sees their result.

What makes this important for SaaS economics is that each stop scales differently and is driven by different aspects of user behavior. CDN costs scale with traffic volume. Database costs scale with data complexity. AI API costs scale with what users ask the AI to do. Egress costs scale with the size of what gets delivered.

Of the non-AI stops, egress tends to be the most variable in absolute magnitude - the difference between a text record and a medical imaging file can be 50,000 times in data size. Database compute is the most consistently underestimated activity across almost all SaaS products. And AI is the most unpredictable, because a single user can spike costs dramatically on any given day depending on what they decide to ask.

A user who clicks once and gets a simple answer barely registers across any of these seven stops. A user who runs complex AI-assisted workflows all day touches every stop repeatedly and at significant cost.

Both pay the same subscription fee.

Until you can see how much each customer costs to serve across all seven stops, you are managing your cloud bill blind. You know the total. You do not know which customers are driving it, which stops are are out of proportion, or whether your pricing reflects the actual cost of your product.

That visibility is not a nice-to-have. As AI makes Stop 5 more variable and less predictable than ever before, it is the foundation of a defensible business model.

Beakpoint Insights connects infrastructure cost to business outcomes, giving SaaS companies cost-per-customer and cost-per-feature visibility across all seven stops, built natively on OpenTelemetry. If you are trying to understand what your product actually costs to deliver, we would like to talk.