I need one picture that shows where the money goes

In every organization, someone invariably asks about the allocation of expenses. Often, it’s the Chief Financial Officer (CFO), but sometimes it’s a Vice President of Engineering or an engineer tasked with handling cloud costs. After reviewing the bill, which is comprehensive yet overwhelming, the request typically emerges: “Stop.

I need one picture that shows where the money goes.” This is a reasonable request, but it’s surprisingly difficult to fulfill using existing tools. This article explains why, where the money often ends up, and what that one visual representation might look like. The bill answers “how much,” but the question is fundamentally “where does the money go?”

A cloud bill is essentially a long table listing every resource, every hour, and the associated cost. For example, an AWS Cost and Usage Report can contain millions of rows. While it’s easy to determine how much was spent on EC2 in July, identifying where each dollar went is a different challenge. Dollars enter a company via an invoice and then travel through several stages: the service provider, the account, the resource type, and finally, to a specific team.

Flat tables don’t depict these pathways. Native tools allow slicing data along a single dimension at a time. For instance, Cost Explorer can show spend by service, linked account, or tags. However, each view is limited, and answering the ultimate question—such as which team’s non-production environments in which account are driving compute growth—requires navigating multiple dimensions simultaneously.

This often results in six tabs and mental calculations, leading many teams to resort to spreadsheets. Someone eventually builds a pivot table, but it’s only useful for a short period before changes like reorganizations or new accounts render it inaccurate.

Where the money usually goes reveals recurring patterns in cloud expenses. Non-production environments that run continuously consume a significant portion of resources; a week has 168 hours, while a typical workweek has around 50. Development, staging, QA, and demo environments left running 24/7 can account for about 120 hours per week of unutilized resources.

Across multiple cloud bills, this non-prod spend is often the largest controllable expense. Unattributed spend is another common issue—resources without proper tagging, ownership, or cost center information, leading to untracked and often unchecked spending. This can range from 10–30% of the total bill. Similarly, “zombie” resources such as unattached volumes, idle load balancers, outdated snapshots, and NAT gateways continuously incur charges, even when their usage has ceased.

Instances running at low CPU utilization due to oversized initial setup or legacy configurations also contribute to waste. In essence, cloud waste can amount to roughly a third of total spend, a pattern that concentrates wherever there’s weak attribution.

Visualizing cloud spending effectively requires a different approach from traditional tables. The most effective tool for this purpose is a Sankey diagram—a flow chart where the width of the ribbons corresponds to the volume of flow. In the context of cloud spend, a useful Sankey diagram can include four columns: Provider → Account → Resource Type → Team.

This diagram effectively shows the entire lifecycle of every dollar spent in the cloud, making it easy to identify where the money goes and where there are inefficiencies. Thick ribbons in the diagram indicate areas of significant spending, such as unattributed costs, which are prominently displayed to draw attention. The key feature of this visual is that it requires no specialized training to interpret, allowing decision-makers to quickly understand the distribution of costs and focus on areas that need attention.

Building such a Sankey diagram requires exporting billing data, aggregating it based on specific dimensions, and then using visualization tools like D3.