Showback for platform teams.

Here is how a platform team builds showback that holds up.

Showback or chargeback: which should a platform team start with?

Chargeback moves real budget: each team is billed for its usage and the cost leaves the platform team’s books. It drives the strongest accountability but needs allocation everyone trusts and finance machinery to support it, so it is a destination, not a starting point. Showback reports the same allocation without the transfer. Start with showback to build trust in the numbers and shift behaviour, then graduate to chargeback once the allocation is accepted and the data is reconciled. Launching chargeback on numbers teams dispute turns every cost conversation into an argument about the method.

How do you allocate the cost a single team does not own?

Direct cost is easy: a namespace’s pods consume measurable CPU, memory, and storage, and you cost them by their resource requests, the same reserved capacity that drives the node bill. The difficulty is shared cost. A cluster runs a control plane, system daemonsets such as logging and networking agents on every node, and idle headroom held for burst and resilience. None of it belongs to one tenant, but all of it is real. The fair method is to allocate shared and idle cost back to tenants in proportion to their direct usage, so a team that consumes 20 percent of the workload capacity also carries 20 percent of the overhead. Show that overhead as a separate line so teams see both their direct footprint and their share of the commons.

What allocation key should you use?

Allocate by resource requests, not raw usage, because requests are what reserve node capacity and therefore what cost real money, and because a key teams can predict is a key they will trust. A namespace label per team is the minimum metadata; richer labels for product and environment let you slice the same data without re tagging. Hold the key stable across reporting periods so a team’s number changes only when its consumption changes, which is the whole point of showback. There are namespace level cost allocation tools that automate this against cluster telemetry; whichever you use, the allocation logic above is what determines whether the report is fair.

How do you use showback to actually change behaviour?

A report nobody reads changes nothing. Send each team its own number on a monthly cadence, show the trend not just the level, and pair the largest direct lines with the rightsizing action that would move them, so the report is also a to do list. Surface the request to usage gap alongside the cost, because that gap is where the saving is. Keep the idle and shared line visible so teams understand that headroom and overhead are paid for collectively, which builds the case for the eventual move to chargeback. This pairs with why Kubernetes bills are so hard to read and pod requests, limits, and real cost, inside the Kubernetes cost guide.

Make shared cluster cost visible and fair

We build showback that allocates direct, shared, and idle cluster cost on a key your teams accept, then turn the report into the rightsizing actions that move it. It is the same discipline behind the 31 percent median reduction we see in the first 90 days, with zero provider commissions. Our guarantee: we reduce your cloud spend or we reimburse our service fee.