Amortised Cost View

A team's cloud report drops 30% in March. Nobody on the team changed anything. The platform group bought a three-year commitment and the discount happened to land on that team's instance family. In April the report rises again because new capacity started on undiscounted rates. Two months of reporting told the team nothing about its own behaviour, and it will now discount the report permanently.

An amortised cost view spreads the up-front portion of a commitment across the term it covers and applies the resulting effective rate to each team's measured usage. The property that matters is not accounting elegance: it is that the number a team sees is a function of that team's decisions alone.

Why it matters

Cost reporting exists to change engineering behaviour. That requires the reported number to respond to the reader's actions and to nothing else. Unblended cost - the rate as billed on the line item - violates this constantly: identical usage is priced differently depending on whether a commitment happened to cover it that hour, which is a purchasing decision made by someone else.

There is a second effect. Amortisation keeps the commitment decision with the group that can make it. When teams see undiscounted rates they start buying their own reservations to look good, which fragments coverage and reduces the organisation's aggregate discount. When they see amortised rates, purchasing stays central and the savings show up where they were earned.

Implementation patterns

• Pick one of the three views and stay on it. Unblended is the billed rate. Blended averages the rate across an account and hides both saving and waste. Amortised spreads commitment cost across its term and charges the effective rate.
• Report platform purchasing separately as commitment coverage and utilisation, so the platform team's performance is visible rather than smeared across product teams.
• Amortise at the granularity you allocate at, usually daily, so a month-boundary purchase does not create a step in anyone's chart.
• Show both the amortised figure and the on-demand equivalent on the platform's own report. The gap is the value the commitment programme created, and it is the only honest way to defend the programme's existence.
• Version the methodology and announce changes in advance. A methodology change is indistinguishable from a real change to everyone reading the chart.
• Set a precision target explicitly - accurate enough to rank teams and to spot trends, typically around 10%, not accurate enough to invoice.

Industry example

Every major provider's cost export ships unblended, blended and amortised columns, and their documentation is consistent about which is for which purpose: unblended for reconciling an invoice, amortised for understanding what usage actually costs. Vendors would not carry three representations if one were sufficient, and the fact that most organisations discover the distinction after their first showback argument is what makes it worth naming. In production the practice that survives is the boring one: one view, documented, unchanged for a year at a time.

Failure scenarios

• Mid-year methodology switch. Every team's number moves by double digits for reasons unconnected to their work, and trust in the report does not come back.
• Teams buying their own commitments because they are judged on unblended cost. Coverage fragments, flexibility falls, and the aggregate discount gets worse while every individual team's chart improves.
• Amortisation hiding an expiring commitment. The effective rate is stable right up to the term end, then steps up with no warning to anyone reading the amortised view. Coverage expiry needs its own alert.
• Allocating the residual. Spreading unattributable shared cost across teams to make the numbers add up destroys the property the whole practice exists to provide.

Trade-offs

Amortisation also costs a little credibility with finance, because the number no longer ties to a payment. The answer is to run both: unblended for the ledger, amortised for the engineering conversation, and say plainly which report is which.

When not to use it

If there are no commitments, there is nothing to amortise and unblended is simply correct. If the report's purpose is contractual - billing an external customer, or a regulated cost-recovery calculation - the view must reconcile to the invoice, and unblended wins regardless of its incentive properties. And if teams have genuine authority over their own purchasing, showing them unblended cost is the right signal, because then the purchasing decision is theirs to make.

Interview question

Q: Two teams with identical infrastructure see costs differing by 25% in the showback report. Neither changed anything. What is the likely cause, what would you change in the report, and what would you refuse to change?

What a strong answer covers: commitment coverage landing unevenly across unblended line items as the likely cause; moving to an amortised view so usage drives the number; reporting coverage and utilisation separately as platform metrics; refusing to allocate the unattributable residual onto teams; refusing to change methodology mid-period without an announced transition; and stating a precision target rather than pursuing exactness.

Quick check

Quiz: Why can unblended cost make a team look worse for a decision it did not make? — Because the billed rate depends on whether a centrally purchased commitment happened to cover that usage hour. Amortised cost applies the effective rate to usage instead, so the team's number tracks its own behaviour.

Flashcard: What does an amortised showback view deliberately hide, and where should that information go instead? — It hides commitment purchasing timing and performance. That belongs on the platform team's own report as coverage and utilisation, with the gap to on-demand shown as the value created.