Unblended, amortized and net AWS costs: choose a comparable number

1. Write down the decision

A finance colleague asks whether the dashboard agrees with the bill. An engineer asks whether running a service became more expensive this week. Those are different comparisons. Write the question at the top of the report before choosing a metric.

For a bill check, begin with the provider statement and its included charges. For an operating review, decide how a commitment should be spread over the period it benefits. Keep a bridge between the two views; do not silently replace one with the other.

2. Understand the cost labels

AWS describes unblended cost as the cost of usage, with discounts separable into their own line items. Amortized cost spreads relevant upfront and recurring reservation charges over their billing period. Net views incorporate applicable discounts. These labels describe accounting treatments, not different measurements of workload efficiency.

A daily spike when a commitment is charged can disappear in an amortized view without any infrastructure change. Conversely, a lower net total may reflect a discount rather than reduced consumption. Keep usage and commercial changes separate in your explanation.

3. Define a comparison contract

Save the report settings alongside the export. Someone should be able to recreate the number without guessing which filters were active. A report title such as “September AWS spend” is not enough to identify its scope.

• Name the payer and included accounts, currency, dates and timezone.
• State the cost basis and treatment of commitments.
• Record whether credits, refunds, tax, support and Marketplace charges are included.
• Record the successful billing-data update and whether the period is final.

4. Explain the difference without deleting it

Compare matching scopes first. Then keep a ledger of documented adjustments: excluded charge types, different period cutoffs or the commitment treatment. An adjustment needs an amount and evidence. “Other” is not an explanation.

Do not force an unexplained residual to zero or distribute it across teams. If a source is missing, mark the report incomplete. Only compute a percentage difference when the reference denominator is meaningful; a zero bill cannot support the usual percentage calculation.

5. Freeze the basis for the next review

After the comparison is understood, save the contract with a version and owner. Apply the same basis to the next month and to the baseline of any savings experiment. If a policy changes, explain its effect before comparing periods.

Separate “we bought a different commitment”, “usage changed” and “the report definition changed” in the action log. This makes a falling chart useful evidence rather than an ambiguous success signal.