A period comparison puts the cost of one period beside another and explains the difference by rate, usage and mix. For the bill it is the first question asked every month: why is the amount different from last month? The answer depends on which periods, which cost basis and which currency are compared; a badly chosen comparison manufactures growth that is not there or hides growth that is. The cost report that shows the comparison is only as sound as those three decisions.
How it works
Every comparison answers a question, and the question chooses the periods:
| Question | Comparison |
|---|---|
| What changed since the last close? | month against the previous month |
| How is a seasonal business developing? | month against the same month a year earlier |
| Where is the trend going, without month boundaries? | trailing periods, such as 30 days against the previous 30 |
| Where does the current month stand? | the first n days against the same n days of the previous month |
Raw totals of two months are rarely comparable. Months have 28 to 31 days; a 31-day month against a 30-day one shows 3.3% growth at constant usage. Weekends and public holidays shift the picture for workloads that run on business days: a month with five weekends against one with four. That is why reports compare daily averages or the same number of days, and why a partial month is never set against a full one.
Then the cost basis. A team's trend reads best in effective cost, where commitments and discounts are spread over the usage they cover; the invoice reads in billed cost, with the purchase in the month it was made. Both are right, but never mixed: one month in billed cost against another in effective cost shows the reservation purchase as growth and its amortisation as a saving.
Then the currency. When the provider invoices in US dollars and the company reports in euros or pounds, every comparison carries an exchange-rate movement. A comparison in invoice currency shows usage; a comparison in reporting currency shows the effect on the result. Which rate applies to which month is a rule fixed before the comparison, not after it.
And the data itself: last month is not finished when the comparison is made. AWS may update the export delivery for the previous billing period within the first two weeks after it ended. Azure typically closes the billing period up to 72 hours after month end, usage charges can change until the fifth day after, and every cost is an estimate until the invoice is issued. On Google Cloud, usage reported late at month end can roll over to the next month's invoice, so the usage month and the invoice month diverge. An account added or removed mid-month changes the scope, and a commitment bought in one month and spread over the following ones moves the two sides differently.
AWS
Cost Explorer's Cost Comparison analyses two selected months, names the largest cost drivers by service, account and region and breaks the change down into usage and discount effects; the Top trends widget on the console home page shows the ten largest changes between the previous two months. Months are calendar months in UTC, and preconfigured ranges such as Last 3 Months exclude the current month; custom ranges come at daily or monthly granularity.
Azure
Cost analysis in Azure Cost Management shows, in its smart views, the change against the previous period next to the total. The average cost KPI compares, for ranges of three months or less, the period up to yesterday with the same number of days in the previous period, and for longer ranges the cost up to but not including the current month; today and the current month stay out of it. Summarised data goes back 13 months.
Google Cloud
Cloud Billing reports show in their header the percentage change against the previous period; the CSV download carries a column named "Percent change in subtotal compared to previous period", except for reports grouped by several dimensions. Two kinds of time range exist: charge period by usage date and billing period by invoice month, which includes taxes and adjustments in the footer.
What it means for your cost
The difference between two periods decomposes into a few causes: more or less usage, a different rate (a list price change, a commitment, a credit), a different mix (dearer instances for the same hours), a different scope (a new account) and a different exchange rate. A comparison that shows only the total names none of them; one grouped by service, account and usage type names them almost on its own. Allocation therefore comes before comparison: a report can only compare scopes it can tell apart.
The change never appears on the invoice as a figure. The invoice names one month; the comparison exists only in the report, built from two invoices or two exports. That makes the data cut-off a footnote to every comparison: which version of last month was used, and what has changed since? A comparison without a date and a cost basis is not reproducible, and one that leads to no decision in the review was a number without a question. What the review decides from it is covered in Monthly cloud cost review: agenda and templates.
A comparison differs from an anomaly: cost anomaly detection finds a deviation from history on the day it happens, while a comparison explains a month's total after the fact. Together they say whether an increase is an event or a trend.
Where it goes wrong
- A partial month against a full one. On the 20th the comparison shows a drop of a third, because a third of the month is missing. Only the same days of the previous month, or the daily average, compare.
- A day-count effect read as growth. March against February shows 10.7% more at unchanged usage, because 31 days stand against 28. The daily average shows zero.
- Billed against effective. The reservation purchase in January appears in billed cost as a jump and in effective cost as a spread over the term; a comparison that switches the cost basis invents a saving or a growth.
- A forgotten credit. Last month carried a credit that has expired; net cost rises at unchanged usage, and the comparison is reported as a usage increase.
- A comparison nobody attaches a decision to. The figure sits in the report, nobody tied a question to it, and next month it is repeated against another version of last month.
Sources
Read on 2026-10-15. AWS Cost Comparison in Cost Explorer, time ranges in Cost Explorer and Data Exports delivery and refresh; Microsoft start using cost analysis and understand Cost Management data; Google Cloud Cloud Billing reports.