Practical guide
Estimate a serverless API budget from users and polling
forecast API Gateway request cost from user activity. Includes a worked example, calculator scope, and practical questions.
Updated · Sources checked
Count automated calls as well as user actions
API volume comes from application behavior, not just active-user totals. Polling, retries, background synchronization, and repeated page components can generate calls without a new button click. Model those sources explicitly before estimating an HTTP API request bill.
A useful traffic worksheet separates interactive actions from time-based activity. This also makes it easier to identify which behavior drives a surprising request count.
Worked example: a dashboard with polling
Assume 5,000 daily users each keep a dashboard open for 30 minutes. Polling once every 30 seconds produces 60 calls per user session: 300,000 polling calls per day, or nine million in a hypothetical 30-day month.
Add eight interactive calls per user per day. That contributes another 1.2 million calls per month, for a total of 10.2 million before retries. If every payload falls within one billing increment, the usage also equals 10.2 million request units.
Changing the polling interval to 60 seconds halves polling calls to 4.5 million. With interactive activity unchanged, total calls become 5.7 million. The difference is 4.5 million calls, approximately 44.1% of the original total. This is a traffic scenario, not a recommendation to accept a particular data-freshness delay.
If the calculator accepts whole millions, bracket this scenario with adjacent supported inputs and keep the precise volume in the planning worksheet. Do not hide rounding uncertainty behind an overly precise displayed dollar amount.
Check experience and backend effects
A longer interval changes freshness. Test it against the dashboard’s actual requirements and consider behavior when a browser tab is inactive. Count retries separately if errors or connectivity changes increase them.
The tool prices supported HTTP API request units only. It excludes backend compute, database work, transfer, and logging. Fewer calls can reduce those workloads too, but savings depend on their own billing models and whether provisioned resources remain running.
Frequently asked questions
Can I estimate from monthly users alone? No. You also need session duration, call frequency, and payload billing units.
Does half the polling mean half the whole application bill? No. Interactive calls remain, and fixed or unrelated service costs can stay unchanged.
Calculate your scenario
Use the api gateway cost calculator. Keep the displayed region, pricing date, billing units, and exclusions alongside your result. The arithmetic examples above illustrate usage or explicitly hypothetical rates; they are not AWS quotes.
Source
Billing structure checked against official AWS documentation on September 6, 2026.