6 attempts, with 15m 30s of waiting
- Total attempts
- 6
- Total wait
- 15m 30s
- Delay cap
- 1h
The initial attempt starts at 0 seconds. Times below include waiting only; request duration and timeouts add to elapsed time.
| Retry | Delay | Cumulative wait |
|---|---|---|
| 1 | 30s | 30s |
| 2 | 1m | 1m 30s |
| 3 | 2m | 3m 30s |
| 4 | 4m | 7m 30s |
| 5 | 8m | 15m 30s |
How exponential backoff works
After each unsuccessful attempt, multiply the previous delay until it reaches your cap. Here, retry 1 is the first retry after the initial attempt. The cap applies even when it is smaller than the initial delay.
delay(n) = min(cap, initial delay × multiplier^(n − 1))
cumulative wait = sum of the delays through retry n
full jitter: delay sampled uniformly from 0 to delay(n)
expected full-jitter wait = sum of delay(n) ÷ 2For five retries starting at 30 seconds with a multiplier of 2 and a 3,600-second cap, the delays are 30, 60, 120, 240, and 480 seconds. Total waiting is 930 seconds (15 minutes 30 seconds), across six attempts. With full jitter, expected waiting is 465 seconds (7 minutes 45 seconds).
Displayed times round to the nearest millisecond; totals use unrounded values. These are configurable examples, not a provider’s retry policy or a complete SMTP mail-server queue policy.
When should I retry an email request?
Retry only failures your provider treats as retryable. Follow its retry-after guidance where available and bound the total attempts. If your provider has already accepted an email and is retrying delivery, submitting a new copy can cause duplicates.
A timeout can leave acceptance uncertain. Use the provider’s documented idempotency behavior and check the original request before resubmitting. Don’t keep retrying permanent address or policy failures unchanged.
Sources and references
Use these primary sources for the underlying behavior. Stampwing’s explanations and calculators are practical aids; your provider’s response and configuration still matter.