Automatic Webhook Retries With Configurable Backoff
A destination that's down for thirty seconds shouldn't cost you an event. Hookbase retries failed deliveries on its own, up to 15 attempts over about 5 days depending on your plan, spacing them out with exponential backoff (or linear, fixed or custom delays on outbound endpoints) — and anything that still fails lands in a Dead Letter Queue you can inspect and replay instead of disappearing.
Four Ways to Space Out a Retry
Inbound routes use the exponential schedule. On outbound endpoints the backoff strategy is set per endpoint, so a flaky third-party API and an internal service you control don't have to share the same retry rhythm. A strategy changes the waits, not the number of attempts: with short linear, fixed or custom delays, the attempts run out well before the plan's window. Exponential waits get ±10% random jitter, so a batch of deliveries that failed together doesn't come back as one synchronized spike.
Exponential (default)
Waits grow with each attempt: 10s, 30s, 2m, 10m, 30m, 1h, 2h, 4h, 8h, 12h, then 24h. Each wait moves by up to ±10% at random. Quick enough to ride out a blip, patient enough to survive a longer outage. Inbound routes always use it.
Linear
A steady 30 seconds multiplied by the attempt number, capped at 10 minutes: 30s, 60s, 90s, 120s, and so on. Predictable spacing when you want retries evenly distributed. Outbound endpoints only.
Fixed
A flat 30 second delay before every retry. The simplest option, and the right one when your endpoint recovers on a known, short cycle. Outbound endpoints only.
Custom delays
Supply up to 10 delays of up to 24 hours each instead of a formula; the last one repeats if there are more retries than delays. Hookbase uses your exact schedule, so you can match a downstream provider’s documented recovery window. Outbound endpoints only.
What Exponential Backoff Looks Like
If you change nothing, this is the sequence a failed delivery follows. Where it stops depends on your plan. Each wait picks up its own jitter, so the real timings drift a little either side of the numbers below.
Example retry schedule (exponential, default)
Summed, that's about 13 minutes on Free, about 28 hours on Starter and Pay As You Go, about 3 days on Pro and about 5 days on Business before a message is exhausted. Want to see the numbers for a different strategy or attempt count? Work them out in the free retry calculator before you configure anything.
How Many Attempts You Get
Your plan sets the number of attempts; it isn't configurable. Inbound deliveries and outbound messages get the same budget, fixed when the delivery or message is created, so an upgrade doesn't lengthen retries already under way.
5 attempts
Free: over about 13 minutes.
11 attempts
Starter and Pay As You Go: over about 28 hours.
13 attempts
Pro: over about 3 days.
15 attempts
Business: over about 5 days.
Some deliveries keep 5 attempts on every plan: inbound routes with failover destinations (failover is decided by attempt count), sources in transient mode, edited replays, and FIFO outbound endpoints, where a failing message holds up the queue behind it. A plain replay starts a fresh set of attempts on your plan's budget.
See pricing for the full breakdown of what each plan includes.
When Retries Stop
Retrying forever is its own failure mode. Hookbase stops when the attempts run out, and waits while a destination is known to be down. Neither one loses your event.
Attempts exhausted → Dead Letter Queue
When a message uses up its retry attempts it moves to a Dead Letter Queue rather than being silently dropped. The failed delivery is still there to inspect — payload, destination, and what went wrong — and you can replay it manually once the endpoint is healthy again.
Circuit breaker open → delivery held
If the destination's circuit breaker is open, the delivery waits as held and spends no attempts, so a known-down endpoint doesn't burn through your retry budget while it's offline. Held deliveries are released once it recovers. Any still held at the end of the plan's retry window go to the Dead Letter Queue. Outbound endpoints have the circuit breaker on every plan; inbound routes on Pro and Business.
Why This Matters
Retry logic looks like ten lines of code right up until the first real outage.
Rolling your own retries
Hookbase
Jitter, not a stampede
When a destination goes down, everything queued for it fails at roughly the same moment. Without jitter, they'd all come back at the same moment too — knocking over the endpoint just as it recovers. Moving each exponential wait by up to 10% either way smears that spike out.
Do the math first
The free retry calculator computes a full delay schedule from a strategy and an attempt count, so you can see exactly how long a policy keeps trying before you commit to it. No account needed.
Stop Losing Events to Temporary Failures
Pick a backoff strategy and let Hookbase handle the rest — with a Dead Letter Queue behind it for whatever still doesn't make it. Get started free — no credit card required.