Who this recommendation is for
This text is written for the people who decide how webhooks from the InnerWeb system reach their recipients: the IT director, the integrator and the application administrator. It assumes the recipient (n8n, a plant system or a messenger) already exists and that the way of delivery has to be chosen.
The switch is the webhook delivery mode setting, with the values direct (default) and outbox. Any value other than outbox means direct, which matters for the typos described in recommendation six.
Recommendations 1 to 3: choosing the mode
Below are three recommendations of a decision-making nature. Each one is justified by the way the mechanisms work.
1. An installation with outbound traffic: stay with direct
The default mode is the simplest: the application server sends the webhook to the recipient by itself. It needs no extra component that has to be supervised. If the network of the installation, for example in the cloud, allows outbound connections to the recipient's address, there is no reason to add a poller.
2. A network that blocks outbound traffic: choose outbox
Outbox mode was built for on-premise installations where the network blocks all outbound traffic from the application host, and an alarm webhook sent directly ended in a timeout and was lost. The poller connects to the application over an inbound channel itself, so the firewall stays unchanged.
3. Alarm webhooks: delivery reliability matters more than simplicity
If the webhooks carry no-movement alarms or other security events, losing a message is costly. In that case outbox makes sense even where the network formally allows the traffic but is unstable: the entry waits in the table instead of being lost.
Recommendations 4 and 5: effects on reports and supervision
Recommendation four: decide where the history for the Communication Channels report will come from. In direct mode, each webhook also writes a finished row with the status delivered or failed as history for this report, and such a row never reaches the poller. In outbox mode, the history is made by the entries themselves, whose state changes with the acknowledgements.
Recommendation five: with outbox, assign an owner of the poller. The mode moves responsibility for delivery from the application to a workflow that has to be monitored. There is no retry limit, no delay between retries and no alert for stuck entries, which means a failed acknowledgement returns the entry to pending and increases the attempt counter, so an undelivered event is retried without limit.
Recommendations 6 and 7: changing the mode and housekeeping
Recommendation six: after every change of mode, always run a test event. Because any value other than outbox means direct, a typo in the setting will silently bring back direct delivery, and on a network with a block the webhook will be lost without a visible error for the user.
Recommendation seven: plan the cleanup. There is no job that removes finished entries from the table, so they grow without end. With high webhook traffic, set up a periodic review and a retention policy on the administrator's side.
The comparison in one place
In short: direct wins on simplicity and needs no poller, but it requires outbound traffic from the application host and does not retry lost messages. Outbox wins on delivery reliability and compatibility with a firewall that blocks outbound traffic, but it requires supervision of the poller and a manual review of the table.
The most common decision mistake is choosing outbox for reliability alone, without assigning an owner of the poller. The second is leaving direct on a network with a block and hoping an exception will be added to the firewall later.
Video
Frequently asked questions
Which mode is the default?
The default is direct. Any setting value other than outbox is treated as direct.
Do I get history in the Communication Channels report in direct mode?
Yes. Each webhook also writes a finished row with the status delivered or failed as history for this report.
Can the mode be changed during operation?
Yes, the setting can be changed. After the change, run a test event and check that the entry arrived as expected.
Not sure which mode to choose?
Describe your installation network to us. We will select the delivery mode and name who should supervise the poller.
Let's talk