Home/ Insights/ Build vs. buy for sending webhooks | What a managed webhook service takes off your plate

Build vs. buy for sending webhooks | What a managed webhook service takes off your plate

SaaS products and business systems usually use webhooks to tell a customer’s system that something happened: an order was confirmed, a payment went through. A single HTTP POST takes a few lines of code. Running webhooks as a feature your customers rely on is another matter: you need retries when delivery fails, signatures so receivers can reject forgeries, a delivery log, and a screen where customers register their own endpoints. More and more companies hand this part to a dedicated service. This article covers what building it in-house involves, what a managed service gives you, and what to check before you choose one.

What building webhook sending in-house involves

Here is what you build when webhooks are a customer-facing feature. The list also works as a checklist for estimates.

  • Queueing and retries: separate “event happened” from “event sent” by putting events on a queue. When an endpoint is down, retry with growing intervals (exponential backoff) over hours or days. Stripe documents that in live mode it retries with exponential backoff for up to three days. The same event can arrive twice, so attach an event ID the receiver can use to discard duplicates
  • Signing and key rotation: sign each payload (for example with HMAC) so the receiver can verify the sender. To handle leaked keys and scheduled rotation, sign with both the old and the new key for a transition period
  • Per-customer endpoints: each customer has URLs, subscribed event types and keys. You need create, update and delete APIs, plus permissions for who may change them
  • Delivery logs and replay: record when you sent what to which endpoint and what came back. When a customer says “we never got it”, you need a screen to check the log and resend
  • Failure alerts and auto-disable: alert your team when an endpoint keeps failing, disable endpoints that have failed for a long time, and tell the customer through another channel such as email
  • SSRF protection: customers type in the destination URL. If someone enters an address on your internal network or the cloud metadata endpoint, your sender will call into your own systems. You need a proxy that refuses internal addresses, or senders placed on a network that cannot reach internal services
  • Static source IPs: banks and large enterprises often only accept inbound traffic from allowlisted IPs. You need fixed egress addresses and a process for announcing changes in advance
  • A customer-facing portal: a screen where customers register endpoints, view their signing keys, browse deliveries and resend. Without it, every endpoint change and every investigation becomes a support ticket
  • Slow receivers: one endpoint that takes tens of seconds to respond can delay deliveries to every other customer on the same queue. You need per-endpoint concurrency limits and a way to pause an endpoint that keeps failing

The Standard Webhooks specification covers much of the sender’s side of this list, including the retry schedule, signing with old and new keys, SSRF protection, disabling failing endpoints and static source IPs. The header and signature format is explained in What is Standard Webhooks.

Companies that use dedicated webhook services

Svix and Hookdeck Outpost are among the services built specifically for sending webhooks. The examples below come from their published case studies and announcements.

  • Clerk (authentication): offers webhooks through Svix. According to the case study, Clerk estimated that building its own would take at least a month for a first version and at least six months with a skilled team for a full-featured, reliable implementation. The initial Svix integration took under two weeks
  • Resend (email API): its co-founder had built webhook infrastructure in-house at his previous company, WorkOS, and decided from the start not to build it again at Resend. Resend sends over 200 million webhooks a year through Svix. It began with Svix’s customer portal as is, and later used Svix’s React libraries to build a portal in its own design
  • Hookdeck Outpost: Hookdeck announced general availability of Outpost on April 23, 2026. Outpost is infrastructure for sending webhooks and events to customers’ destinations. It is open source under Apache 2.0, and the managed service and the self-hosted version run the same code. Besides webhooks, destinations include Kafka, AWS SQS and GCP Pub/Sub

Signature formats are converging too. In December 2023, Standard Webhooks was announced, initiated by Svix with a technical steering committee that includes Zapier, Twilio, Lob, Mux, ngrok, Supabase and Kong. With a shared format, receivers can verify signatures with the reference libraries, and senders can change their sending infrastructure without asking customers to change code.

What a managed service gives you

  • Time to market: no retries, signing, logs or portal to build. Your application calls an API when an event happens and registers endpoints
  • Reliability: retry schedules, disabling of failing endpoints and isolation of slow receivers are there from day one
  • Observability: delivery logs and endpoint responses are visible in a dashboard, so “we never got it” can be answered from the log
  • Customer self-service: customers register endpoints, check keys and resend on their own, which cuts support requests
  • IP-allowlisted customers: if the service offers static source IPs, customers only need to allow those addresses

What to check before you choose a service

  1. Pricing model: most services charge by message volume. Check whether retries count, and whether sending stops or overage is billed when you pass the limit. Run the numbers on your expected volume
  2. Where the payload goes: payloads pass through the service and are kept as delivery logs for a period. Check retention, storage region, and how the service is treated as a processor if payloads contain personal data. Sending only IDs and letting receivers fetch details from your API reduces what you hand over
  3. Signature format: if signatures follow a public specification such as Standard Webhooks, your customers’ verification code survives a later move to another service or back in-house
  4. Outage impact: if the service is down, so is your sending. Check the uptime SLA, whether it keeps accepting messages during an incident, and how it publishes status
  5. Contract and support: check the language of the terms, the governing law, billing currency and support channels against your vendor review process

Building in-house vs. using a managed service

ItemBuild in-houseUse a managed service
Initial workDesign and build queueing, retries, signing, logs and screensIntegrate the send API and endpoint registration
Ongoing maintenanceYour team monitors the platform, handles incidents and manages keysThe provider runs the platform; your team handles failing endpoints
Retries and delivery logsYou define the schedule, retention and replay screensBuilt in; retention usually depends on the plan
Customer-facing portalYou build it, with full control over the designUse the provided portal, or build your own on the API
Static source IPsSet up egress networking and manage the addressesPick a service and plan that offers them
Cost modelEngineering time for build and upkeep, plus serversMonthly fee based on message volume or similar
Where payloads are storedYour systemsThe provider (check retention and region)

Building in-house fits better when your volume is so high that per-message pricing does not work, when internal policy forbids payloads leaving your systems, or when the sending infrastructure itself is what sets your product apart. If payloads cannot leave your systems, you can also run an open-source sending platform in your own environment.

About Webhook Admin

We offer Webhook Admin, a service that takes over webhook sending. Call the API once where the event happens, and it signs the message in the Standard Webhooks format, sends it, and if delivery fails, retries it. By default it makes up to eight attempts (including the first) over about 28 hours, and you can choose the number of retries and the interval per endpoint. Endpoints that keep failing for five days are disabled automatically, and failures are reported to Slack, Teams, Chatwork, email and more. Static source IPs and a customer-facing portal are included (plans and pricing are on the product site). If payloads cannot leave your systems, self-hosting in your own environment is available on request.

Sources

FAQ

Where does the effort go when you build webhook sending yourself?

Not in the HTTP request itself. It goes into retries when delivery fails, signing and key rotation, delivery logs with manual replay, alerting on and disabling endpoints that keep failing, and managing each customer's endpoints. Some customers will also require a static source IP.

If a managed service sends our webhooks, does the payload pass through a third party?

Yes. The payload goes through the service and is kept for a period as part of the delivery log. Check the retention period and the storage region before signing up. Sending only IDs, with no personal data in the payload, reduces what you hand over.

Can we move from a managed service to another one, or back in-house, later?

If the signature follows a public specification such as Standard Webhooks, your customers' verification code keeps working after the move. With a proprietary signature format, your customers will need to change their receivers when you migrate.

Back to Insights

Let's talk.

Everything you share is handled in confidence. We reply within two business days.