Retry logic decides whether your automations heal or fail
The dashboard shows green, the workflow ran, but the subscriber never received their welcome sequence. Somewhere between step three and step four, an API timeout occurred, the automation made a choice, and no alarm went off.
Retry logic decides whether your automations heal or fail silently
Workflow tools handle failed steps differently—some replay automatically, others quit and log nothing.
When a webhook times out or an API returns a 500 error mid-workflow, your automation platform makes a decision: retry the step, skip it, or halt the entire run. That decision is governed by retry logic—rules baked into the platform that determine how many attempts are made, how long the system waits between them, and what happens if every attempt fails. The problem is that these rules vary wildly between tools, and most operators never see the settings until a workflow breaks in production.
Zapier retries failed steps up to three times over roughly two minutes, then marks the task as errored. Make retries once, immediately, with no backoff. n8n offers configurable retry attempts and intervals, but defaults to zero retries unless you explicitly enable them. ActivePieces and Pipedream sit somewhere in between. If you’re chaining automations across platforms—Zapier triggering Make triggering a custom script—one platform’s aggressive retry can collide with another’s rate limit, creating duplicate records or burning through API quotas. The feature article walks through what each major platform does when a step fails, how to audit retry behaviour in existing workflows, and when to add manual error-handling nodes instead of trusting the defaults.
TACTIC
When social schedulers hit API rate limits, posts vanish or queue silently
Every social platform meters API calls differently—Twitter caps at 50 posts per three hours for most app tiers, LinkedIn counts read and write operations separately, and Instagram’s Graph API has per-user and per-app limits that reset on different clocks. When your scheduler crosses the threshold mid-batch, some tools retry after the window resets, others drop the post entirely, and a few queue it locally without telling you it never published. Buffer pauses the queue and emails you. Hootsuite marks posts as “failed” but doesn’t retry. Knowing which behaviour your tool defaults to means you can audit your calendar before a campaign publishes half its content.
WORTH READING
AI token counting varies between platforms—and you’re billed on the discrepancy
OpenAI’s tiktoken library counts tokens one way, Anthropic’s API counts them another, and most third-party wrappers round up. The mismatch means a prompt you tested at 1,200 tokens locally might bill as 1,350 tokens in production, and that difference compounds across thousands of requests. Some platforms count system messages and function definitions toward your quota, others don’t. A few round to the nearest hundred tokens for billing but display exact counts in logs. If you’re running AI workflows at scale—summarising newsletter archives, generating social captions, answering reader questions—auditing token counts across platforms can cut your monthly bill by fifteen to twenty per cent without changing a single prompt.
FROM THE ARCHIVE
WordPress staging environments clone your database but skip critical connections
Staging copies your database tables and uploads directory, but most hosts don’t replicate external API keys, webhook endpoints, or CDN rewrites. Push a plugin update live and your transactional email integration breaks because the staging environment was still pointing at a sandbox API key. Payment forms submit test transactions. Backup scripts write to the wrong S3 bucket. The cloning process duplicates everything inside wp-content and wp_options, but it doesn’t rewrite hardcoded URLs in serialised data or update DNS-dependent integrations. Knowing what transfers and what stays bound to production means you can test changes without discovering silent failures three days after deploy.
Know someone who would like this? Forward today’s email—every operator we reach is one closer to running an online business with a little less friction.