Brief & Hire
Writing the brief

My automation keeps sending duplicates: what to ask a freelancer to fix

The short answer

An automation that sends the same email, post or row twice is almost never “glitching”. Something made a step run twice, and the automation had no record to check before acting again.

So the fix to ask for is not “make it stop”. It is three things: a record of what has been done, keyed by something that stays the same; a check of that record before every action; and a check of the real result after every error. Then a test that proves it.

Before you message anyone, collect the evidence below. It turns a vague “it sends duplicates” into a scoped fix with a fixed price. If you want the concept first, what “safe to re-run” means explains it in plain words.

First, collect the evidence

Freelancers diagnose duplicates from timing. Give them:

The gap between copies is the best single clue. Seconds apart usually means two copies of the automation running at once, or a trigger delivered twice. Minutes apart usually means a retry. Hours apart usually means a catch-up run, or the same item entering the automation a second time.

The five usual causes

1. A retry after an error that was not a failure. The service did the action and still returned an error, or the answer was lost on the way back. The automation retries and does it again. This happened to a publishing job of mine in August 2026: five posts were recorded as failed after API errors, and all five had already been published. The job was retrying finished work.

2. A trigger delivered twice. Some apps resend an event when they do not get a quick reply, which webhooks can do. The automation sees two events and acts on both.

3. A catch-up run on top of a scheduled run. A job that missed its time runs when the machine wakes up or the network returns, and the next scheduled run is right behind it. Both pick up the same item. My publishing pipeline has a catch-up slot for missed posts, and most of its rules are about when not to run:

Situation Catch-up? Why
The slot already published No It would double that day’s posts
Next regular slot is less than 30 minutes away No Two runs could pick the same item at once
A catch-up is already booked No Do not stack them
Two catch-ups already today No Alert a person instead of looping
Nothing went out and the cause was the network Yes Fifteen minutes later, once

4. Two copies running at once. A long run is still going when the next one starts, or the automation is installed twice. Both read the same list of “to do” items before either marks one done.

5. The same thing under two names. The automation’s record is correct, but the same content entered twice with different IDs. When I restarted my publishing queue in September 2026, I checked it first: old-format items were still queued ahead of the new ones, and one topic was in the queue in both formats. Started as it was, the same topic would have gone out twice, with no duplicate check catching it, because to the automation they were two different files.

What to ask the freelancer to fix

Once the cause is known, ask for the fix in these terms. They work whatever tool the automation is built in.

  • A record of what is done. A sheet, a table or a file, on your account, with one line per completed action. The key should be stable: an order ID, an email address plus a date, a post ID. Not a row number, which shifts.
  • Check before acting. Every step that sends, posts or charges looks at the record first and skips anything already there. My publishing job writes a “published at” time on each item; any later run, scheduled or manual, sees it and stops quietly.
  • Check after an error. On an error, look up whether the action happened before retrying. Only an action that really did not happen is a failure.
  • One runner at a time. If a run is still going, the next one waits or skips.
  • Catch-up rules written down. When a missed run is made up, and when it is not, as in the table above.
  • A definition of “the same”. What makes two items duplicates: same ID, same email, same title? Ask for this in writing, because cause 5 lives here.

Test the fix before you pay

The acceptance lines for workflow automation already include “run a finished job again by hand”. For a duplicate fix, ask for all five:

Test 2 is the one most often skipped, and it is the cause of most duplicates that survive a “fix”. The acceptance checklist for unattended automations has the wider version of these tests.

Cleaning up the copies already sent

Ask about this separately, because it is not the same job. Some services let the automation delete or correct what it sent; many do not. The publishing API I use has no delete call, so a duplicate post stays up until someone removes it by hand. Before the freelancer starts, agree who removes the existing copies, whether customers need a short note (“you may have received this twice”), and whether the record of what is done needs back-filling with items sent before the fix, so the new check knows about them.

Who to ask, and how to frame it

A duplicate fix in an existing workflow is usually a small, scoped order. A seller who offers short fix sessions, such as Ramon C with live one-hour sessions, can be a good fit when the workflow is in their tool. A larger pipeline with several sending steps is better quoted as a fix plus a review. If it keeps happening after fixes, the one-off fix or ongoing developer guide helps you decide whether it is time for a maintenance arrangement.

A message you can paste:

My automation [what it does] sends duplicates. Three examples with times are below, plus the run log. I would like:

  1. the cause, in a sentence;
  2. a record of completed actions on my account, keyed by [ID], checked before every send and after every error;
  3. only one run at a time, and written rules for catch-up runs;
  4. acceptance: re-run by hand, forced error after a send, two copies at once, a missed run, and the same item sent in twice. No duplicates in any of them. What would this cost, and what would you change?

A good reply names a likely cause from your timings before quoting. A reply that quotes without asking a single question about the timings is guessing.

Questions people ask

Why is my Zapier, Make or n8n automation sending emails twice?

Usually because a step ran twice: a retry after an error, a trigger delivered twice, a catch-up run on top of a scheduled run, or two copies running at once. The automation sends again because it does not keep, or does not check, a record of what it already sent.

Will turning off retries stop the duplicates?

It stops one cause and creates another problem: a short network blip now loses the work instead of repeating it. The better fix keeps retries and adds a check, before each action, of whether that action is already done.

How much does it cost to fix duplicates in an automation?

It depends on how many places the automation acts and whether the services it uses let you look up what was already done. A single workflow with one sending step is a small fix; a pipeline that publishes, emails and writes records in several places takes longer, because each place needs its own check.

From the field