· Natrajan
How to stop duplicate SMS in lead follow-up
Why a new lead sometimes gets the same welcome text twice or three times, and a simple setup that sends exactly one.
A new lead fills in your form. Your automation texts them within a minute: "Thanks for reaching out, when's a good time to talk?" Good.
Then a second identical text arrives. Sometimes a third. The lead now thinks you're a bot, and the first impression you paid for with ad spend is gone.
Duplicate SMS is one of the most common complaints from businesses that automate lead follow-up. The cause is almost never the SMS provider. It's that something upstream ran twice, and nothing in the workflow noticed.
This post walks through where duplicates come from, how to stop each one, and how to tell whether the text actually did its job.
Where duplicate texts come from
There are three usual sources, and each needs a different fix.
1. The form submits twice. A lead double-clicks Submit, the page is slow and they press it again, or a browser retries a request that timed out. Your form tool sends two webhooks for what is really one lead.
2. The webhook is delivered twice. Most webhook senders retry when they don't get a fast success response. If your automation took too long to answer the first delivery, the sender assumes it failed and sends it again. Now two workflow runs start for the same submission.
3. The send step is retried after an unclear failure. The workflow calls the SMS provider, the request times out, and the workflow retries. But the first request may have gone through. Sending an SMS is not safe to repeat: every successful call is a new message.
The first two are duplicate triggers. The third is a duplicate action. Fix both.
Fix 1: give every lead an idempotency key
An idempotency key is a value that is the same every time the same real-world thing happens. For a form, the best key is the submission ID your form tool already creates. If there isn't one, combine the form name, the lead's phone number and the date.
When the automation receives a trigger with a key it has already seen, it should do nothing new and return the original result.
In OutcomeOS, a webhook trigger accepts the key in an Idempotency-Key header, or you point the trigger at a field in the payload, such as the submission ID. A second delivery with the same key returns the run that already exists. It does not start a new one, so it does not send a second text.
This one setting removes duplicates from double submits and webhook retries.
Fix 2: don't retry a send that may have worked
Retries are good for safe steps, such as looking up a contact. They are dangerous for steps that create something, like a text, an email or a charge.
The rule we use: if a send step fails in a way where you can't tell whether it went through (a timeout, a dropped connection), don't retry it automatically. Hold it for a person to check.
In OutcomeOS each step has its own retry policy and idempotency key. A step that isn't safe to repeat and fails ambiguously goes to "Requires attention" instead of running again. You look at the provider's message log, see whether the text went out, and either mark it done or send it once.
When you replay a failed run, steps that already succeeded reuse their earlier output, so the welcome text isn't sent again on the way to the step that broke.
Fix 3: one contact per person
Duplicates also hide in your contact list. If the same lead appears as two contacts, two follow-ups start.
OutcomeOS matches incoming leads to existing contacts by email or phone, so a lead who fills in your form twice in a day is one contact, with both submissions on their timeline.
Measure the reply, not the send
Stopping duplicates is half the job. The other half is knowing whether the one text you sent worked.
"SMS sent" tells you the workflow ran. It doesn't tell you the lead answered. For lead follow-up, the result you care about is a reply within a set time.
In OutcomeOS that becomes an outcome:
- Starts when: a new lead arrives (form link, webhook or API).
- Achieved when: the lead replies. Replies come in through your Twilio number. OutcomeOS checks Twilio's signature on each incoming message and records it as "Customer reply received" on that contact.
- Time limit: for example, 2 hours.
- Alert: an email (or a text to you) when leads are still waiting near the limit.
Now your dashboard shows how many leads replied, not how many texts went out.
A hypothetical example
Say a home-services business gets 100 form leads a week. Before the fix:
| Before | After | |
|---|---|---|
| Leads | 100 | 100 |
| Welcome texts sent | 118 (double submits and webhook retries) | 100 |
| Leads who got 2+ identical texts | 15 | 0 |
| What the report says | "118 texts sent" | "41 of 100 leads replied within 2 hours" |
These numbers are made up to show the shape of the change. The second report is the one that tells you whether your follow-up is working, and which leads someone should call.
Checklist
- Find a stable ID for each lead (form submission ID, or form + phone + date).
- Send it as the idempotency key on the trigger.
- Mark SMS, email and payment steps as not safe to retry automatically.
- Review "Requires attention" daily; check the provider log before resending.
- Track "lead replied within X hours" as the result, not "text sent".
- Get an alert while leads are still waiting, not a report after they've gone cold.
How OutcomeOS helps
OutcomeOS drops duplicate triggers with idempotency keys, holds unclear send failures for a person instead of retrying them, and tracks each lead until they actually reply. It works with Twilio SMS and WhatsApp, signed webhooks, form links and an API.
You can start free, or see how it compares with the tools you use today.