Recover Abandoned Shopify Carts with WhatsApp and n8n
By Catalina Jaramillo · AI Commerce & Revenue Operations Consultant · 2026-11-09
You recover abandoned Shopify carts with WhatsApp and n8n by connecting Shopify's checkout webhook to an n8n workflow, waiting a set delay, checking that the order still hasn't been placed, and sending a WhatsApp template message through a provider like Meta Cloud API, Twilio, or 360dialog. The workflow itself is the easy part; the timing, the opt-in, and the message logic decide whether it recovers revenue or gets your number blocked.
What does the abandoned cart flow actually look like step by step?
A working flow has six stages, and each one can fail independently. Build it in this order so you can test each stage before adding the next.
The trigger is not the cart creation event. Shopify fires checkouts/create when a customer reaches checkout, and checkouts/update when they change something. You want the update event with a completed checkout left empty, or the create event followed by a delay and a verification step.
- Trigger: Shopify webhook on checkouts/create or checkouts/update, filtered to carts with an email or phone and a non-zero total.
- Delay: a Wait node in n8n, typically 30 to 60 minutes for the first message. Shorter feels pushy, longer loses the intent.
- Verification: before sending, query Shopify for orders matching that checkout token or customer. If an order exists, stop the branch. This single check prevents most embarrassing mistakes.
- Channel check: confirm the customer has a phone number and a valid WhatsApp opt-in. If not, route to email instead.
- Send: call your WhatsApp provider's API from an HTTP Request node with an approved template message.
- Log and branch: write the result back to Shopify (tag, note, or metafield), then schedule the second message only if the first was delivered and no order appeared.
Which WhatsApp provider should you connect to n8n?
n8n has no native WhatsApp node that covers business messaging, so you call an API. The three realistic options differ in cost, setup effort, and how much control you keep.
Meta Cloud API is the cheapest per message and the most direct, but you manage template approval, phone number verification, and quality rating yourself. Twilio and 360dialog wrap that complexity in a friendlier API and dashboard, at a higher per-conversation price. For a store doing a few hundred carts a month, the wrapper usually pays for itself in saved setup time.
- Meta Cloud API: lowest per-message cost, direct control, you own template approval and number quality.
- Twilio: fastest to connect, good n8n documentation, higher unit cost.
- 360dialog or similar BSP: middle ground, useful if you want support in Spanish or Portuguese.
- Whatever you pick, use template messages for the first contact. Free-form messages only work inside the 24-hour customer service window, which an abandoned cart almost never falls into.
Carolina, Catalina's AI agent, reviews your business, shows which system solves it and what it would cost. If there is a fit, she books a call with Catalina.
Tell Carolina about your case →What should the messages say, and how many should you send?
Two messages is the practical ceiling for most stores. A third rarely earns its cost and increases the chance the customer blocks the number, which damages your sender quality rating for every future message.
The first message should reference the specific product, not the cart in general. The second should add something the first didn't: a question, a shipping detail, or a reason to return. Repeating the same nudge twice trains people to ignore you.
- Message one, 30 to 60 minutes after abandonment: product name, one line, direct link back to checkout.
- Message two, 20 to 24 hours later, only if no order and no reply: a question or a concrete objection handler, such as delivery time or payment method.
- Never send between roughly 10pm and 7am in the customer's timezone. Schedule the Wait node against local time, not server time.
- Include a clear opt-out line. It protects your number and it is required by WhatsApp's business policy.
- Keep the link short and trackable so you can tell which message produced the order.
What breaks these flows in production?
Most failures are not code failures. They are data and policy failures that only appear once real traffic hits the workflow.
The most common one is duplicate sends. If Shopify retries a webhook, n8n processes it twice and the customer gets two identical messages. Deduplicate on the checkout token before the send node.
- Duplicate webhook deliveries causing double messages. Deduplicate by checkout ID.
- Missing or invalid phone numbers. Validate the format and country code before calling the API.
- Customers who already ordered through another device or session. Always re-check orders before sending.
- Template rejection or paused templates. If your template gets paused, the send node fails silently unless you handle the error branch.
- Timezone drift. A Wait node set to 24 hours is not the same as 'tomorrow morning in the customer's timezone'.
- No logging. Without a record of what was sent and what converted, you cannot improve the flow.
What does it cost to build and run this?
The software is cheap. The design, the template approval, and the measurement are where the real cost sits. Below are the actual engagement prices for this kind of work, not market estimates.
A single-channel WhatsApp agent setup starts at USD 790 configuration plus USD 199 per month. A broader e-commerce and post-sale system, which is where abandoned cart recovery usually belongs, runs USD 5,500 to 8,500 configuration plus USD 1,200 per month. If you only need the workflow wired up and measured, an express automation runs USD 500 to 1,500.
- Agent Starter for WhatsApp: USD 790 configuration + USD 199/month.
- Express automation: USD 500 to 1,500.
- E-commerce / Shopify plus post-sale system: USD 5,500 to 8,500 configuration + USD 1,200/month.
- Strategic diagnosis with Catalina: USD 490, useful before committing to a build.
- Ongoing optimization: USD 1,000 to 2,000+ per month.
How would Catalina approach this?
Catalina Jaramillo is an AI Commerce and Revenue Operations consultant based in Pereira, Colombia, working remotely with companies in any country. She founded Professional Glam, a business that grew to four physical locations and now operates digitally with 9,296 documented online orders, so she has run this problem from the operator's side, not only the consultant's.
Her working method is diagnosis first, then design, then a scoped pilot with metrics captured before and after. For abandoned carts that means mapping where the store actually loses checkouts, checking whether the phone data is usable, and only then deciding whether WhatsApp, email, or both should carry the recovery. Carolina, her AI agent, coordinates international conversations in the client's language and confirms the interpretation format before each meeting.
The pilot is deliberately small: one segment, one or two messages, a fixed window, and a clear before-and-after comparison. If the numbers do not justify expanding, the honest answer is not to expand.
- Diagnosis: where carts are lost, what contact data exists, what the current recovery looks like.
- Design: channel choice, message sequence, timing rules, and the verification logic.
- Pilot: one segment, one or two messages, measured against a defined baseline.
- Decision: expand, adjust, or stop, based on the pilot numbers rather than on assumptions.
Frequently asked questions
FAQ
Can n8n send WhatsApp messages without a third-party provider?
No. n8n orchestrates the workflow, but WhatsApp business messages must go through an approved API such as Meta Cloud API, Twilio, or a BSP like 360dialog. You connect to that API from an HTTP Request node inside n8n.
How long should I wait before sending the first recovery message?
Between 30 and 60 minutes after checkout abandonment works for most stores. It is long enough that the customer has clearly stopped, and short enough that the intent is still fresh. Anything under 15 minutes tends to feel intrusive.
How many messages should the flow send?
Two is the practical limit for most stores. A third message rarely earns its cost and raises the risk of the customer blocking the number, which lowers your sender quality rating and affects every future message you send.
Do I need an opt-in to message abandoned carts on WhatsApp?
Yes. WhatsApp business policy requires prior opt-in for template messages. In practice this means capturing consent at checkout or through a signup, and storing it alongside the phone number so the workflow can check it before sending.
What is the biggest mistake in these flows?
Sending without re-checking whether the order was already placed. Shopify webhooks can arrive late or twice, and customers often complete the purchase in another session. A verification step before the send node prevents both duplicate messages and messages to people who already bought.
Carolina, Catalina's AI agent, reviews your business, shows which system solves it and what it would cost. If there is a fit, she books a call with Catalina.
Tell Carolina about your case →