Ask customers how the order went, and hear the answer in their words
A Loop Voice agent calls a few days after delivery, asks two or three short questions and turns anything that went wrong into a ticket for your team.
- Order fulfilledA wait covers delivery and some use
- A share of ordersA split step samples the orders
- Two or three questionsA follow-up when an answer is short
- Problems become ticketsIn Loop Voice and your helpdesk
- Answers read weeklySearch, filter or export to CSV
The short answer
A post-purchase survey call is a short check-in a few days after delivery: two or three questions, the customer's own words in a transcript, and a ticket when something went wrong. In Loop Voice it runs as a flow that starts when an order is fulfilled. Our September 2026 benchmark doesn't track surveys as a use case, so there are no Loop Voice survey response rates to publish yet; the numbers to watch are below.
The closest real calls: two post-purchase check-ins
We don't have a recorded survey call to share yet. These two check-ins for a pet chew store show the same moment: one customer reported the wrong item and got a replacement; the other said the chews were "Perfect as is."
Full transcripts are in the call library.
A star rating says whether a customer is happy, not why, and a form can't ask "what happened?" when the answer is a 3. A short call can, while the delivery is still fresh.
When it works, and when to skip it
Works well for
- Products where fit, taste or results vary from one customer to the next.
- New products, where the first weeks of feedback decide what to change.
- Stores that will act on what they hear: a product page fix, a packaging change, a support follow-up.
- Customers who agreed to calls (calling rules).
Skip it for
- Every order, every time: survey a share of orders or one segment, so regular customers aren't called after each purchase.
- Orders with an open support ticket: your team should call, not a survey.
- Questions your data already answers, such as whether the parcel arrived on time.
Set it up in Loop Voice
Start from order fulfilled
Build a flow on the order fulfilled event from Shopify, WooCommerce, BigCommerce, Magento, Shopline or Checkout Champ, and add a wait step of a few days so the product has arrived and been used.
Write two or three questions
For example: how did the order arrive, how is it working out, what would you change? Turn on the gather feedback objective, and tell the agent to ask a follow-up when an answer is short.
Decide where answers go
Problems become a ticket with the AI summary, the order and the recording, in Loop Voice and in Gorgias, Zendesk or Redo. Keep review asks out of this flow. Run review requests separately, with the same ask for every customer whatever they said on the survey, and nothing offered in return.
Sample and cap
Send only a share of orders down the call path with a split step, and cap calls per customer so nobody is surveyed after every order.
Read the answers every week
Search and filter the call log, read the summaries, export it to CSV, or ask Loop IQ about the week's calls.
What to set on the survey flow
- Wait step
- Count the wait from fulfillment, covering shipping plus enough use to have an opinion: arrival is enough for a packaging question, while a product that works over weeks needs weeks. If you also run review requests, whose template calls 3 to 5 days after delivery, space the two waits apart so one order doesn't bring two calls in a few days.
- Split step
- The split sends a set percentage of orders down the call path and the rest down a path with no call. Name the paths survey and no call so live tracking shows how many went each way. Example: at 20%, a store with 500 orders a month calls about 100 customers, and fewer of them will pick up.
- Objectives
- With gather feedback on, leave smart upselling and create urgency off, because an offer makes the survey a marketing call (calling rules).
- Frequency cap
- Caps count per agent, so this one doesn't see calls from your upsell or review agents. Set it by quarter or year so a repeat customer is surveyed once in that span; the account-wide limit of 3 outbound calls in 24 hours is only a backstop and doesn't space calls across days.
- Escalation path
- Give the survey agent the same escalation path as your support agent, so a complaint heard on a survey lands in the queue your team already works: a Loop Voice ticket by default, plus Gorgias, Zendesk, a support inbox by email or SMS.
- Recording retention
- Recordings are kept for 7 days by default, and retention can be extended to 90. A weekly read fits the default; if you review survey calls monthly, extend retention first, or the first weeks' recordings will be gone by the time you sit down with them.
What to ask
Keep it to two or three questions and let the customer talk. A useful set: How did the order arrive? How is it working out? What would you change? Have the agent ask a follow-up when an answer is short or vague, which is the one thing a form can't do.
Skip questions your data already answers. You know when the parcel was delivered; ask whether it arrived in good shape. And leave out the 0 to 10 score unless you already track it elsewhere and need the same scale.
One of the recorded check-ins shows the open question at work: "Anything you want your next coconut yak chew shipment to be better at, longer lasting, a different size, or are they perfect as is?" The customer: "Perfect as is."
Turn problems into tickets
The point of asking is to fix what you hear. When a customer reports a problem, the agent captures what happened and creates a ticket with the AI summary, the order and the recording, in Loop Voice and in Gorgias, Zendesk or Redo. On the other recorded call, a customer reported the wrong chews; the agent confirmed the details, opened a ticket and arranged a replacement to the same address.
Decide who follows up, and how fast, before you switch the flow on. A complaint collected on a Tuesday and answered the next week tells the customer the call was a formality.
Read the tickets as a list, not one by one. If three customers in a week mention the same packaging problem, that's a fix for your warehouse, not three apologies.
Surveys and review requests are different calls
Asking for a review on a survey call turns feedback into a favor, and using the answers to choose who gets asked is review gating, which Google's review policy doesn't allow. Run review requests as their own flow instead: Loop Voice's review request template calls 3 to 5 days after delivery, and it should ask every customer the same way and offer nothing for the review.
Keep sales out of it too. Treat any call that includes an offer as a marketing call (calling rules); the post-purchase upsell call is a separate flow with its own timing.
Built for the questions between intent and action
Open answers
The customer's own words, in the transcript and the call summary.
Follow-up questions
A short answer gets a follow-up question when you write that into the agent's instructions.
Problem routing
Complaints become a ticket with the order and the recording attached.
Opt-outs
A customer who doesn't want calls goes on the do-not-contact list before the next step runs.
No survey numbers yet, so measure your own
Loop Voice's September 2026 benchmark doesn't track surveys, so there are no survey response rates to publish. Track these on your flow:
- Answer rate
- Customers who picked up, out of everyone called. Across all shoppers called in the September benchmark, 15.9% picked up at least once.
- Completed surveys
- Answered calls where the customer got through every question.
- Problems found
- Tickets created per 100 answered calls, and what they were about.
- Opt-outs
- Customers who asked not to be called again.
Source: Loop Voice Ecommerce Benchmark
Works with
Common questions
When should a survey call go out?
A few days after delivery, once the product has been used. Loop Voice's review request template calls 3 to 5 days after delivery, and a survey fits the same window. Build the flow on order fulfilled and add a wait step that covers shipping time.
How long should a survey call be?
Two or three questions. Anything longer belongs in an email the customer can answer when it suits them.
Can unhappy customers go straight to support?
Yes. When a customer reports a problem, the agent creates a ticket with the AI summary, the order and the recording attached, in Loop Voice and in Gorgias, Zendesk or Redo.
Where do the survey answers go?
Every call keeps its transcript, a short summary and QA notes on objections, pain points and next steps. You can filter, search and export the call log to CSV.
Can the survey call ask for a review?
Keep it separate. Asking for a favor on a feedback call changes the answers, and inviting only the customers who sounded happy is review gating: the FTC says gating can violate the FTC Act, and its 2024 rule on consumer reviews bans incentives tied to a review's sentiment. Send every customer the same ask through a separate review requests flow, offer nothing in return, and confirm the details with counsel.
Do survey calls need consent?
Treat them like any call made with an AI voice, and treat any call that includes an offer as a marketing call. Our guide to calling rules explains which consent each kind of call needs.
Read more before you set it up
- Post-Purchase Upsell Calls: When They Pay Off, When They AnnoyWhen to sell after delivery, and when to listen.Read
- How to Set Up an AI Phone Agent for 24/7 Store SupportEscalation rules and helpdesk routing for tickets.Read
- Real Abandoned Cart Recovery Calls: What Actually WorkedBoth post-purchase check-ins, transcribed.Read
- What Is an AI Voice Agent for Ecommerce? How It WorksHow the agent hears, decides and speaks.Read
Keep the conversation moving
Ready to launch post-purchase survey?
Setup takes under 10 minutes. No code, no headcount.