A Loop Voice agent calls about 10 minutes after checkout drop-off, helps with whatever stopped the order and texts the checkout link. Calls nobody answers use no credits.
Modern Hearing: $300,000+ in attributed revenue from 8,000+ calls, with 11% of calls leading to a sale.
The short answer
Cart recovery calls work when the shopper meant to buy and something got in the way: a card that kept failing, an add-on they didn't want, a question about the product. In Loop Voice's September 2026 benchmark, 16.5% of the 14,036 shoppers called after abandoning checkout picked up at least once, and 13.7% of everyone called ordered within 5 days, through any channel. Timing lined up with one of the widest gaps: shoppers first called 5 to 15 minutes after abandoning ordered at 22.9%, against 4.1% when the first call came 15 to 60 minutes later.
Thirteen recorded calls from stores on Loop Voice, each cut to the moment that decided it. We picked calls that led to an order; most calls never reach a person, and in September 16.5% of shoppers called picked up at least once.
Full transcripts are in the call library.
A recovery email arrives hours later and can only say "come back." If the card was declined, the code wouldn't apply or an add-on looked mandatory, the shopper needs someone to say which button to press. In 9 of the 12 cart calls we transcribed, the shopper had already decided to buy and was stuck on the checkout itself.
Install the Shopify app, or connect WooCommerce, BigCommerce, Magento, Shopline or Checkout Champ. Abandoned checkouts, carts and orders reach Loop Voice through your platform, and nothing is installed on your storefront.
Call from a dedicated number, or from the Super Number on the Scale plan, which shows a number with the shopper's area code. In September, dedicated numbers were answered on 7.3% of calls and Super Number calls on 5.3%.
The template waits 10 minutes, calls, texts the checkout link, waits, calls again with a discount and sends a last note. It stops the moment the shopper orders, says stop or has a negative call.
Pick the Shopify discount codes the agent may use and when, the calling window in each shopper's time zone, and the most calls any customer can get in a week, month, quarter, six months or year.
Run the agent against the simulator's angry, rushed and skeptical shoppers, then have it call your own phone from the test console. Test calls stay out of your analytics.
When shoppers on September's calls gave a reason for leaving, a third were just browsing (33.4%) and a quarter were waiting for a discount or checking the price (25.6%). Another 17% gave a reason the call could fix on the spot: a site or checkout error (11.7%) or a product question (5.5%). That's 824 conversations across the stores in the September 2026 benchmark.
Browsers are a big reason answered calls end without an order, and why every call needs an easy way out. The fixable 17% are where a call earns its cost, because a call can fix their problem on the spot.
In the recorded calls above, the fixes were small: retrying a declined card with the agent on the line, declining a protection add-on that looked mandatory, getting a discount code to apply. Our call-by-call breakdown quotes the lines that worked.
Timing showed one of the widest gaps in September's data. Shoppers who abandoned during calling hours and got a first call 5 to 15 minutes later ordered at 22.9% within 5 days. A first call 15 to 60 minutes later went with a 4.1% order rate, and 1 to 4 hours later with 4.7%. Checkouts abandoned after calling hours and called the next day did better than in-hours checkouts first called 15 minutes to 4 hours later, at 6.5%.
The benchmark's clock counts only minutes inside calling hours (8 am to 9 pm in the shopper's time zone, or a state's tighter window; see calling rules), and these are comparisons across shoppers and stores, not a controlled test. Read them as a direction: keep the default 10-minute wait unless you have a reason to change it.
Stopping matters as much. Across all calls in September, 9.5% of shoppers who picked up asked not to be called again. Every opt-out, whether "don't call me" on the call, STOP by text or a carrier block, goes on the do-not-contact list before the next step runs, and the flow also ends on an order or a negative call.
A quarter of shoppers who gave a reason were waiting for a discount or checking the price, so the agent will hear the question. The default flow keeps the code for the second call, and the agent can only use the Shopify codes you allow.
Our call script template sets one rule: offer the code only when price is the objection. In the recorded calls the code never grew: when a shopper said 10% might not be enough, the agent said it couldn't go beyond that code and offered help with his plants instead.
We don't publish conversion by discount level yet. Each level needs 200 conversations where price was the objection, and September's calls didn't reach that for any level.
Calls and texts run as one sequence. In September, a text was the credited touch on 77.8% of attributed abandoned-checkout orders and an answered call on 22.2%. A text takes the credit whenever the shopper ordered after a text with no answered call in the window, so judge the channel on both.
Each order is matched to the call or text by its Shopify order number, inside an attribution window you set. That shows what the calls touched. To see what they added beyond your emails, hold out a random share of shoppers: the holdout test guide shows how to size it.
If Klaviyo sends your emails, this timing plan keeps the call, the emails and one text from colliding, with every message stopping on the order.
Declined cards, add-ons the shopper didn't want, discount fields and pages that won't load, with the agent on the line until the order confirms.
Fit, compatibility, shipping times and returns, answered from your site, policies and product catalog.
Only the Shopify discount codes you allow, at the point in the flow you choose.
Texted during or after the call, so the shopper can finish on their own phone.
$300,000+ attributed revenue · 8,000+ calls · 11% led to a sale
“Our customers have a lot of questions before they commit to a hearing aid. Loop answers those questions live on the call and converts customers who would have just disappeared. We're recovering a strong amount revenue through Loop and learning a lot about our customers in the process.”
Jorraine
CEO, Modern Hearing
Read the story
$200,000+ attributed revenue · 1,184 calls · 7.1% led to a sale
“Loop AI Voice has been a fantastic addition to our Shopify store. Since we sell high-ticket products, every recovered order can have a significant impact on our business, which is why this app has been especially valuable for us. The results have exceeded our expectations, and we've recovered orders that may have otherwise been lost. We're also extremely impressed with the customer support team.”
Dylan B.
Warrior Will Power
Read the story
$30,000+ attributed revenue · 2,789 calls · 8% led to a sale
“Loop's voice agents are recovering abandoned carts that our email flows never could have converted. The sales we're getting back are real and meaningful. The team is fantastic, executes changes fast, and has been a great partner. Highly recommend.”
Daniel Castle
CEO, Castle Flexx
Read the story
$45,000+ attributed revenue · 8,000+ calls · 5% led to a sale
“Loop has been very helpful to convert abandoned cart and remind customer when we restock products, helps them find the right size, and even suggests pieces that go with what they were looking at to increase the AOV.”
Kapeel B.
Founder, CityMan USA
Read the storyLoop Voice publishes a monthly benchmark from stores live for at least 30 days, with the sample size and 95% interval behind every rate. These are the September 2026 results for abandoned checkouts.
16.5%
of shoppers called picked up at least once
14,036 shoppers called
15.1%
of shoppers who picked up ordered with the call as the last touch, within 5 days
95% interval: 13.6% to 16.7%
+5.7 pts
estimated lift in 5-day orders over enrolled shoppers the flows never reached
95% interval: 4.4 to 6.9 points
$234
average converted abandoned-checkout order
Against $210 across all orders in the same stores
The lift is an estimate, not a randomized holdout: no store ran one in September, so contacted shoppers are compared store by store with enrolled shoppers the flows never reached. The abandoned checkout lift comes from 5 of the benchmark's 7 stores, so results on your store will differ.
Works with
The default flow waits 10 minutes, and you can change the delay. In September's benchmark, shoppers first called 5 to 15 minutes after abandoning ordered at 22.9% within 5 days, against 4.1% when the first call came 15 to 60 minutes later. Calls only go out inside the calling window you set, in each shopper's own time zone.
It depends on the consent you collected. In the US, treat a cart recovery call as a marketing call: the FCC treats AI-generated voices as artificial voices, so you need prior express written consent that covers calls made with an AI voice, calls only inside the shopper's local calling hours, an opt-out and Do Not Call checks. Our guide to calling rules covers the wording and the checks.
Unanswered calls use no credits. The default flow texts the checkout link, waits, calls once more with a discount and sends a last note, then stops. You can also cap calls per customer by week, month, quarter, six months or year.
With the guardrail for clear identity as an AI turned on, the agent says it's an AI assistant whenever a caller asks. Put your store's name in the opening line too, so every call starts by naming the business.
By Shopify order number: an order counts when it matches a call or text inside the attribution window you set. That shows which orders followed a call. To see how many the calls added beyond your emails and texts, run a holdout; the holdout guide explains how.
Plans start at $119 a month for 120 credits, about 80 answered conversations. An answered conversation uses about 1.5 credits, and calls nobody answers use none. Pricing has the larger plans.
Yes. The simulator runs your agent against AI-played customers, such as an angry shopper or someone who asks if it's a robot, and scores each conversation without dialing anyone. The test console then calls your own phone.
Setup takes under 10 minutes. No code, no headcount.
