About this case: The client is a smart coffee machine brand selling in North America. We haven’t named it.
Key Takeaways
- The brand’s return rate fell from 28.5% to 13.5%.
- Many of the returned machines weren’t broken. A shipping seal was still in the water tank, or there was air in the line on first use.
- We put a chat window inside the brand’s app, so buyers could ask for help on the screen that showed the error.
- The human agents reported the most common false faults to the brand, and the brand added an illustrated card to the top of the box.
A new machine makes a loud noise and gives no water. The app shows a red error. The buyer looks for a way to ask someone and finds an email address.
On this account, a common cause was a shipping seal that was still in the base of the water tank. The machine was fine. One such return costs the seller roughly US$80–120 in reverse freight and disposal fees, and the one-star review stays on the listing.
Two gaps between the error and the answer
The brand had an English manual and an email support address. Neither helped at the moment the buyer needed it.
The first gap was the entry point. The error appeared in the app, and the app had no way to reach a person. The buyer had to leave the app, write an email and wait an average of 4.5 hours for a reply. Starting a marketplace return takes less time than that.
The second gap was the language of the fix. The manual had been translated directly into English and read like an engineering formula. Buyers skipped it. A fault that takes seconds to clear was being reported as “the hardware is dead.”
Consumer electronics as a whole has this problem. According to TechSee’s research on no-fault-found returns, 68% of consumer electronics returns are no trouble found, and more than 60% of those trace back to something that went wrong during first-time setup (TechSee, 2026).
What we changed
Chat inside the brand's app
We embedded the chat window from HeroDash, our support platform, in the brand’s app. A buyer taps the help icon and the conversation opens in the human agent’s workspace. Buyers can send a screenshot or a short video of the fault.Device details beside the conversation
The workspace sidebar shows the buyer’s account, region, purchase channel, device model, serial number and warranty period. The human agent doesn’t have to ask which machine the buyer owns.Fixes rewritten in everyday English
The human agents studied the manual and then rewrote each troubleshooting step as a short sentence a buyer understands on the first read.Role-play before the first live chat
New human agents went through role-play scored on five things: troubleshooting knowledge, empathy with an anxious buyer, local service manners, privacy and information-security compliance, and avoiding arguments.One chat, start to finish
Our account team describes a US buyer named Chris who hit an error while trying the machine and tapped the help icon in the app.
A human agent in one of our centers in the Americas picked up the chat. The sidebar showed that Chris was in the United States, had bought through Amazon, and which model and serial number he owned. The first message went straight to the fault:
“Hi Chris! I see your machine is reporting a steam air lock—don’t worry at all, the heating block is 100% fine, it’s just a tiny air bubble trapped from factory water testing!”
Chris followed the instructions and opened the valve. Five seconds later the air cleared and the noise stopped. He wrote back: “WOW, it works! Thanks man!”
The human agent then asked for an in-app rating, and Chris gave five stars.
The support team’s notes changed the packaging
After handling many chats like this one, the human agents listed the false faults they saw most: the seal left in the water tank, and air in the line on first use. They wrote short tips for each and sent them to the brand.
The brand and the support team turned the main unboxing checks into an illustrated card. It now sits on the top layer of the box, where the buyer sees it before the machine. A buyer who reads it can clear the fault without contacting support.
We describe this pattern in our guide to customer inquiry volume: the cheapest support contact is the one that a change to the product or the box makes unnecessary.
Results
Smart coffee machine brand: before and after
The response-time row compares two channels. Before the change, buyers emailed and waited. After it, they chatted in the app. The channel changed along with the speed.
The drop in returns has more than one cause. The in-app chat, the rewritten fixes and the card in the box all arrived during the same engagement, so the drop can’t be credited to any one of them.
When this setup is more than you need
In-app chat makes sense when the product has an app and the faults show up there. If your product has no companion app, a QR code on the quick-start card that opens a chat does a similar job for less integration work.
If your returns come from real hardware defects, faster support will make those customers less angry and they’ll still send the product back. Check what your inspection team finds in returned units first. The share that arrives in working order is the part support can affect.
Three questions to ask about your own returns
- Of the units that came back last quarter, how many had nothing wrong with them?
- When your app shows an error, how many taps does it take to reach a person?
- Does anyone on your support team send the product team a list of the faults that weren’t faults?
Callnovo.ai ran this account with human agents trained on the product and with HeroDash embedded in the brand’s app. For how we scope product support for hardware brands, see outsourced IT support services. Our managed support teams and published pricing are on the site.
Sources
- TechSee, no-fault-found returns research
- Callnovo, smart coffee machine support case (WeChat, in Chinese), September 11, 2026
Performance metrics reflect results from a Callnovo.ai partnership. Client identity withheld.