한국어
한국어
Columns

Why do social robots fail? (1/2)

Part 1. What determined the fate of social robots

By Seungbin Jeong · · Updated

What does it take for a social robot to succeed commercially?

I have kept coming back to this question since I began researching social robots. I tried to find the answer in successful products, but few successes come with verifiable sales figures and evidence of business continuity.

So I turned to the failures to see what to avoid. I compared 32 social robots documented in this site’s archive: 13 that had been discontinued, withdrawn from certain markets, or whose manufacturers had shut down, and 19 that were still on sale.

At first, I assumed the main cause of failure was expectations inflated by unnecessary features and overstated marketing. But some robots kept selling even after reviewers judged them to fall short. As I compared the cases, my attention turned to price and to who was buying. Robots with similar limitations took different paths depending on who paid, how much, and for what purpose.

This essay comes in two parts. Part 1 follows the cases to see what shaped each robot’s fate. Part 2 looks at what the results mean.

Robots that raised expectations

Some social robots were presented as if they could do far more than they actually could. After launch, users found a different reality.

Jibo began crowdfunding in 2014 as “the world’s first social robot for the home.” While its launch slipped by nearly two years, Amazon Echo and Google Home moved into homes first, and when Jibo finally shipped in 2017, the comparison was inevitable. Reviewers liked its personality, but it could not escape the criticism that it did less than a smart speaker.

Pepper arrived with SoftBank chairman Masayoshi Son introducing it as “the world’s first personal robot that reads emotions.” But after a one-week trial at a grocery store in Edinburgh, the owner’s daughter dismissed it as “just a walking sign.”

Sony’s Xperia Hello! recognized the faces of up to ten family members and spoke to them first. It also handled messaging, video calls, photos around the home and TV control. Sony marketed it as “a member of the family,” but a review from a Japanese tech outlet on launch day concluded that it could not recommend it as a general-purpose device.

Robots that set different expectations

Among the robots that have stayed on the market for a relatively long time, some never promised users much in the first place.

When GROOVE X unveiled LOVOT, it said plainly that it was not a convenient robot that would do things for people. Panasonic deliberately left convenience features out of NICOBO so that users would not expect too much of it. Qoobo has no face and no legs; it responds with a single tail.

Instead of the ability to perform tasks, these products offered emotional connection and responsiveness as their core value. From the start, users were asked to expect a different role.

Experiments also suggest that expectation setting affects how users evaluate a robot. In 2010, before showing participants Pleo and AIBO, Paepcke and Takayama told one group that the robot was highly capable of interacting with people and the other that it was not very impressive. After using the robot themselves, the group that came in with high expectations rated its capabilities lower than the other group did. It was a small study of 24 people, but the researchers recommended introducing personal robots in a way that keeps expectations from being inflated.

Based on these findings, one could hypothesize that for a social robot to survive in the market over the long term, setting users’ expectations at an appropriate level works better than adding features or promising a lot.

Robots that fell short but survived

Not every robot followed the same path, of course. Some kept their businesses going despite the gap between the features their makers claimed and the actual user experience. What set them apart?

To find out, I checked the core features and benefits claimed at launch against independent reviews and research.

In a 2025 review, the children’s AI robot Miko was described as closer to “a small tablet dressed as a robot,” and the reviewer added that without the paid subscription it would be hard to recommend to most families. For the Chinese AI doll BubblePal, parents told MIT Technology Review that its answers ran long, it struggled to understand children’s speech and their kids quickly lost interest. France’s Buddy was crowdfunded in 2015 as a $749 home robot, but backers still had not received one by 2019.

All three drew criticism for the gap between their launch claims and the actual experience. Yet Miko and BubblePal stayed on sale, and Buddy survived by pivoting from the home market to education.

Who pays, and how much

What set the three apart was price and who bought them. Miko 3 sells for $299 on the official US store, and BubblePal for around 400 yuan in China. Buddy left the home market, won an education ministry tender and was deployed in more than 1,500 schools. Some individual backers paid a lot early on, but the business later came to rest on the ministry contract.

The expensive home robots that drew similar criticism took a different path. Jibo launched at $899 and Xperia Hello! at ¥149,880, and both services were eventually shut down. Pleo launched at $349, and its original maker, Ugobe, went bankrupt about a year and a half later. (Another company later bought the rights and sold it again, so the product itself did not disappear from the market entirely.)

Among the cases in the current archive, I had trouble finding a high-priced home robot that was criticized for the gap between expectation and actual experience and still sold for years under its original business model.

Price can also shape the value users expect from a robot. Someone who paid $899 for Jibo likely expected a higher level of performance and experience than they would from a cheap toy or a small gadget. If the actual experience falls short, the disappointment can be greater. And as that dissatisfaction spreads through reviews and word of mouth, it can sway potential buyers.

When the buyer changes, so do the criteria for judging a robot. Individuals may care most about the enjoyment of living with a robot or everyday convenience, while schools and businesses weigh its usefulness and operational results against the reason they adopted it. Even if a robot does not deliver all the features expected of it, an institution may keep using it if it is valuable enough for a specific task.

Pepper illustrates this difference. Among companies that deployed it for in-store customer service, only 15% said they planned to renew as their first three-year contracts were ending. Meanwhile, it stayed in use in Japanese schools and labs, appearing in classes at about 1,300 schools by March 2022. A successor, Pepper+, later launched in Japan. The same robot can be judged differently depending on its purpose and setting.

Institutional buyers do not guarantee a business’s survival, of course. Beyond the purchase price, they have to account for the effort of managing and running the robot. In a school, the time teachers spend preparing lessons and looking after the robot is an important factor too. Beyond price and the type of buyer, many other factors, such as operating costs and competition, affect whether a business lasts. And the fact that a product is still on sale or in use in certain settings does not by itself prove commercial success.

Still, a trend stands out in this comparison. Even robots judged to fall short of expectations sometimes stayed in business or in use when they sold at relatively low prices or when their buyers and settings changed. This suggests that understanding the sustainability of social robots means looking not only at features and user experience, but also at who buys a robot, for what purpose, and how much they pay for it.

Part 2 reframes these findings as a relationship between expectations and cost, and looks at why that balance is so easy to break for social robots.

Summary

  • The failure of social robots cannot be explained by a lack of features alone; the gap between actual performance and user expectations is a key problem.
  • Managing expectations or defining a clear role helps narrow that gap, but it does not guarantee that a business will last.
  • Miko and BubblePal kept selling despite negative reviews, and Buddy shifted its business from the home market to education.
  • These cases suggest that a robot’s sustainability depends not only on its features and user experience, but also on its price, who buys it and for what purpose.
  • The key is balancing the value users expect, the value actually delivered and the cost they pay.

Research cited