Chinese restaurants that adopted a system for alerting staff to negative reviews saw higher average ratings in an October research revision. The study suggests that a complaint can inform operational work, not just a public reply.
Researchers Guangyu Cao, Shenghao He and Ginger Zhe Jin analyzed 122 restaurants using reviews from Dianping, a Chinese review platform. In their baseline estimate, newly posted weekly reviews averaged 0.358 stars higher after adoption, compared with a pre-adoption mean of 4.265.
The working paper, first released in March 2026 and revised in October, has not been peer reviewed. The results come from one review-monitoring system and a selected group of restaurants. They do not establish that buying similar software will raise every business’s rating.
Alerts and work tickets do different jobs
The system, introduced in 2019 by an unnamed provider, collected reviews and notified restaurants about negative ones. A separate work-ticket function let staff record and manage issues raised by customers.
Of the 122 restaurants in the analysis, 104 adopted the alert function during the observation window and 18 supplied review data but did not adopt it. Among adopters, 81 also implemented the ticket-management module. The researchers compared changes over time around the staggered adoption of the system.
A notification tells a manager that something may have gone wrong. A ticket can assign the issue to someone, record what was done and allow colleagues to check whether a repeated complaint has been resolved. Neither step guarantees the complaint is accurate or that a particular remedy will work.
The authors found stronger changes among restaurants with lower initial ratings and in areas customers had already criticized. They also found fewer public managerial replies after adoption, consistent with some attention shifting toward internal work.
Better ratings are not the same as verified service quality
The researchers examined whether the improvement could be explained by manipulation of reviews. Their tests included excluding five-star reviews and comparing archived reviews with those later visible on the platform. They found no evidence that the patterns they tested accounted for the estimated improvement.
Those checks do not prove that every review was authentic or that every operational problem was fixed. Ratings reflect the experiences and decisions of people who choose to review. A different mix of customers or an unrelated change at a restaurant could also complicate interpretation of an observational comparison.
Our earlier reporting explained how the distribution behind an average star rating can contain information that a single number hides. A restaurant tracking its own performance can likewise examine recurring complaints, not just its average.
Closing the loop on feedback
A manager can begin by grouping complaints by subject, such as waiting time, order accuracy or cleanliness, and checking whether multiple customers describe the same issue. The next step is to assign a person who can investigate and record what changed.
A public response can acknowledge a customer’s experience, but it should not substitute for correcting a recurring problem. At the same time, a quiet internal fix will not necessarily resolve an individual customer’s complaint or compensate someone affected.
The study does not measure the profitability of the software or provide a universal return on investment. Before paying for a tool, a business can compare its cost and staff time with the problems it expects to address, then monitor repeat complaints and service outcomes over a period that makes sense for its volume of reviews.
