A generic sentiment model tells you a review is negative, but a bank, a hotel or an online shop wants to know something more specific: is the problem the card fee, the room cleanliness or a late delivery? Industry-specific review classifiers are AI tools that read a review through the topic set of that particular sector. This post walks through the nine industry classifiers among BranchVoice's free AI tools, what they flag, and how the same approach runs automatically inside the platform.
Why a generic model is not enough
Every sector has its own complaint vocabulary. "I waited" can mean slow service in a restaurant, the call centre at a mobile operator, or a loan application at a bank. Generic buckets such as product quality, price and staff place every review somewhere, but they do not tell a team what to do next. What a customer experience director needs is a split fine enough to send the review straight to the unit responsible: the courier, the product page, or seller communication?
That is why the industry classifiers ship with a predefined label set per sector, written in that sector's operational language. The AI decides on each label independently, so a single review can be both "late delivery" and "size and fit".
Nine sectors, nine label sets
The tools are grouped under "Industry classifiers" on the AI tools page. Each one works with 9 to 13 labels chosen for its sector:
| Sector | Key labels |
|---|---|
| E-commerce | Late delivery, damaged packaging, wrong or missing item, not as described, size and fit, returns |
| Mobile app | Crash or freeze, login and account, payment, performance, notifications, update regression |
| Mobile game | Gameplay, difficulty balance, pay to win, in-app purchases, ads, multiplayer |
| Software | Ease of use, features, reliability, bugs, integrations, licensing, onboarding and docs |
| Online course | Instructor quality, content, course structure, level match, up to date, audio and video, certificate |
| Restaurant and hotel | Food quality, portion, service speed, staff attitude, cleanliness, ambience, room comfort |
| Banking | Mobile banking, cards, fees and charges, loans, branch and ATM, fraud and security |
| Automotive | Performance, comfort, technology, fuel efficiency, safety, maintenance and repairs, after-sales |
| Mobile operator | Coverage, internet speed, billing, tariffs, number porting, contracts and cancellation, outages |
The label sets follow how operations and product teams actually divide work. In banking, fraud and security reports are a label of their own, so unauthorised transaction complaints can go to the security team without being lost among general dissatisfaction. In the telecom set, network problems (coverage, speed, outages) are kept apart from commercial ones (billing, tariffs, contracts).
How to use a tool and what the result shows
Open your sector's tool, paste a customer review, complaint or message into the text box (up to 1,500 characters) and press "Analyse". Every page also offers sample reviews to try. The result lists the labels detected in the text with their probabilities and shows the ones that are not clearly present separately.
The tools are free and need no sign-up. To prevent abuse, the number of runs is limited; if you hit the limit, you can carry on after a short wait. They understand multilingual text, Turkish and English first. Your text is not stored; when the same text is sent again, the result is cached briefly so it comes back faster.
If the label set does not fit, the "customise labels" panel lets you define your own: up to 12 labels, each with a short name and a description of what it means. The tool then asks about your labels instead of its own. It is a practical way to test an idea before building a permanent topic set in the platform.
How the same logic runs in the platform
The free tools classify one text on demand; BranchVoice applies the same approach to your whole review stream. The platform regularly collects reviews from Google Maps, complaint sites (such as Şikayetvar, Turkey's largest consumer complaint platform), social media, app stores and marketplaces, and runs every review through analysis groups. An analysis group is the set of questions the AI answers for each review, and different channels use different groups. App reviews, for example, go through the mobile app group, which asks about issue type, what users liked and the impact of updates; see mobile app review tracking for details.
Labels you tried in the customisation panel can become permanent in the platform: you define your own analysis groups and topics, and every new review is analysed with those questions. Analysis groups and topic taxonomy explains how this works. Classification results feed dashboards by store and period, alert rules and reports; for the overall logic, read AI review classification.
What a regional manager does with the result
The value of industry labels is that they turn a review into an action. If "portion" complaints pile up at one restaurant branch, that is a kitchen standards issue; if "service speed" rises on weekend evenings, it is time to look at the shift plan. An e-commerce team groups "not as described" reviews by product to see which product page needs new photos or a better size chart. A mobile operator routes outage and coverage complaints to the network team and contract or number porting conversations to the retention team.
In short, the tool tells you what a review is about in your sector's words; the platform does it every day, for every store, and in a way that reaches the right person.
Next steps
Open your sector's classifier on the AI tools page and try a few of your own reviews. To apply this analysis automatically to reviews from all your sources and build your own topic set, request a demo.