A review written on Google belongs to a specific store by definition. A complaint on Şikayetvar that says "the cashier at your Antalya store shouted at me" is posted to the brand's general page, and the branch it concerns is hidden inside the text. In a multi-store chain, failing to route those reviews to the right regional manager means the most serious complaints end up in a "general" pool that nobody owns. This post explains how BranchVoice handles matching reviews to stores, which rules the AI applies, and what happens when it is not sure.
Two kinds of review source: store-level and brand-level
Platforms differ in where a review is posted:
- Store-level sources: Google Maps and Yandex Maps. Each store has its own business listing, and a review belongs to the listing it was written on. The match is deterministic and the AI is not involved.
- Brand-level sources: Şikayetvar (Turkey's largest consumer complaint site), Ekşi Sözlük (a Turkish collaborative dictionary where brands are discussed at length) and X. The review is written about the brand; if there is any store information, it is in the text ("the store in Cevahir mall", "the Bursa Nilüfer branch").
When a brand-level review enters BranchVoice it is first linked to a virtual store called "General". That means the review has not yet been assigned to a branch. During analysis, the AI looks for a branch clue in the text and, if it finds one, moves the review to the matching physical store.
How the AI matches reviews to stores
Matching happens in the same step as AI review classification; it is not a separate process. The flow is:
- Candidate extraction: The review text is compared with your store list. Up to ten physical candidates are selected using city, district, mall name, address fragments and the aliases you have defined for each store. The virtual "General" and "E-commerce" stores are always among the options.
- Decision: The AI considers the review and the candidate list together and either picks a single store or answers "none". When a store is picked, the confidence of the decision is measured as well; if it falls below the defined threshold, no link is made and the review stays in "General".
- Mall conflict check: If the review names a shopping mall, the selected store's name, address or aliases must contain that mall. Otherwise the link is rejected. This rule removes the most common cause of mis-matches in chains with several stores in the same city.
- Channel decision: In the same step, the AI decides whether a brand-level review describes a physical store or an e-commerce experience. "My parcel has not arrived in three weeks" goes to the E-commerce virtual store, not to any branch.
Every change of link is recorded as an event, so you can later see when a review was linked to which store and by which method (source listing, AI or manual).
Critical reviews with no clear match: the Branch Matching queue
The most important exception to the matching rules concerns critical reviews. When a review with a high risk level (hygiene, safety, misleading pricing) cannot be linked to any branch, it is automatically placed in the Branch Matching queue under Alerts. The queue shows the review text next to the candidate stores; someone on the operations team picks the right branch and the review is linked manually.
This queue is a deliberate design choice. Rather than sending a critical complaint to the wrong manager on an uncertain guess, it costs one person thirty seconds. Low-risk reviews that name no branch simply remain in "General" and are counted in brand-level reports.
How to improve matching accuracy
The quality of the AI's decisions depends on the quality of your store data. Three simple steps make a difference:
- Define aliases. Customers rarely use a store's official name; they write "the Forum store" or "the Kızılay branch". Adding those phrases to the alias field on the store card means candidate extraction scans for them too.
- Complete address and mall details. The mall conflict check relies on the mall name appearing in the store's name or address. Filling in those fields during store list import is easier than fixing them later.
- Review the Branch Matching queue weekly. The queue does not pile up, because critical reviews are rare. Checked regularly, it ensures every brand-level complaint ends up at the right branch.
What it means for a regional manager
The Marmara regional manager of a cosmetics chain used to hear about Şikayetvar complaints in a weekly digest from head office. With matching in place, a complaint that begins "at your store in Kocaeli" appears in that store's review feed and on the regional dashboard the day it is written, and reaches the manager by email through critical review alerts if it is critical. In the store comparison view, Şikayetvar and Ekşi Sözlük reviews now sit in the same branch breakdown as Google reviews.
| Situation | What BranchVoice does |
|---|---|
| Google / Yandex review | Linked directly to the store from the source listing |
| Brand review with a branch clue | AI picks from candidate stores and links if confidence passes the threshold |
| Brand review with no clue | Stays in the "General" virtual store |
| Describes an e-commerce experience | Goes to the "E-commerce" virtual store |
| Critical and unlinked | Waits for a manual choice in the Branch Matching queue |
Next steps
Matching reviews to stores is the step that turns brand-level platforms into something you can manage branch by branch; without it, Şikayetvar and Ekşi Sözlük remain brand reputation sources only. Once your store list and aliases are ready, historical reviews are matched with the same rules. To see it working with your own stores, request a demo.