A chain with dozens of stores collects negative reviews every single day; the hard part is telling ordinary noise from "the store someone needs to call this morning". Critical review alerts make that distinction for you: when a metric crosses a threshold you set, an alert opens and the responsible person gets an e-mail. This post walks through how an alert rule is built in BranchVoice and how a regional manager runs their day with it.
Why critical review alerts need rules, not reading
Opening the review list every morning works for a few weeks, then quietly stops. Attention runs out, the store count grows, and the real signal disappears among hundreds of routine reviews. A rule-based alert flips the model: you define "what should I care about" once, and the system tests that definition against every store after every sync.
The second benefit is consistency. When the regional managers for Marmara and the Aegean work with the same thresholds, the number of alerts head office sees is comparable across regions, and the "our region just reads reviews more carefully" debate ends.
The anatomy of an alert rule
In BranchVoice a rule answers four questions: which metric, which threshold, which scope, and who gets told.
Metrics
Star-based metrics are computed from sources that carry a rating (Google Maps, Yandex Maps):
- Review count, 1-star count, negative count (2 stars and below)
- Negative rate and positive rate (percentages)
- Average rating (over the reviews in the window) and Google store rating (the live rating on the listing)
- Unanswered negative age: how many hours the oldest unanswered negative review of the last 30 days has been waiting
Metrics from the AI analysis are available too: experience score, negative content rate, high-risk and critical review counts, unanswered urgent reviews, churn-intent rate. You can also write rules on topics from your own analysis groups; for example, the "long queue" option of a "checkout problem" topic being chosen five times in seven days.
Threshold and window
Each condition has an operator (greater, less, equal), a value and a time window: 6 hours, 24 hours, 7 days or 30 days. A rule holds at most two conditions, joined with "and" or "or". The most useful combination in practice pairs a rating condition with a volume condition: "average rating fell below 3.0 in 24 hours and at least 5 reviews came in". Without the volume condition, a single 1-star review at a quiet store can sink the average and raise a pointless alarm.
Scope and severity
A rule applies to the whole brand, to the regions you pick, or to individual stores. Severity has three levels: critical, warning and info. It drives filtering on the alert list and the colour of the e-mail header; "info" rules are typically used for positive thresholds (positive rate climbed, let the regional manager know).
Notifications and repeat control
When an alert opens, everyone in the company gets an in-app notification (the bell icon); if the rule lists recipients, an e-mail goes out at the same time. The e-mail does more than say "threshold crossed": it includes the metric-versus-threshold table, a comparison with the previous period, how many negative reviews are still unanswered and a few quotes from the reviews that triggered it. A regional manager can make a first assessment without opening the app.
Receiving a fresh e-mail every hour for the same store would devalue the alert, so two brakes are built in:
- Cooldown: after an alert is resolved, the same rule will not reopen for the same store for the period you set (24 hours by default).
- Occurrence counter: if an open alert's condition is met again on later evaluations, no new alert is created; the existing one shows a "repeated N times" badge with updated numbers. If the alert is still untouched, a single reminder is sent on the third repeat.
Once a rule is saved, "send test e-mail" shows recipients exactly what the template will look like. A rule can be paused, edited or deleted at any time, and a saved change is re-evaluated straight away, so you see what the new threshold produces within the minute.
The life of an alert
Every alert opens as "new"; it becomes "investigating" when a regional manager picks it up and "resolved" once the store has been spoken to and action taken. Every transition and every note stays on the timeline, so when the same problem resurfaces at the same store three months later, the earlier notes are right there.
Alert rules work alongside review anomaly detection, which catches a store drifting away from its own history: rules watch the thresholds you defined, anomaly detection watches the deviations you did not think to define. Both are listed on the same Alerts page, in separate tabs.
A starter rule set for chains
| Goal | Scope | Condition | Severity |
|---|---|---|---|
| Sudden rating collapse | Whole brand | Average rating ≤ 3.0 in 24 hours and review count ≥ 5 | Critical |
| Response discipline | Region | Unanswered negative age ≥ 48 hours | Warning |
| Hygiene and safety | Whole brand | Critical review count ≥ 1 in 7 days | Critical |
| Newly opened stores | Selected stores | Negative rate ≥ 30% in 7 days | Warning |
| Positive momentum | Region | Positive rate ≥ 85% in 30 days | Info |
Treat this as a template; the right thresholds depend on the sector and on how many reviews each store receives. For background on how Google reviews are collected and attached to stores, see Google Maps review management.
Next steps
Alert rules turn review data from a report somebody looks at into a phone call somebody makes. Once your stores are connected, the first rule takes a few minutes to set up and the thresholds settle within a week. To design a rule set that fits your store structure, request a demo.