This is a setup guide, not an argument. If you want the argument about whether alerts are better than a dashboard — including the cases where they clearly aren't — that's a separate piece.
What follows is the wiring. It takes about ten minutes, most of which is deciding whose phone rings rather than clicking anything.
Minutes 0–2: connect the sources
Google first, because for most independents it's the overwhelming majority of volume. Then TripAdvisor if you're anywhere near tourists, and your booking platform if it collects feedback separately. Facebook last and possibly not at all.
Pick the locations you actually manage. A group operator who routes all six sites to their own phone gets sixty texts a week and stops reading them by the second Thursday.
Minutes 2–4: decide whose phone
This is the part that determines whether any of it works, and it's a management decision rather than a configuration one.
Add the duty manager's mobile, not just yours. If the only number on the list is the owner's, then every alert becomes something the owner has to delegate, which means every alert waits for the owner to be free.
One group per site. Quiet hours on — an alert at 1am wakes someone who can do nothing about it and teaches them to silence the whole channel.
Minutes 4–6: routing and escalation
Four rules, and resist adding a fifth:
- One and two-star reviews go to the duty manager immediately.
- Three-star goes to the general manager during working hours only.
- Anything containing allergen, illness, injury or refund goes to both, immediately, whatever the star rating. This is the rule that earns its keep — the four-star review mentioning a child being unwell is the one you cannot afford to have sitting in a queue.
- If nobody acknowledges a one-star within fifteen minutes, it escalates.
The fifteen-minute escalation is worth more than it looks. Not because fifteen minutes matters to the guest — it doesn't — but because it converts "someone should deal with this" into "someone specifically has not dealt with this", which is the only version that produces a reply. The wider question of who replies is worth settling before you switch any of this on.
Minutes 6–8: the message templates
Keep them short enough to read on a lock screen without opening anything.
To the duty manager:
New 1★ at {Location}: "{Snippet}". {Link}. Reply TAKE to own it.
Escalation:
No response on the 1★ at {Location} (18 min). Reply TAKE to claim.
A first private message to the guest, as a draft rather than a send:
Hi {Name}, I'm {Manager} at {Venue}. I've read your review and I'm sorry — that isn't the evening we wanted you to have. Could you email me at {address} so I can sort it out properly?
Closing the loop internally:
{ID} closed by {Owner}. What we did: {action}. Theme: {tag}.
That last one is the template everybody deletes and then wishes they hadn't, because the themes are what you brief from next week.
Minutes 8–10: test it and hand it over
Fire a test alert. Confirm everyone gets it, including the person who's off today, because their phone is the one that will be on the rota on Saturday.
Have one manager do one acknowledgement and one public reply while you watch. Not explained — done. Ten minutes of watching someone use it is worth more than any amount of training, and it's usually where you discover that the notification is going to a phone that lives in a locker during service.
The Friday summary, and what belongs in it
The alerts handle the urgent. The weekly summary handles the pattern, and it should be short enough to read standing up:
- Best of the week — one quote, with the name of whoever earned it.
- The two themes — theme, count, which shift they cluster on, one action each.
- Anything still open — with a name against it.
- Time to first response — this week against last.
That's four lines and a number. Anything longer gets skimmed. There's a longer argument for keeping the brief this short, and it applies with more force to something automated, because automated reports lose the reader faster than human ones.
What to watch for a fortnight, and when to switch it off
Three numbers, and only three:
Time to first response on one and two-star reviews. Under an hour is a reasonable target during trading. Under twenty minutes is a target for people who like targets.
Proportion answered within 24 hours. If this doesn't move, the alerts aren't the constraint — the constraint is that nobody has been told replying is their job.
Who acknowledged. The number that matters is the share acknowledged by the duty manager rather than by you. If it's still 90 per cent you, the routing worked and the delegation didn't, and no amount of configuration fixes that.
Give it two weeks. If time to first response hasn't moved, turn the alerts off rather than leaving them running as background noise, and go and look at whether replying is actually on anybody's rota. A notification nobody acts on trains the whole team to ignore notifications, which is a genuine cost, and it's worse than not having it.
One thing this setup cannot do
Alerts make you faster at responding. They do nothing about whether the underlying problem gets fixed, and speed of reply is not the same as service recovery. Six weeks of twelve-minute replies to the same cold-food complaint is a well-run reputation process attached to an unfixed kitchen problem. Alerts are the front door; the themes in the Friday summary are the actual work.