What is real-time guest feedback, and how do you act on it?

Real-time guest feedback is anything a guest tells you while they are still on the property: a reply to a short check-in message, a complaint in a WhatsApp thread, a request that arrives for the second time. Acting on it means the information reaches someone who can still change the stay, with the authority to do so, before the guest has decided what the stay was. A post-stay survey measures the outcome. Real-time feedback is the last point at which you can alter it.

Why a post-stay survey arrives too late

The survey is not wrong. It is late. It reaches the guest after the stay it asks about, so whoever answers is describing a memory, and the details that would let you fix anything - which room, which shift, which morning - have usually gone by then.

The people who could have acted see the result weeks later, as a score in a report. That is useful for spotting a pattern across a season: a floor that keeps scoring low on noise, a breakfast that improved after a menu change. It does nothing for the guest in room 214 tonight, whose air conditioning is loud and who has not yet decided whether to mention it in a review.

Real-time feedback does not replace the survey. It sits in front of it, in the window where an answer can still change what the survey will eventually say.

Where real-time guest feedback actually comes from

Mostly not from asking. It arrives in the messages guests already send: "is the pool open", "the AC is loud", "how do I get a taxi", "we still have no towels". Each one is information about the stay, and the ones phrased as complaints or repeated requests are feedback in everything but name.

That matters for where you collect it. Across the 190+ properties vGuest runs guest messaging for, 97% of conversations arrive on WhatsApp (vGuest portfolio data, August 2026). The channel guests already have open with you is the feedback channel; a separate survey app, a QR code to a form or a kiosk by the lift asks the guest to move somewhere new to tell you something they would have told you where they were.

In practice there are three sources, and a property that only watches the first is missing most of the signal:

1. Unprompted messages. Requests, complaints and questions during the stay.
2. One prompted question. A single short message at the right moment (below).
3. Behaviour. The same request twice means the first one was not resolved, whatever the reply said. A guest who asks for the Wi-Fi password three times is telling you something about the Wi-Fi, not about their memory.

The one question worth asking, and when

Ask once, after the first night: "How was your first night? Anything we should fix today?"

The timing is the point. On arrival there is nothing yet to judge. At departure it is too late to act. After the first night is the only moment when a fix still changes the stay and the guest has not yet formed a verdict on it. A second night with the problem solved is remembered as a hotel that listened; a second night with the problem unsolved is remembered as the problem.

What not to send: a form, a satisfaction scale, or anything daily. One open question gets an answer a person can act on. "Rate your stay so far from 1 to 10" gets a number nobody can act on and, from a guest with a loud air conditioner, a 6 that tells you nothing about the air conditioner.

How do you analyse real-time guest feedback?

There are two jobs, and they need different tools.

The first is to spot the message that needs a person now. A complaint, a repeated request, wording such as "still", "again" or "nobody came". An assistant can flag that wording and hand the conversation to a member of staff with the whole thread attached, which is what vGuest does when a conversation shows signs of a problem or confusion: the reply then comes from a person, not the assistant. The hand-off from assistant to staff is where feedback becomes actionable, because the person taking over can fix the thing while the guest is still upstairs.

Be careful what you read into the flag. Wording is a prompt for a human to look, not a verdict. A terse guest is not an angry guest, a guest who writes "fine" may mean it, and a sentiment score on a four-word message is a guess. Treat the flag as "read this one now", never as "this guest is unhappy".

The second job is to count what recurs. The question that arrives every day belongs in your property information, so the assistant answers it before anyone has to. The complaint that repeats on one floor belongs with maintenance, not with the front desk that keeps apologising for it. Sentiment analysis across reviews and messages is the weekly view of the same data; the daily view is simply a list of what came in and which of it came in twice.

Acting on it: authority first, then close the loop

Information without authority is a report. Before you ask a single guest anything, decide what front-line staff may do without asking anyone: move a room, send housekeeping up, waive a breakfast charge, call maintenance directly. Write the list down. A team that has to check before fixing a cold room will stop asking guests whether their rooms are cold, because the question creates work they cannot finish.

Then close the loop in two messages, not one. The first is an acknowledgement: "Thanks, someone is on the way." The second, the one most properties skip, is the confirmation: "That should be sorted now - is it right for you?" The second message is where the guest learns that saying something changed something, and it is the only reason they will say something next time.

Every one of these should land as a task someone owns. A guest messaging platform that routes a request to housekeeping or maintenance as a job with a name on it is doing real-time feedback whether or not anyone calls it that; a shared phone that everyone can read and nobody owns is not.

Where real-time guest feedback goes wrong

• Asking too often. A question every day is a survey by another name, and guests stop answering the second one.
• Asking without the authority to fix. Worse than not asking. The guest told you, you did nothing, and now they know you knew.
• Reading silence as satisfaction. A guest who does not reply to "how was your first night?" has not said it was fine. Watch the unprompted messages instead.
• Trusting the score more than the message. Sentiment tools are useful for volume and useless for a single guest. Read the message.
• Collecting it where the team does not look. A web chat widget nobody owns, or a survey tool whose results go to head office, gathers feedback that changes nothing on the property.

What to do first

Tomorrow morning, send one message to every guest who arrived today, on the channel they booked or messaged you on: "How was your first night? Anything we should fix today?" Tonight, decide who reads the replies and what they may fix without asking. That is the whole system; everything else is scale.

The Guest Messaging topic collects everything we have written on running the channel this feedback arrives through, and the Guest Experience topic covers what to do with it once it has.