What a 90-Day Guest Messaging Pilot Should Prove

A pilot is not a trial of the software. It is a trial of whether your property can run it. Most pilots that fail do so for reasons that were decided before the first guest message.

Set the baseline first

Nothing in a pilot means anything without a number from before.

For two weeks before go-live, count at the desk: how many inbound guest questions, on which channels, how long each takes, and which question is asked most. Note the response time on a typical afternoon and on a busy Saturday.

That fortnight of counting is the whole evaluation. Without it you will end ninety days with a feeling, and feelings do not survive a budget conversation.

Measure four things, not satisfaction

Overall guest satisfaction moves for too many reasons - weather, a renovation, one bad review - to prove anything about a messaging tool in ninety days. Measure things the tool directly touches.

1. Share of conversations resolved without a person. The headline number. Across the vGuest portfolio it runs around 70%, with the best-performing property at 98.7%. A new deployment should be climbing toward the portfolio figure by month three.

2. Volume of the most repetitive question still reaching staff. Take the top question from your baseline count. Count it again in month three. This is the single most persuasive before-and-after, because it is one number and everyone at the desk knows the question.

3. Response time on everything else. When the repetitive questions stop consuming the queue, the non-repetitive ones get answered faster. This is often the larger operational gain and it is invisible unless measured.

4. The escalation path, end to end. For every request the assistant handed over: where did the notification land, who saw it, how long until a person responded. If this number is bad, the tool is not failing - the routing is.

The three decisions that decide the outcome

Before day one, not after.

Who owns the knowledge base. A named role at the property, close to guests, with authority to change answers without a committee. An assistant nobody owns is out of date by week six, and the pilot then measures stale information rather than the product.

Where escalations go. A channel staff actually watch - for most teams that is a WhatsApp group - with a fallback if nobody responds within a set time. A pilot that routes escalations to a dashboard nobody opens mid-shift will conclude the tool does not work, when the tool worked and the notification did not.

What it must not answer. Complaints, accessibility requests, anything with legal or financial weight. Decide the list, and check the assistant honours it.

Month by month

Month one: settling. The knowledge base gets corrected daily as real questions expose gaps. Resolution rate is low and rising. Do not measure yet; read escalations and fix answers.

Month two: real operation. The rate should be climbing. This is when the repetitive question count is worth taking, and when the escalation path shows whether it is routed correctly.

Month three: measure. Take all four numbers against the baseline. Include a busy period if the calendar allows, because that is where the value concentrates - the share of basic questions rises exactly when staff have least capacity to absorb it.

What a good result looks like

The top repetitive question reaching staff drops by a large majority. Resolution share is in the region of the portfolio number or clearly heading there. Non-repetitive response time is faster than baseline. And the escalation log shows requests reaching a person within minutes, not the next morning.

If three of four are there, the product works at your property and the fourth is fixable. If none are, the likeliest causes in order are: no baseline, no owner, wrong escalation channel. All three are the property's to fix, and all three are cheap.

What to insist on from the vendor

A real pilot at real volume with your own guests, not a demo environment. Weekly access to the escalation log and the unanswered-question list. A named contact who reviews month one with you. And a clear statement of what the price will be after the pilot at the volume you actually saw.

A vendor confident in the product agrees to all of this. It is, after all, the only way they get the case study.