Your review score tells you how guests felt, on average, weeks after they left. It doesn't tell you which stays went wrong, why, or what to fix first. And it only counts the guests who chose to review. Many with a poor stay don't. They just don't book again.
Most of what you need is already in your inbox. Guests tell you when the wifi drops, the shower's cold or the flat wasn't clean, usually during the stay, in their own words. A guest experience index turns that into a number you can manage.
What the index measures
The index is simple: the share of completed stays where nothing went wrong.
For each stay, read the guest's messages and their review, if there is one. Note anything they raised, and how serious it was. A stay with no issues counts towards the index. A stay with any issue doesn't.
Here's an illustration for one month:
- 412 completed stays
- 354 went smoothly, with nothing raised
- 58 raised something, of which 9 were serious enough to spoil the stay
That's an index of 86: 86% of stays had no complaint at all. On its own, that's a useful health check. The value is in the breakdown.
Service and property, scored separately
Split every issue into one of two groups, because each is fixed by different people.
| Service: what your team did | Stays affected | Property: what guests found | Stays affected |
|---|---|---|---|
| Check-in and access | 6 | Facilities | 11 |
| Communication | 3 | Comfort | 8 |
| Billing and admin | 5 | Cleanliness | 34 |
Illustrative figures.
Service issues point at your processes: door codes that didn't work, slow replies, a confusing deposit. Property issues point at the homes and the people who look after them: cleaning, broken appliances, a lumpy mattress.
Mixed into a single review score, a cleanliness problem and a communication problem look the same. Split out, they go to different people with different fixes.
Find the one fix that saves the most
The last step is the one that makes the index useful. Look at where issues cluster.
In the example, cleanliness cost 34 stays. Look closer and 13 of those were at one property. That isn't a portfolio-wide cleaning problem. It's one cleaner to talk to, or one property with a layout that makes it hard to turn over in time. One conversation could recover a third of your cleanliness complaints.
That's the habit to build: each month, name the single change that would have saved the most stays, and do it.
How to run it by hand
For a small portfolio, a spreadsheet works.
- List last month's completed stays from your PMS, one row each.
- Read each stay's conversation and review. Note any issue the guest raised, even if it was resolved quickly.
- Tag each issue with one category (check-in and access, communication, billing and admin, facilities, comfort, cleanliness) and a severity: raised or spoiled the stay.
- Work out the index: stays with no issues ÷ all completed stays.
- Sort by category, then by property. Look for clusters.
Count resolved issues too. A wifi problem fixed in an hour was still a problem the guest had to report, and a pattern of them at one building is a router to replace before it reaches a review.
Where AI helps
By hand, this takes a few minutes a stay. At 400 stays a month that's a couple of days' work, which is why most operators never do it, or sample a handful.
AI makes it practical to score every stay. An agent can read each conversation and review, recognise when a guest raised something (including the polite ones, like "no worries, we managed without hot water"), classify it, judge severity and roll it up by property. Then it can point at the cluster that matters.
A few things to check if a tool offers this:
- Does it read the messages, or only the reviews? Reviews are a fraction of the signal.
- Does it separate service from property? If not, the fix is hard to route.
- Can you click through to the stays behind a number? You should be able to read the actual conversations.
- Does it name an action? A dashboard of percentages is less useful than "13 cleanliness complaints at one property".
- Are the categories yours? If you care about noise or parking, you should be able to track them.
What to do with it
Use the index alongside your automation rate, not instead of it. Automation tells you how much work comes off your team. The index tells you whether guests are getting a better stay. If automation climbs and the index falls, something's being answered that shouldn't be, or answered coldly. We cover how that happens in why AI guest messaging goes wrong.
For how this fits with the rest of an agentic operation, see agentic AI for short-term rental operations.