Hotel operator guide

Hotel guest feedback analysis across multiple properties

Analysing guest feedback across a portfolio means doing four things the single-property review inbox never does: classify every comment into a fixed operating taxonomy, score every property on the same scale, separate portfolio-wide patterns from property-specific ones, and attach each issue to an owner with a date to recheck it.

A single hotel can be run from its review feed. A regional director with fourteen hotels cannot, because the feed gives them fourteen unrelated narratives, four scoring conventions and no way to tell whether the Tuesday complaint about check-in queues at one property is an isolated staffing gap or the third instance of the same rostering decision.

The work below is the method, not the software. It can be done in a spreadsheet for a small estate, and most groups should prove the method that way before buying anything that automates it.

Start from a fixed issue taxonomy, not from sentiment

Sentiment scoring tells you that a guest was unhappy. It does not tell you what to fix, and it cannot be compared between properties because the mix of what guests bother to write about differs by market, rate and segment.

A fixed taxonomy does both. Every comment is classified into the same small set of operating categories across the whole estate, so a category rate at one hotel means the same thing as the category rate at another. Keep it small enough that classification is consistent — a taxonomy nobody can apply twice the same way is worse than none.

Cleanliness — room, bathroom, public areas, turndown consistency.
Arrival and front desk — queueing, check-in time, room readiness, key and billing errors.
Maintenance and room condition — HVAC, plumbing, lighting, wear, lifts.
Noise and disturbance — corridor, adjacent rooms, external, building works.
Food and beverage — breakfast throughput, quality, opening hours, in-room dining.
Value and rate perception — what the guest felt they paid for versus received.
Service recovery — whether a raised problem was resolved during the stay.

Score every property on the same scale

The purpose of scoring is comparison, so the scale must survive different volumes. A 180-room airport hotel and a 62-room select-service asset will not produce comparable raw counts of any issue; they will produce comparable rates.

Express each category as a share of classified feedback in the period, not as a count, and hold the period fixed across the estate. Then record the volume alongside the rate, because a rate built on eleven comments is a signal to watch, not a finding to act on.

Same window for every property — a rolling 30 or 90 days, chosen once.
Rate per category, with the underlying volume shown next to it.
A minimum volume threshold below which a property is flagged 'insufficient evidence' rather than ranked.
No manual adjustment for a property having had 'a bad month' — that is what comp-set context is for.

Separate portfolio patterns from property problems

This is the distinction that changes what you do on Monday. A portfolio pattern is an issue whose rate is elevated across many properties at once; it is nearly always caused by something above property level — a brand standard, a rostering model, a procurement change, a systems rollout, a capital deferral. Sending it to individual GMs produces fourteen local workarounds and no fix.

A property problem is an issue whose rate is elevated at one or two assets while the rest of the estate is flat. That belongs to the GM, with a date.

A workable test: rank properties by the rate for a category. If the elevated properties are a minority and the rest cluster low, treat it as local. If most of the estate sits above where you would accept it and the 'good' properties are the exception, stop asking GMs and go look at what those properties do differently — then go look at what changed above them.

Give every issue an owner and a recheck date

Analysis that ends in a report ends. The only step that converts feedback into operating change is naming one owner per issue and fixing the date the same measurement is repeated.

Recheck against the same category rate over the same window length, not against a general score. General scores move for reasons unrelated to the fix — a rate change, a seasonal segment shift, one unusually loud group — and they let a failed fix pass as a success.

One issue, one owner, one date. Shared ownership is no ownership.
The recheck window must be long enough to accumulate evidence at that property's feedback volume.
Record the verdict explicitly: improved, unchanged, worse. 'In progress' is not a verdict.
Issues that fail two rechecks are escalated as structural, not reassigned.

An illustrative example

The numbers below are invented to show the shape of the output, not drawn from any portfolio.

Suppose arrival and front desk issues run at roughly 6% of classified feedback across eleven hotels, but three properties sit near 20%. That is a property-level pattern: three GMs, three dates, one recheck. Now suppose instead that nine of eleven sit between 17% and 24% and the two low ones both moved to a different arrival rostering pattern last quarter. Nothing about that is a GM problem, and fourteen local action plans will not touch it.

What to review each cycle

Portfolio ranking, worst-first, on one comparable scale.
The top three categories by portfolio-wide rate, with the spread between properties.
Properties flagged for insufficient evidence — a collection problem, not an operating one.
Every open issue past its recheck date, with its verdict.
New issues that were not present last cycle.

Doing this across a whole portfolio

AnswerSignal Hospitality runs this method continuously: permitted guest feedback is classified per property, every hotel is scored on the same scale and benchmarked against its comp set, and assigned issues carry a recheck. It is a monthly subscription and setup is done with you rather than self-serve.

View AnswerSignal Hospitality Pricing Begin portfolio setup