Marriott Orlando Reviews: Dirty Rooms Cause Bad Stays, the Front Desk Barely Does
A causal analysis of 9,446 public Google reviews of 45 Marriott-family hotels in Orlando. The topic guests mention most has almost no effect on a bad stay; cleanliness and people do.
Most CX dashboards rank issues by how often guests mention them. This report takes one public company, rebuilds that dashboard from public reviews, and then asks a different question of the same reviews: which issues actually change the chance of a bad stay or a great one. The two rankings disagree.
What was analyzed
| Source | Public Google reviews. No survey or Marriott data was used. |
|---|---|
| Reviews | 9,446 reviews, 8,504 of them with a star rating |
| Scope | 45 Marriott-family hotels in the Orlando metro |
| Date range | January 2025 to July 2026 |
| Structured fields | 23 dimensions per review |
Key findings
- The front desk is mentioned constantly but barely moves the score. Front desk and check-in is the dashboard's number-two priority, yet its causal effect on a one- or two-star review is about two points, and it is not statistically significant.
- Cleanliness is a real cause. A dirty room raises the chance of a one- or two-star review by 34 points, with every effect passing placebo, random-cause and subset tests.
- A person beats a price seven to one. An employee who goes beyond raises the chance of a five-star review by 42 points, against 6 points for a good price. Price and location are among the weakest levers, and a surprise fee can only cost a guest, never win one.
- The descriptive view looks healthy. Rebuilt from the reviews, the dashboard shows NPS +32 and CSAT 68%, flat for 18 months. Strong stays and broken ones cancel out in the average.
- Hotels differ widely. The share of poor stays runs from 7.89% to 41.18% across the 45 hotels.
Exhibits from the report
From the report: What Your CX Tools Can't Tell You
From the report: What Your CX Tools Can't Tell You
From the report: What Your CX Tools Can't Tell You
How the analysis was done
Every review was enriched in one pass into 23 structured fields, such as the cleanliness signal, a fee or billing signal, staff interaction and whether the guest says they will not return. The causal estimates come from a DoWhy model on the 8,504 reviews with a star rating. The treatment is defined from the review text and the outcome is the star rating, holding the hotel, tier, language, review length and reviewer type constant. Each estimate is cross-checked with propensity-score weighting and stress-tested with placebo, random-common-cause and subset refutation tests. The dashboard view uses the standard proxies: NPS from the share of five-star minus one- and two-star reviews, and CSAT from the share of four- and five-star reviews.
Reviews come from guests who chose to post publicly, so they are not every guest. Public company facts in the report are cited separately and never merged with the review estimates.
Related analysis
- Correlation vs Causation in Customer Experience Data
- How to Build a Customer Experience Dashboard in 4 Steps
- Which Complaints Cause Bad Reviews? Hotel, Gym and Café Data
- Voice of Customer Tools Are Good at Listening. They're Bad at Explaining Why.
- Every report in the Artifacts Gallery
To run the same analysis on your own reviews, tickets or surveys, start with your data.