Methodology
How the Counterwise Score works
We rate each rental desk at a specific airport or train station, not each brand. Here is exactly how a desk gets its score.
Published by Counterwise · Last reviewed
1. What we measure
The Counterwise Score (0 to 10) is a weighted average of five sub-scores. Each sub-score is also 0 to 10, and higher is always better for the traveller.
| Sub-score | Weight |
|---|---|
Deposit & damage disputes How often travellers report surprise damage charges, slow deposit refunds or disputed scratches. | 30% |
Counter upselling Pressure to buy extra insurance, upgrades or add-ons at the desk. Higher means less pressure. | 20% |
Booking honoured Whether you get the car category and the price you booked, without new fees at the counter. | 20% |
Queue time Typical wait from reaching the desk (or shuttle stop) to holding the keys. | 15% |
Car condition Cleanliness, mileage and mechanical state of the cars handed out. | 15% |
Deposit and damage disputes carry the most weight because they cost travellers the most money and stress, often weeks after the trip.
2. How to read a score
- 8.0 – 10 · GreatBook with confidence.
- 6.0 – 7.9 · MixedFine if you read the red flags.
- 0 – 5.9 · PoorExpect friction at the counter.
Colours always come with a text label, so the scale works for everyone, including colour-blind readers.
3. Data sources
For each desk we use only data from licensed sources or official APIs, and only feedback that clearly refers to that specific location:
- The desk’s own Google Maps listing: its overall star rating (through the Google Places API) and up to 300 of its most recent reviews (through a licensed review-data provider). Review texts are read by our AI during the analysis only: we never store, quote or republish them, and never show reviewers’ names.
- Web search results (through search APIs) for the desk’s deposits, damage charges, queues and traveller discussions such as forums. We cite the pages we relied on in the “Sources” list on each desk page.
- Published terms and conditions (deposits, fuel policy, card rules), researched by us, with a source link for each fact.
Reviews about the brand in general, another location or a booking website are left out. We show the number of reviews analysed on every desk page. A practical fact we could not confirm from a source is shown as “Not confirmed yet”: we never guess it.
4. How reviews become scores
An AI model reads each desk’s recent reviews (newest first, each with its age and star rating) and the web search results, and scores the five topics above. Recent reviews count more than old ones: a review loses half its weight every 12 months. For every topic the AI must say how many reviews or pages speak to it; with fewer than 3, the topic is marked “not enough data” and left out of the score. The AI also drafts the short summary and the red and green flags, in its own words. Our code then checks every answer: links must come from the search results, nothing may be copied word for word from a review, and no personal names may appear.
Star ratings are rescaled the same way for every desk: a rating at the 5th percentile of all desks we have rated becomes 0, one at the 95th percentile becomes 10. While we have rated fewer than 20 desks, those percentiles lean partly on a starting estimate. The score then blends the AI analysis with the rescaled star ratings.
Desks with few recent reviews are pulled toward the average of their airport or station, so one or two extreme reviews cannot make a desk look excellent or terrible. Desks with fewer reviews than our minimum carry a “Limited data” label; desks with too little evidence get no score at all and are not listed until we have enough.
Consensus. We compare up to four independent signals: the AI’s review analysis, a separate score from web reports only, the rescaled Google rating and the rescaled average of recent review stars. When they point the same way the desk page shows “High consensus”; when they disagree, “Low consensus”, and the red and green flags deserve a closer read.
An editor checks every summary and every red flag before publication.
The sample scores currently on this site are fictional and were made directly as a weighted average of the five sub-scores, without this process.
5. Update frequency
Scores are refreshed every month. Each desk page shows its last update date. If something changes significantly between refreshes, for example a franchise changes hands, we update that desk sooner.
6. Independence policy
- Rankings are never for sale. No company can pay to appear, move up, remove a red flag or change a summary.
- We list every desk at an airport or station, including those we earn nothing from.
- Affiliate commissions are handled separately from scoring. The scoring process never sees commission rates.
- Rental companies may send us factual corrections (opening hours, deposit amounts). They cannot change scores.
See our affiliate disclosure for how we make money.
Prototype note: this page describes the method our data pipeline uses. All scores currently shown on this site are sample data for fictional brands.