SmartCost Ranking Factors

We understand that large quantities of information are not useful if the quality of information is low or the information is difficult to understand. The following section describes how we calculate the rankings and opinion summaries found throughout the site.

Where do we collect our data from?

The data we process and analyze is collected using web crawlers (spiders) that are constantly checking thousands of data sources for reviews, ratings and price changes.

Among these data sources we crawl:

  • Raw hotels information - such as their location, list of amenities, images from our booking partners and/or the hotels themselves, etc..
  • Real-world parameters - such as the hotel renovation date, management and chains structure, closest metro, neighborhood popularity and more
  • Booking website reviews - people who stayed at hotels provide valuable feedback. Large travel agency sites (OTAs such as Booking.com, Expedia, etc) have lots of reviews per hotel, in some cases varies from hundreds to thousands of textual feedbacks.
  • International booking sites - we monitor websites across the world, processing feedbacks in multiple international languages. Then we translate (normalize) these reviews to English for single point of comparison.
  • Social networks - we monitor the social reviews each hotel receives through the various channels - Facebook, Twitter and Instagram. This allows us to build an updated data-points graph of current reviews.
  • Social travel networks - there are multiple social networks dedicated to the travel industry (such TripAdvisor).
  • Commercial reviews sites - we monitor websites dedicated for commercial reviews - Yelp, Foursquare, Google Locals and so on.
  • Professional travel reviews - in case the hotel was reviewed in by a professional travel site such as Frommer's Guides, Fodor's Travel, Lonely Planet, Michelin Green guide, Rough Guides, Timeout etc.
  • Travel bloggers - we monitor travel bloggers which review the places they stay at
  • Discussion boards - we also scan few travel discussion boards trying to find data related to the hotels we monitor.

How We Calculate the Rankings

Processing textual feedbacks is a complicated process that uses NLP methods (natural language processing, usually referred to as an Artificial Intelligence - AI).

Ironically, product and service ratings (or star ratings) were originally intended to help you differentiate one product or service from another. Unfortunately, the vast majority of products and services have an overall rating between 3.5 to 4.5 (out of 5.0) making differentiation difficult.

At Low Cost Hotels, we address this problem by assigning each property a distinct ranking across the entire city and also within their hotel class. The rankings show you how the hotel performed amongst other hotels in the same price range but also how the property ranks across the entire city. A simple way to explain our ranking is to imagine each hotel is competing in a race. As they cross the finish line each hotel is ranked in the order they finished with #1 being equivalent to first place. We chose to display rankings instead of ratings because we found it frustrating when products and services all appear to be above average.

We generate rankings for attributes such as service, room and value that are common to almost all hotels. In the spirit of differentiation, our rankings work much like a competition; therefore, the lower the ranking the better the performance.

Our rankings are based off of an average generated for each hotel with enough reviews to be considered statistically valid. The ratings are mathematically defined as a weighted average because we assign differing importance (or weights) to various review factors such as:

  • The quality of the site in which the review originated
  • The date the review was posted (the newer, the better)
  • The quality of the review and the reputation of the author
  • The rating of the original review
  • Specific comments made inside of each review

Lastly, we normalize the ratings to prevent cases where all hotels in a specific city have nearly the same rating (similar to grading on a curve).

#50 out of 100 is not the same as #50 out of 50:

Unlike other sites, our rankings also take into account whether or not the other hotels you are comparing against have enough reviews to be credible. In other words if there are 100 hotels in the city but only 50 of them have reviews each hotel is ranked against the other 49 hotels not the total population of 100. Of course, LowCostHotels still includes all 100 hotels for you to investigate however a hotel without a ranking indicates that we didn't have enough reviews to accurately rank the hotel.

Opinion Summaries

We believe that most people planning a trip spend most of their time searching and browsing a sea of information trying to determine which hotel is the best choice for their next vacation. With this in mind we developed a technology that automatically analyzes and summaries hotel guest reviews so you can find your perfect hotel quickly and easily.

How we summarize reviews

We built our technology to summarize guest reviews by giving a few gifted computer engineers too much caffeine and unlimited number of computers to create an artificial intelligence or more specifically natural language process (NLP). This process semantically analyzes portions of each hotel guest review against an ontology to determine the subject matter (i.e. hotel room, location) and the sentiment of the statement (i.e. clean room or dirty room). More simply, our system compares each statement against a database of common words and statements found in hotel reviews. For example phrases like "paper thin walls", "earplugs", "quiet", and "peaceful" are commonly used to describe how noisy or quiet a hotel room is.

Finally, we identify common patterns resulting from this analysis to develop the good and bad summary you see on the site.

The result is the ability to look at hotels top-down instead of trying to solve a mystery a review or sentence at a time. At a glance we can determine things like the quality of the free breakfast or how good the hotel is for a romantic getaway.

Incorrectly Categorization

Software is rarely perfect so occasionally you might encounter a situation where a phrase is incorrectly categorized. For example consider the following sentence: "The hotel's idea of excellent customer service and my idea of excellent customer service could not be further apart.". It is possible that our software might see the words excellent customer service and make the determination that the sentence is positive. We are always improving our systems ability to understand the English language, so over time you should see less and less of these errors. We also put in safeguards to prevent the software from making incorrect recommendations as a result of errors in our language analysis.

Language Inconsistency

In order for a feature to be listed in the "The Good" or "The Bad" section of the web site the feature must have enough evidence from enough credible reviews. We built our system with a margin of error to account for the occasional mislabeling of statements. This not only improves the quality of the recommendations but also means that you are looking at an accurate summary of the underlying data.

You Are The Real Judge

None of this matters at all if you do not find LowCostHotels useful. Please, tell us what you like and what you do not so we can improve. You will find that we are hooked on opinions especially those from our own users! Please send an email to [email protected] if you have any questions or suggestions on how we can improve.