Over the years, HomeWarrantyReviews.com has managed to establish itself as the #1 review website and the primary source for consumer research on home warranty companies. And here, we shall explain the process we employ to rank the various companies.
The ranking process is completely automated and we cannot alter the output of the ranking algorithm as per our wishes. If a company is ranked high on our site, it is purely based on consumer ratings and a few external factors explained below. We do not favor any company based on our advertising or marketing relationship with them.
First things first, what is the difference between ranking and rating?
The inputs for rating a company are the star ratings awarded by its customers when they place a review on our website. We have an automated system for keeping track of the ratings that a company receives. So the higher the number of good reviews a company receives, the higher the star rating will be. The higher the rating, higher the ranking!
To rank the companies on our site in a specific order, we employ a well accepted sorting formula. The algorithm we use for ranking the companies is the Bayesian Weighted Average sorting that ensures fair and unbiased results. The algorithm will also consider few other external factors explained in the next section.
While the good reviews heavily influence the ranking, companies are also evaluated on several other criteria.
Several hurdles are faced when a ranking is to be provided from user generated content. At times a company may achieve high ratings and ranking through cheating the system. Doing so involves posting fake reviews. Such reviews may also be used to make a fellow competitor look bad. We are very good with finding such reviews using semi automated methods and such companies will be imposed with a ranking penalty after they are caught.
Bayesian Logic is the most commonly applied logic for decision making and inferential statistical needs. It deals with probability inference, which means that it makes use of information from previous events to predict future events. Named after the famous Thomas Bayes, the Bayesian ranking method helps implement the most common way to rank entities – on a scale of 0 to 5.
This is a rather complex algorithm to explain to the layman, so we’ll use a nice example to convey how it works. Let’s assume that we only have three entities to rate – A, B and C. Now, the Bayesian rating system is built around “beliefs”, where a belief describes the probability of something’s value. Like how we say 50% chance for A and 20% for B and so on. The percentages here are the so called beliefs.
Now the basic strategy for awarding ratings to A, B and C as per Bayesianism would have the following steps:
Step 1: You begin with a mock (but equal) “belief” for each entity
Step 2: As the new data arrives, the value of the entity is updated
Step 3: Make use of the latest belief as the construct for sorting criterion
So irrespective of how wrong the first belief is, once the user generated data is considered, the ratings will be precise. The first belief also called the prior-belief, often describes a Bernoulli’s parameter. So the best way to select a value would be to use Beta distribution.
To implement the 2nd step, each time a good review is received, update the counter for good reviews. Now the prior belief becomes a beta distribution with the new values.
To implement the third step, create the sorting construct. In simple words, the sorting construct is something that answers this question: “Where does an entity belong in a list of entities whose average rating is known with 100% certainty?”
To make a belief a sorting construct, a loss function is formulated. Let’s call the loss multiple L. Now with the U – Up reviews and D – down reviews, the value of x, the sorting criteria can be solved using the following formula:
Ix (U+1, D+1) = 1/ (1+L)
The value of x then decides the rating an entity gets.
The ranking system we use is completely unbiased and automated. The system takes the inputs available and generates a ranking, with no human interference. This means that only customers can influence the warranty providers’ ratings on our site. This rating later forms the basis of evaluation for new and potential customers. The system ensures that the companies actively take an effort to provide good service to their customers.
While star ratings are the primary criteria for ranking, we do consider other ranking factors explained above. So, it is possible that you may see a company with a slightly lower ranking placed higher in the list. However, in such cases, the rating for such a company will not be significantly lower than its neighbors.
Now, you may ask why we would go through all this trouble just to provide ranking for warranty providers. The answer is simple, at HomeWarrantyReviews.com, the mission is to bring forth good specimens in an industry which is often under the scanners for all the wrong reasons. We have, over the years realized that home warranty customers need good guidance in selecting a reputable company and that word of mouth alone is not to be relied upon. So go ahead and use our website for an enhanced home warranty buying experience.
Thank you for reading this and we sincerely hope we didn’t scare the bejesus out of you with all the Bayesian explanation. Cheers!!