Search (tradeoff vs parametric)


Retrieve highly relevant results that others miss:

Parametric search also known as constraint based or SQL-type searches, can be represented in an n dimensional space, where n is the number of attributes, by a box based on the search constraints (query conditions). The search retrieves the data that meet the query constraints in other words the elements that are in the box.

For example, the selection of a car that has a performance that exceeds 200 horsepower and priced under $25,000 would eliminate a vehicle with 250 Horsepower that costs $25,500. It only returns the cars within the constraints or “inside the box”.

Optimization theory, teaches that ideal results are usually found close to the pareto optimal boundaries. These boundaries are typically the intersections of curves, planes and graphs representing constraints, utility functions, etc…

In the case of a parametric search, this intersection corresponds to the corner of the box. With SQL-type searches, data that falls outside the box is eliminated and does not show up in the result set. This is why parametric search often requires several trial and error attempts to locate the required data.

This means that a parametric or SQL-type search will miss highly relevant results that happen to “land just outside the box”.

In comparison, Auguri’s search works by identifying an ideal result (often hypothetical) and measuring the distance of each element to the ideal in an n-dimensional vector space. Auguri will retrieve ALL the elements that are close to the ideal result. As a result every relevant data will be retrieved.


Allow users to determine how to prioritize their results

Until the advent of Auguri computers were really good at searching in a black or white environment (a condition is either met or it is not). But they were notoriously bad at dealing with tradeoffs and “gray zones”. With Auguri this changes. Auguri introduces a new element to the search: the relative importance of the criteria. This additional dimension yields a highly more relevant result. For example if you are searching for a car with more than 200 Horsepower (Performance > 200HP) under the price of $25,000 (Price < $25K) traditional SQL-type queries would retrieve the same result irrespective of the affinity of the users to performance and their sensitivity to price.

On the other hand, Auguri will retrieve different results depending on whether the user is a “car buff” who is primarily interested in horsepower and would tradeoff price for a higher performance engine or if the user is price sensitive.

Auguri takes into account the relative importance of every criterion by weighing every criterion this translates in a stretch of the axis accordingly.

Leverage interactive aspect of the web to capture user needs

The challenge of the single field metaphor is that it makes it difficult to prioritize the results of the search to meet the specific needs of a user. This problem is compounded by the fact that (i) typical searches have very little context (i.e. all the Google engine knows are the words that are typed and there is very little else that is known about the user), and (ii) the amount of data on the web increases geometrically thus yielding in excess of 2,740,000 results for a search for “Large Screen Laptops”. GYM are focused on building smarter more intelligent ranking algorithms.

To address this challenge we believe that a paradigm shift is warranted. The new approach we propone, is based on leveraging the interactive aspect of the Internet. The best ranking algorithms developed by the most talented engineers of GYM in an attempt to guess the intent of a user will not yield as relevant a result as an interaction with the user to better understand what they are seeking and most importantly letting them decide how to rank, order and prioritize their results.
This new approach is particularly more attractive given that the associated revenues are highly dependent on relevance for click-through for example. In addition the interaction gives the chance to display more ads.

The new approach is predicated on building a platform that allows the user to select the way they want to prioritize their results instead of a relying on the guesswork of the sophisticated algorithms of GYM’s engineers. The Auguri patented technology and framework that was developed over the past 7 years is a perfect fit for this platform.
To illustrate our point, if someone walks into a video store and asks the clerk to recommend a movie, GYM’s approach would be to identify statistically the most popular movie. Our approach would be to engage the customer by asking about their mood, the genre they are interested in, if they have a preferred actor, whether they favor a new release or if they have a certain period in mind, etc… and accordingly match the movie that best meet their needs.