Wiki Lead : Drew Hill

Wiki Contributor: Amit Elazari, Pooja Jain

Topic Audio :



This week, we discussed data transparency with data scientists at Glassdoor–Andrew Chamberlain, Ling Cheng, and Morgan Smart. Glassdoor was founded with a very simple question in mind: what can be done to bring more transparency to a world where very little exists? Today, their mission is to help people find jobs at companies they love, and that they will continue to love for ten to fifteen years or more. Moreover, by allowing candidates access to information pertaining to their future potential employers (such as salary, benefits etc.), Glassdoor solve some of the information asymmetry problem associated with job searching, creating a “two-way window” where to employee can retrieve information regarding the employer.

The Glassdoor Marketplace
Glassdoor generally functions as a give-to-get curator of job-related data. Their website enlists employees to review their company's past and present, in terms of environment, salary, perks, and pace, among other things; provides a space for interviewees to rate their experience and post helpful hints to fellow job-seekers; and provides a space in which companies can efficiently acquire great talent. In doing so, the “Glassdoor Marketplace”, pictured below produces value for both sides of the career market (job seekers and employers). In fact, Glassdoor claims that the unique content they provide allows companies to hire talent of twice the quality as they’d otherwise find. This is evident in the 600k companies represented, 34 million monthly unique users, and 190 countries in which they are active.

external image B1R9dEAYQkQ6pDTA-iyXEhHV8savVD8XrZTCPtbHLweG8HZcQ6p_iceiEEZ63YiiW8ucDQDJw5gOv_priyCfgQhdTEY9wUQs06li4yAhsK1x-Fmw2W-AbdoTbyJoYmz0uPsptEln

Transparency as a force for social good

Glassdoor has set out to alter the way the job market works by opening the kimono of the workplace (™ SDR2016). Before they arrived, it was near-impossible to make data-driven decisions about what sort of salary to ask for, which company best fits one’s personality and work pace, or how to prepare for an interview. The Glassdoor Marketplace shines light on workplace practices-- good and bad alike-- to reward employers who treat their workers right and pressure those who don’t into making changes. This example illustrates how “data rights”, and more specifically the right for transparency, is utilized in order to change industry norms and to create benefits “for the people”.

But data transparency is tricky. The data provided in the Marketplace stand to be immensely useful for improving both humanitarian circumstances. One of the primary reasons such data were previously so hard to come by was fear of retribution by employers. Employees with truly dismal -- or even just plain mediocre-- opinions of their place of employment were likely to hold those opinions to their chest, lest their employer take punitive action and fire, demote, or ostracize them. Employees even faced legal compliance issues if they released their salary information. It was not until Glassdoor offered a fully-anonymized forum for employees to post their unbridled opinions, openly discuss interviews, and freely post the salaries of their given titles at a given employer. In this regard, Glassdoor creates a window into otherwise opaque workplaces, while carefully protecting the personal identifiers of those within them. Moreover, Glassdoor’s “__Know your Worth__” tool, allows users to understand how much they should be compensated, according to their experience and skills, thereby providing the users an “order of magnitude” to compare their information to others. This illustrates how the right of transparency sometimes requires an equivalent right to view your information in contrast to other users information, and compare it, in order to reach conclusions.

As Glassdoor expands the transparency of the job market, employers are also beginning to realize substantial gains. Unrestricted and unbridled reviews from employees offer businesses an invaluable insight into practices that affect employee productivity and bottom line. Glassdoor makes its data available to companies for download (by request) and examination. Such analyses allow them to pinpoint and target job titles with low satisfaction and make improvements. At a higher level, these data provide information to stock-holders and other investors who wish to understand the nature of the company and perhaps evaluate, based on temporal and/or cross-sectional trends in satisfaction, whether there’s room for added value (or whether there may, in fact, have been too much value placed on the company to begin with). Worthy of note is that Glassdoor allows businesses to respond to reviews; they are not allowed to alter or unduly influence them.

Social Data Provides Unparalleled Research Opportunities

As previously mentioned, Glassdoor delivers information to over 34 million unique monthly users on over 600k companies represented 190 countries. This provides them with a great trove of data with which to iterate on their product. This was demonstrated to us in class by a demonstration of an internal project performed to identify and address bias in employer reviews. As Morgan explained, reviews at Glassdoor and other companies, like Amazon, are often distributed in a J-like curve (pictured below). Efforts have been made to adjust for this behavior post-facto (so that aggregate ratings aren’t not unduly skewed in one direction or another), but Morgan and her colleagues wished to understand how they might prevent it. They ultimately believed that those users most-incentivize to review an employer, interview process, etc. are those who either had extremely positive or extremely negative experiences. Those with run-of-the-mill feelings often didn’t feel like it was worth the effort to say something, well, so-so about a company. To test this theory (and a possible solution!), Morgan et al designed a test to examine whether a give to get model, in which consumers would be forced to create an account AND submit a piece of content in order to see more than three pieces of content. This model further contributes to maintaining an important cycle: The More Content, The More Users --> The More Users, The More Content.
external image gaAnmormVT8Reev2RJS6Tu_sfDfJjkQKSJy3gYQ4O8H5EoGp6C-0U20-F7lsCVLuCPvOd0heccfTzxUKKEksn22n8fZ3uC3APfmDAziZSPtGCYGrSni0H9f_qeOfa0-U0rVNoKJh

The study, summarized in the figure below (top), examined 363,168 unique US employer reviews published on between 2013-2016 to see if they might be identify the relative impact of Give to Get on reported satisfaction with the mployer. The results confirmed their hypothesis (of Give to Get as a means of boosting mid-rated reviews) rather dramatically, as shown in the bar chart below. Reviews of 3s and 4s were much more common in those who signed up via Get to Give, rather than those who signed up with a bone to pick or praise to sing.

external image UhTNwTzph8xW9TN89Y5YSUyUEyXPHg9lxYg4hUR3q6QziNKIBt80-2nHivwSfkQLq1e8YudCHAjmBUtNBSlx67-5_pxWXcS7cLzmDcFvgUkOwYeeDgvn4G6sn9zcNcNTPoRRt3Ma

external image NJgn28j7UVGJKzXap3v5id_t4INCp46SBWBnLTsLTXCUs1kpf1K3zyGwRmawFI0lp6kfLrbMNuKXYDLUw__sOR5vJ9DiaREm4Ke-btHt2vJL73LmexU4zbhN6lkxofnKebDyQIgu

The massive amounts of data Glassdoor collects also puts them in a unique position to provide highly-diverse and statistically powerful insights into the marketplace. Such research has the potential to advance their brand, of course, but it can also create great utility to society. For example, Glassdoor recently put these data to work understanding the gender pay gap. In another study, they found a significant causal relationship between employee satisfaction and company performance, demonstrating to the world that the Capitalistic, profit-wise approach to employee treatment is to treat them better. Interestingly, these data also allowed Glassdoor to uncover a very low impact of wage on satisfaction. These results, which showed correlated a 10% increase in wages with a mere *1%* increase in satisfaction, demonstrates that maybe money really doesn’t buy happiness, and that we should all maybe look for a job we love rather than a paycheck we like.


Glassdoor is an exemplar of how people, in this case, potential employees, could gain “a seat at the controls”, and howtransparency could result in more agency. Glassdoor provides the users with means to analyze and interpret their data and compare it to other users, in order to reach important conclusions regarding their future career. Achieving transparency in the information age will require formulating a “two way window” between users and data refineries, thereby allowing users to access and inspect the data collected by refineries, and further understand how refineries utilize such data.

The right to inspect data refineries should also include a right to see “a data safety audit” and “privacy efficiency rating” as well as “return-on-data” score. Various legal regulations require that refineries will provide a notice in a case of a security breach, in certain circumstances (see generally). In addition, visibility and transparency are foundational principles of “Privacy by Design” mechanisms (see here). Nevertheless, there is still a long way to go until the a-symmetry relations between users and refineries will change. Demanding our right to transparency, is an important step in order to evolve from data of the people to data for the people.