Employee turnover is usually the #1 cost to businesses.

InsiderOpinion reduces employee turnover and maximizes the effectiveness of human resources.

We accomplish this by automatically identifying the expertise and satisfaction of people within any organization or network. All the platform requires is access to a network's communication.

This empowers:

HR to monitor satisfaction, expertise, and influence.
Management to monitor and compare organization-wide trends.
Employees to find expertise when you need it.
Schedule a call if this sounds interesting to you!

Through our web interface it's possible to:

  • Search people by expertise
  • Track of satisfaction and influence of people
  • Search relevant content, discussed by the network
  • Monitor trends of any topic or product
  • Monitor Promoter score of any topic or product
  • Compare trends, people and products

Setup typically takes less than a day. Simply send communications to the our platform we do the rest!

Technical Details, for the Bold

Deployment & Scalability

We try to keep everything simple. That's why our setup uses three of the most well known applications / frameworks: Ruby on Rails, Python Flask, and PostgreSQL. All of which are available at nearly every enterprise.

For a more detailed breakdown of InsiderOpinion, there are three components:

Ingestion API - Python Flask application, which can scale to as needed. Requires 220 Mb of RAM each, each app instance can handle an average of 70 comments per second with a 2.2Ghz CPU (recommend one app instance per thread).

Query API - Ruby on Rails application, which can scale as needed. Requires 4 Gb of RAM each, formats responses, hosts the basic UI and API.

Database - PostgreSQL database, can scale up to the comments of a 100,000 person company on a single instance. This demo is using an AWS RDS db.m4.xlarge (using general purpose SSD). Primarily IOPS bound, the current setup can handle up to 100 million comments and 1 million profiles, without issue. With the ability to scale up as needed.

All of the components above are easy to deploy and scale very well. For reference, the current demo contains 32 million comments, and the platform can process up to 2800 comments per second.

Yet, the whole platform only costs $600 / month in infrastructure on AWS (with on-demand instances).

API: Ingestion

Below is an example of a comment from a discussion sent to our server:

The system will then respond, unless otherwise specified, with the keywords (i.e. subject matter), sentiment, and a score (numerical measure of sentiment):

After ingesting comments, the platform is ready to go -- just like this site!

In addition to search, it's also possible to use the APIs to integrate into third party applications.

For instance, we have a slack integration / bot which upon a question being asked (or left unanswered) in a channel the bot will identify an employee can answer a given question.

For further details, feel free to contact us!

Expert Rank

We built this system by categorizing experts online and analyzing their discussions through data mining, sentiment analysis, and more generally machine learning. We call our algorithm for ranking ExpertRank, paying homage to PageRank.

Without going into too much detail here, ExpertRank works by determining an authors expertise in a network. Then using their opinions around their expertise to rank a topic, piece of content (web page, document, etc.), person, or object. This is in contrast to PageRank which weights web pages, domains, or institutions as credible, we rank people.

This leads to the neat feature where you can search for documents or web pages on our platform -- without our platform ever needing to access the documents or web pages.

In other words, our platform never accesses the documents / web pages, but we make them searchable! They only appear because experts are discussing them, as it relates to their expertise. From that, we can learn what the documents are about.