About the tool!
Frrole is a social intelligence platform that mines precise content and generates deep insights from Twitter. The engine that powers Frrole has the ability to analyze hundreds of millions of universal data sets in real-time. For all these data sets, Frrole performs extensive analysis across semantic, metadata and statistical dimensions. It leverages standard and custom-built algorithms around ML, NLP, NER and clustering to do this analysis.
While most Social Analytics/Intelligence products provide results based on statistics and the first level of NLP, Frrole goes two levels deeper building semantic context for each topic and tying it up with information available in the general and historical data sets.
Who founded the product/company?
Frrole was founded by Amarpreet as a consumer news product. Nishith and Abhishek joined him in early 2013, and the team pivoted to focus on the enterprise. Nishith and Amarpreet share the same alma mater, and had met via their alumni group. Abhishek had met Amarpreet through a Trilogy ex-colleague of his.
What platforms does it cover?
Currently, Frrole analyzes Twitter data because of its availability as the largest public data set. In medium-to-long term, we have plans to include other social networks into our offering.
Features of the tool?
Frrole’s media offering built on top of its intelligence allows TV, digital media and brands to integrate very precise social data right into their products & campaigns and create visual applications like election trackers, movie reviews, second-screen displays for TV shows etc. This enables them to create experiences that are immersive, real-time and social. The USP is almost a brand new generation of content, which helps our customers create audience engagement and reach at a social scale.
What analytics support does the tool provide?
Typically, data analysis can be broken down across 4 levels: Statistical, Linguistic, Semantic & Contextual.
A very large majority of data analytics products have a good ability to answer questions of statistical or linguistic nature, but not the latter two. Frrole has built the ability to go much deeper, leveraging semantic and contextual analysis, which although can never be perfected, we have been able to make some strides there. For ex., Frrole is able to break down each tweet into nouns, adjectives, verbs and understand the context around it. Going one level deeper, Frrole also understands whether the noun is a person, a place, a sports team or something else.
For ex., a brand can know the sentiment of its recently launched campaign across top 10 cities in India in real-time, a TV Show can bring whole nation into its studio by displaying their tweets, votes and related buzz while the show is live on air, a retail company can read signals on twitter about their potential users and take appropriate actions.
Does it give an option to download reports?
Currently, we don’t provide any dashboard platform and reports mechanism. We provide relevant data points, like engagement and reach achieved, but not a detailed report.
What kind of sentiment analysis does it do and how accurate is it?
We use certain global sentiment libraries which we have trained further and modified based on our two years of experience. We are able to calculate total positive/negative sentiment for a collection of tweets, distribution of tweets into positive/negative/neutral and overall mood of the users based on combined sentiment value of all tweets.
Sentiment for individual tweets might not be very accurate always, since sometimes tweets comprise of very neutral terms like ‘Yeah, right’, the sentiment of which depends on the tone in which it has been said. But when you aggregate all tweets, our sentiment results are pretty accurate to upto 70%, which is a global standard.
What is the algorithm for sentiment analysis?
We breakdown each tweet into words and divide them into categories like nouns, adjectives, verbs etc. Each word is then analysed for positivity/negativity on a scale of -3 to +3 (good might be +1, fantastic might be +3, pathetic might be -2). The overall sentiment is arrived at by cumulative result of these individual sentiment values. We also take into account if a sentence has a negative qualifier for a positive word, for ex, “This is not good” will be a negative tweet, because even though good is a positive word, it has been qualified by a negative word.
What are the brands/agencies that are using this tool?
Headlines Today, UTV Disney, Times of India, Indian Express, Times Now, Celebrity Cricket League, Lakme Fashion Week, Nasscom Product Conclave, BrizzTV (for Airtel Social TV), U2opiaMobile, Zoom TV are among our intitial set of clients. There are other potential customers from media, brands and retail verticals that we are in discussion with at various stages.
Pricing & Packages
Our pricing is based on no. of data streams a customer is using for a specific application/program/campaign. We define a data stream as a well-defined and specific collection of tweets/insights pertaining to any hashtag/location/category/keyword etc. Currently, we charge at $100 per data stream per month, and a typical application uses 8-10 data streams.
We believe that social data contains lot more value than what current breed of Data Analytics products is able to extract. A layer of deep semantic intelligence, on top of statistical and shallow semantic analysis, is what is required to mine this value.
Starting with Twitter data, our goal therefore is to build social intelligence like it has never existed before. In long term, we aspire to do for social web what Google did for web.
Amarpreet and Nishith are MBAs from IIM Kozhikode and Abhishek is B.Tech/M.Tech from IIIT, Gwalior. Amarpreet has worked with Trilogy, Nokia and HP and now takes care of business & product at Frrole. Nishith has worked with Jaguar Land Rover, Tata Motors and is responsible for marketing & front end products. Abhishek spent a few years in academics and research before Frrole and heads engineering & technology here.
Are you hiring?
We are hiring for sales, product and technology roles. Plus, we are always on lookout for short term assignments/internships if we find a strong candidate. Currently, we are also looking to hire a front-end web development and a design intern.