Active Hackathon

Social media analytics: Insights that standard KPIs fail to generate

A look at the social media stats of e-commerce companies in India makes it clear that Social media has been one of the key differentiators in their growth story. Most of these companies successfully managed to acquire a good customer base by offering deep discounts, running campaigns, contests, quizzes, encouraging them to share Selfies and directly engaging with them.

Company Since Followers count Count of Tweets
Flipkart July,2009                 619,000                    23,400
Snapdeal Jan,2010                 285,000                    24,200
Amazon Mar,2013                 273,000                    20,200
Myntra Nov,2009                 188,000                    77,600
Jabong Dec,2011                 170,000                    38,200
Ebay Mar,2009                 160,000                    57,700
Reliance Digital May,2010                    64,100                    14,300



Sign up for your weekly dose of what's up in emerging technology.

There are two purposes for which any e-commerce company invests in social media, either Promotion and customer acquisition or Customer services.

Brands carry out promotions and customer acquisitions through contests and campaigns such as encouraging people to post ‘selfies’ on festive occasion, creating ‘virtual Big Boss’ on Twitter (as Snapdeal did when it was sponsoring Big Boss), etc. However, these methods work once awareness about the brand is in place.

For an upstart, creating awareness is the first step followed by campaigns. Yet, measuring success merely by understanding how fast it has acquired new customers can often be misleading. Making promises is easy but delivering is challenging.

If you want to know the real performance report card, skim through their support handles and mentions rather than the promotional handles.

For tweets in campaign trends like #JabongShopForFree doesn’t really capture true picture of customers. These are those people aspiring to win the promised gifts free. Judging whether they are your potential customers in future based on such participations can be dicey. Take for example:

Most of the articles floating around on web seem focused on promotional aspects of these companies. KPIs like share of voice, engagement ratio, Potential reach, Number of followers etc, in my view, may not really give you a true picture. Take, for example, this article on Mint claiming “favorable mentions as %age of total twitter mentions” for 6 leading e-commerce brands. There are two problems with such kinds of claims:

  1. They may be mixing Promotional handles with customer support handles or focusing only on promotional handles of companies.
  2. They may be using a generic algorithm to decide which tweet is favorable and which one is not.

As a thumb rule, in my view, companies should maintain two twitter handles:- one for Promotional purposes and other for customer support. Flipkart seems to be the first company to have done it and we know how quickly they shot to fame.

Though lately, Myntra and Snapdeal have also dedicated twitter accounts for support as well as for promotions. It makes your life easy since dedicated account for customer support means that most of the tweets are asking for your attention and you are at the risk of losing those customers in case you don’t address their concerns. Ebay, Jabong, Amazon India have yet not come up with dedicated accounts for customer support. This means that if you skim their twitter handles, you will find more mixed reactions compared to Flipkart that has dedicated accounts.

Challenges in Social media analytics: Understanding right keywords for sentiment analysis

Coming to the problem of using generic algorithm, studying “Mentions” and trying to do sentiment analysis is a thankless job because most of the mentions that companies get are from users screaming about one issue or other and very rarely does a satisfied customer come back and thank for smooth delivery or good product. I personally seldom do. I was fully convinced that one cannot use “twitter mentions” to do sentiment analysis given the fact that most, if not all, the tweets are terribly negative.

How about Timelines? Well, tweets on timelines are mostly boring beyond limits as most of them are templated. However, among those boring tweets, at times, i did get some of the positive tweets. Interesting thing to note here is that each twitter timeline can have its own set of words that will make its tweets qualify positive and you cannot pick up a set of words from Flipkart’s timelines and apply it on Myntra and conclude that they don’t have positive tweets on their timelines. This is the beauty and this is the complexity of twitter sentiment analysis.

Based upon my analysis of more than 3000 tweets of each company, I have listed the positive keywords for some of the biggest e-commerce brands in India below. Frequency is determined by number of times these words occur per 100 tweets and accuracy is nothing but number of times these positive keywords are actually positive. For example, Positive keyword picked for Reliance digital is glad but not all tweets containing the word “glad” are positive like this one.

Twitter Handle Positive keywords Frequency Accuracy
@flipkartsupport Glad Low 45%
@Snapdeal_Help [Smiley] :) Medium 74%
@ebayindia No Positive keyword N/A N/A
@bigbasket_com Awesome Medium 69%
@cromaretail Compliments Low 80%
@ShopClues No Positive keyword N/A N/A
@myntrasupport Woohoo High 95%
@reliancedigital Glad Medium 18%
@jabong No Positive keyword N/A N/A
@limeroadcom Delight Low 81%
@pepperfry Kind Medium 93%
@Urbanladder So Happy High 95%


Thumb rules for running a social media account: Know what not to do

This brings us to an interesting question: What kind of insights do Twitter Mentions and Timelines generate that can be used to improve customer satisfaction and cut costs? Upon diving deep into things, I have prepared a list that could be of help not only to e-commerce companies but also to marketing research companies and new companies planning to make space on social media. Here you go one by one:

  1. Do not give arbitrary replies like we will get back to you. Always share timelines like-We will get back to you by tomorrow 6 PM. Take for example, this conversation with Ebay and the kind of digital footprint the company is leaving.
  2. Do not give replies in templates. People hate templates and like customized and to the point information. My analysis says that Flipkart’s replies are mostly templated leaving behind a bot like impression. On the contrary, Myntra responds consciously. Urban Ladder is doing best in terms of giving personalized messages with a human touch.
  3. Do not misinform and mislead people. Flipkart seems to be overdoing things and there are too many customers complaining about different versions of truth being spoken by different people. At times, they provide wrong address of service center. On another occasion, they provided wrong information about policies
  4. Learn to empathize and take feedback for making life easy for your customers.
  5. Do not spam.

Social media exposure risk: What your social media accounts reveal about your business?

Social media can make or break your business in today’s digital age. A bit of analytics and a bit of analysis can reveal everything about your business – your strengths as well as weaknesses. If companies understand the importance of wealth of information hidden in their social media accounts, they can take this as an opportunity to understand underlying challenges impacting business and improve operational efficiency to directly impact business.

I spent around 10 days studying a dozen companies and discovered at least 2 big challenges that each company can address and make significant impact. Here are some of the most apparent ones:

  1. Flipkart:
    1. Has some serious problem with website. The order return button doesn’t display for many customers.
    2. Also, it seems to have some problem with payment gateway.
    3. Has many customers complaining “out of stock” products displaying available for sale. I have personally also faced this issue once when i bought a book.
    4. Has some serious issues going on with sellers.
    5. Is changing policies with respect to cashback points, deliveries and not creating enough awareness about it. There are ambiguities with respect to return policy, warranties and wallet money.
    6. Abruptly cancelling orders
    7. Product issues, multiple replacements
  2. Snapdeal
    1. Many complaints about product quality.
    2. Templated replies without co-ordination from operations and delivery team, delays in delivery.
  3. Ebay India
    1. Indifferent to Social media complaints from customers. Most of the concerns are replied but not resolved.
  4. Jabong:
    1. Delays in delivery
    2. Ambiguous clauses in price, tax and other charges.
  5. Lime Road
    1. Poor delivery.
    2. Poor marketing campaign leading to spamming.
  6. Pepperfry
    1. Website issues, and order tracking issues.
    2. Serious delivery problems and inventory issues.
  7. Croma Retail
    1. Delays in installation
    2. Poor and unresponsive website
    3. Pricing issues and indifferent on social media.
  8. Urban ladder
    1. Website issues, problems while placing order.
    2. Refund delays

Above observations have not been done by randomly picking up any tweet and quoting them. Upon studying more nearly 3,000 tweets which is mostly a data collected over one week, I have listed some of those concerns which have surfaced for at least 5 times.

I think it is high time companies competing in digital space considered taking social media analytics seriously to differentiate themselves with rest of the competition. If not quarterly, at least an annual review of your social media accounts will definitely yield rich dividends.

Disclaimer: The views expressed in the article are author’s personal responsibility. His employer and Analytics India Magazine did not participate in this analysis.

More Great AIM Stories

Mihir Kumar
Coming from Digital Family, Married to Analytics, Dating Social media and Subject Matter Expert on Trade Promotions Management in Mindtree.

Our Upcoming Events

Conference, in-person (Bangalore)
Machine Learning Developers Summit (MLDS) 2023
19-20th Jan, 2023

Conference, in-person (Bangalore)
Data Engineering Summit (DES) 2023
21st Apr, 2023

Conference, in-person (Bangalore)
MachineCon 2023
23rd Jun, 2023

3 Ways to Join our Community

Discord Server

Stay Connected with a larger ecosystem of data science and ML Professionals

Telegram Channel

Discover special offers, top stories, upcoming events, and more.

Subscribe to our newsletter

Get the latest updates from AIM