In the time that SetuServ has been working with twtData, they say that their ability to process workload has considerably improved, largely due to the speed at which twtData can fetch and process data for them, allowing them to focus on their core business, deriving insights from data, for their clients.
SetuServ is a data science company that was founded in 2017. Their expertise are in text-based analytics with an emphasis on natural language processing.
“Unfortunately, we did not have expertise in mining Twitter data. We could do the mining ourselves,...we have the technical ability but we had some speed issues. We were not able to do it quickly and because our expertise is in machine learning, we also did not want to spend a lot of time here.” said Srikanth, Analytics and Data Science Manager at SetuServ
Not wanting to take time away from their primary business offering, machine learning, they turned to twtData to provide Twitter follower and tweet data and develop bespoke solutions which would significantly increase their company's efficiency.
Data requested by SetuServ
1,100 Twitter handles/accounts followers were downloaded, which amounted to more than 1 million rows of data. We were able to provide this data in a matter of hours.
How was the data used?
Creating Network Graphs and Brand Share Reports used to take a significant amount of time; something which was vastly reduced when SetuServ started working with twtData.
Brand Share Reports
Being able to measure brand share means that SetuServ can measure their clients against their closest competitors.
SetuServ, as part of their offering to clients, produces Brand Share reports. These are like the social media equivalent to market share. In theory, given that Social Media often mirrors attitudes offline, if you were to overlay a Brand Share Report with a Market Share Report, you may get similar results.
Here is an example of a Brand Share graph. The chart below shows what share of mentions each of the clients drugs is getting on social media. As Srikanth explained, “So, essentially, we are helping them [their clients] create brand share reports… How frequently people are talking about you vs competitors”. The below example shows that, in this case, Drug D had the largest brand share over the three years data that this graph encapsulates.
How is Brand Share Calculated?
Brand Share is calculated by the number of mentions per quarter of each year. So in this case, it would be calculated by how many times each drug is mentioned. This creates a percentage share of the brand, much like market share is calculated as a business metric.
Topic Classification and Sentiment Analysis
“We get tweets,...likes etc and on that data, we perform lots of analytics. We do topic classification on tweets, retweets… topic classification on likes and we do sentiment analysis. Using topic classification and sentiment analysis, we can actually solve a lot of business questions.”
Srikanth Reddy, an Analyst and Data Science Manager at SetuServ, gives an example of one of these business questions and how social media activity affects sales below:
“…One of the analyses we did for a pharma company is about finding their product share vs competitors' share. They wanted to know or actually, we proposed to them that their sales are ultimately linked to how customers are… how frequently customers are talking about them. If a customer is talking about you, primarily positively, then it translates to higher sales and if a customer is talking about you negatively, it translates to negative sales.
“If a customer is not talking about you,...it also translates to no sales. So we proposed to them that we will measure your brand share. We’ll measure your social media brand share…and we’ll also give you sentiment analysis, like how people are talking about it.” explained Srikanth.
Here’s a case study where Topic Classification and Sentiment Analysis were put to use to help with the marketing strategy of major sports brands.
Case Study One - Helping fitness brands gauge positive or negative customer feedback from Twitter
Below, Srikanth talks about how they measure brand share for specific brands of shoes and what they do with this data before handing it over to the client.
“...let's take Nike, Addidas, Reebok, Puma. We can find out the brand share of these shoes. How many people are talking about these four brands.” Srikanth explained.
SetuServ uses social media data to see how positive or negative sentiment on Twitter affects sales. In doing so they are able to see if customer sentiment on a social media platform correlates with the direction of the sales of the shoes.
Gauging what customers like about products
Using Topic Classification, SetuServ can identify areas that are particularly focused on by a given target audience and then derive insight into how/what the customer feels just by analysing their Twitter activity.
SetuServ download tweet data, for example, tweets that mention the keyword ‘Reebok’ with the help of twtData and then perform topic classification. They can then create topics around particular elements of the product, for example, style, fit and comfortability.
“We can create all the topics that are relevant for footwear, and… understand what… people like about your products, what they don’t like about your products. Which models are doing good, which models are not doing good etc?” explained Srikanth.
The company also helps their clients do “live market research”, using sentiment analysis and topic classification, as opposed to traditional methods of market research, such as surveys or focus groups.
“You’re just collecting that information, giving you organic survey results. These are more organic because people are going onto Twitter and talking about it.”
Sports brands are not the only application of Sentiment Analysis which can be applied to business problems in many industries.
Case Study Two - Using Sentiment Analysis to monitor drug perception among physicians and doctors
Below is an example of a Sentiment Analysis chart produced by SetuServ, looking at the sentiment surrounding three different drugs before and after the American Society of Clinical Oncology (ASCO) conference in 2020. The conference took place from the 29th of May to the 2nd of June 2020 and this data goes from the latter part of 2019 to the first half of 2022.
The graph below helps their client, ASCO, gauge the perception of their drugs among their customers, the doctors who treat patients with them, according to sentiment on Twitter.
SetuServ analysed tweets data provided by twtData using a combination of downloaded tweets from select keywords and hashtags from certain accounts. With this, they were able to produce graphs such as the one below.
This graph measures the Net Sentiment, the net value of all the opinions about a product on social media. The Net Sentiment is calculated by calculating in one of two ways:
Method 1: Total positive mentions added to total neutral mentions, minus total negative mentions
Method 2: Total positive mentions minus total negative mentions
In this case, the graph shows that while there was an initial decrease in positive sentiment, both Drug 1 and Drug 3 regained positive social sentiment over the next two years.
SetuServ uses the data provided by twtData to identify trends in their customers' market. Given that what is trending on Twitter often mirrors what customers are talking about offline, there is a benefit to capturing this data for SetuServ’s clients.
Below you will find an example of a Network Graph produced for the client.
What is a Network Graph?
A Network graph is a mathematical structure showing connections between, in this case, Twitter users, however, points of data in general. In this case, the graph shows how different users are connected to each other (who follows who mostly). Entities (users) are shown as nodes (dots on the graph) and then the relationship between them is shown with lines connecting them.
SetuServ uses follower data for creating Network Graphs. Using a Network Graph, they can take all the followers of the Twitter profiles they are looking at and create a complex spiderweb type diagram. They use this graph to determine who are the most influential followers (in this case, who has the most connections to others in the web).
Here’s an example of a Network Graph that SetuServ produced with the data provided by twtData. In this case, it maps links between different physicians and doctors' Twitter profiles.
Below, you can find three further case studies where these measurements and metrics are put into practice.
Case Study Three - Helping Gym Brands find fitness influencers using Network Graphs
Setuserv was tasked to find brand ambassadors and influencers for the Gym Brand Company to promote their products or services. They talked about what they do for a gym brand using tools like network graphs to find the most influential people.
“Then we can use Twitter followers among gym going people to see who is the most popular celebrity or …the most popular…gym influencer that most gym going people are following.”; something which Srikanth says the Network Graphs are vital for.
“That is the goal of marketing. So, you want your information to reach as many people as possible, and this Network Graph will help us with that.”
Case Study Four - Using PharmaSignals to support pharmacists and patients
PharmaSignals is one of SetuServ’s products, for which they utilised Twitter follower data provided to them by twtData. SetuServ utilised twtData’s services to find influencers within their customers target market.
In this case, the follower and following data is used for every user in the dataset, looking at the influence that each user has by the size of their following and connections they have with each other.
This is what is shown in the graph above. The connections between each user shown by lines and the size of their following shown by the size of the nodes (the dots) on the graph.
Finding out the size of someone’s following gives companies looking for influencers an idea of who to target and a sense of the type of content being published by others. This will inform the content strategy of SetuServ’s clients. They are able to use this data to target the right audience to improve business performance through social media promotion.
Case Study Five - Using Following and Follower Information to choose and target influencers to partner with
SetuServ gave another example, discussing how they can work with a marketing company to narrow down influencers and celebrities to target by downloading and analysing followers for their client.
Harvesting “the following information of their amateur athletes and see[ing] who are... [they] following.
“Who is the celebrity that these amateur athletes are most following? Then…we can help the companies decide which brand ambassador makes the most sense”
This helps companies focus where to target their marketing budget when it comes to social media marketing and who to choose for the widest impact. This is something that could be applied to many industries that use influencers, however, in this case, there is a focus on sports.
How can we help you?
SetuServ has partnered with us, like many other Digital Marketing Agencies, to service all their Twitter data needs.Do not hesitate to contact us via email at sales@twtdata.com to discuss how we can help you.