“For us, what is important is that we gauge the sentiments confidently…So that (the sample size) is something we would not be able to give you in a tangible or sizeable way,” he told ThePrint.
“If I have to give you an example, we would definitely have around 3,000 to 4,000 samples, but it won’t come in one sort of form. We don’t work in traditional forms,” he added.
The BJP won 48 of 70 seats in the Delhi assembly elections. The AAP secured 22, while the Congress drew a blank.
Moreover, its data wasn’t limited to the polling day but was collected over two months—December and January—leading up to the assembly election on 5 February.
The Hyderabad-based company—which entered the political forecasting space with the 2019 Andhra Pradesh elections—says it collects data through on-field interviews, tele-calling, interactive voice response or IVR, WhatsApp surveys, and Artificial Intelligence (AI) based sentiment analysis on social media platforms.
The organisation has a website and KK, who used to be a software employee, has a YouTube channel.
Unlike many exit polls that are commissioned by the media or political strategists, KKS’s exit poll for the Delhi assembly election was internally funded. “We want to make this a national brand, so we wanted visibility,” KKS’s spokesperson explained.
From former journalists and engineers to lawyers, MBA graduates and corporate trainers, exit pollsters in India are a mixed bag of professionals, stepping into the spotlight in the crucial hours before parliamentary and state assembly election results are announced.
However, exit pollsters operate in a legal grey area, with few guidelines on disclosing sample sizes and the methodology used to support their predictions. While experts advocate for these polls to be used as research tools to interpret the final mandate, exit polls have transformed into media spectacles.
Another organisation that released its exit poll numbers for the Delhi assembly elections is Journo Mirror, run by Mohammad Ali and Shahid Ansari, who met during their Master’s programme at the Delhi School of Journalism.
The Delhi-based Journo Mirror predicted 45 to 48 seats for the AAP and 18 to 20 seats for the BJP.
Their sample size? Just 1,000 people, surveyed by a team of around 10 reporters and students. The data, Ali told ThePrint, was collected through in-person interviews, telephone surveys and ‘group discussions’ over some time, not just on the day of voting.
Ali and Ansari started the organisation, along with a few others, while they were still pursuing their Master’s degrees in 2020. While Journo Mirror is primarily a news website, it also began releasing exit poll numbers from the general elections last year, when they claimed to have predicted 240 seats for NDA.
KKS and Journo Mirror, along with several other pollsters, were both widely quoted by TV channels on 5 February, after the voting for the Delhi assembly elections concluded.
However, none of these news agencies mentioned their sample size—which ranged from 1,000 to 3 lakh—or their methodology or funding. They also refused to share their raw data with ThePrint.
Moreover, many pollsters who released their predictions after the Delhi assembly elections did not have an online presence. Without a website or a social media page explaining the origins of these companies, there was little to back the credibility of their numbers and surveys.
A look at 10 companies that released exit poll numbers before the Delhi assembly election results were announced tells a similar story.
The industry is cluttered with pollsters mushrooming by the minute, with new names cropping up with each election cycle. However, they raise more questions than answers about their survey methodology, expertise and funding disclosures.
Some don’t believe in sample sizes, others work with parties and even conduct their surveys before the polls.
Also read: What was the centerpiece of BJP’s perfect plan to win Delhi? A slogan that disappeared after LS polls
Curiouser and curiouser
Mind Brink Media was one of the few agencies that predicted the likelihood of the AAP retaining power in Delhi. It gave 44-49 seats to the AAP, 21-25 to the BJP, and 0-1 seat to the Congress.
The company does not have a social media footprint. But according to ZaubaCorp, a corporate database, the company was incorporated in 2019 and lists two directors, one of whom is Navneet Kumar.
Kumar, who worked with ETG Research in the past, confirmed he was associated with Mind Brink but refused to respond to any queries on the organisation or its poll methodology.
He told ThePrint the company tried its best to gauge the mood of the voters.
“But clearly, we were off from reality on the ground, suggesting that somewhere our methodology and approach was wrong,” he said in a message.
“I personally don’t see a point in discussing a failed methodology and process. We will keep trying to improve our methodology in future and maybe if we are able to get it right in future, we can discuss it,” the message added.
The funding funda
Media houses often commission exit polls, collaborating with agencies that conduct them. However, for pollsters and their agencies, election surveys aren’t their sole source of income.
For instance, Yashwant Deshmukh, founder-director of CVoter, says election and exit polls don’t even account for 10 percent of the firm’s work.
The organisation conducts socio-economic research, like any other market research or social research company, he told ThePrint.
“Election research just gets out of proportion coverage in the media, that’s why people know about it,” he added.
The veteran TV journalist also told ThePrint that several pollsters work for political parties, and they pass on the data at the end of the polling to news channels for public visibility.
However, unlike Western democracies, these associations between pollsters and political parties remain undisclosed in India, he said.
“That is a significant difference between India and Western democracies. In the latter, pollsters would say from day one that they are working for the Republicans or with the Democrats. In India, you have pollsters moonlighting with political parties, and then come on TV only to gain some street cred,” he explained.
Take CNX, a Noida-based AI-powered public opinion survey company, for instance, which predicted 10 to 19 seats for the AAP and 49 to 61 for the BJP.
Its co-founder, Bhawesh Jha, told ThePrint that while the company doesn’t work directly with political parties, it was commissioned by an election strategist/consultant to track the Delhi election campaign, which is a process that typically starts three to four months in advance.
The 2010 Bihar election was his first project. Before 2019, the company was initially called CNX Media, focusing on delivering exit and opinion polls to news channels but was changed to CNX Data Lab when they expanded their portfolio to include a variety of services such as in-depth election campaign surveys, real-time governance tracking, and market research.
“In 2008, US Presidential elections were going on when I got an opportunity to work on research through the Florida University. Since I was a student of mathematics, I got interested in such data research. And I always had an inclination towards politics…That’s how I got into this,” Jha recalls.
According to Jha, CNX’s Delhi exit poll predictions were based on a refined sample size of 21,000, with 300 samples collected from each Assembly constituency. It had 270 surveyors on the ground, conducting computer-assisted personal interviews.
The company has four or five in-house analysts, but Jha says he conducts the final analyses himself.
CNX was working with India TV before the 2024 Lok Sabha elections for exit polls, but since then news channels have stopped commissioning such polls, Jha said. As a result, CNX has been releasing its exit poll numbers on social media.
“News channels have been using these numbers for their ‘poll of poll’ segments. Given that we are considered one of India’s leading pollsters, it is our responsibility to disclose our exit poll numbers,” he added.
IITians, latitudes and longitudes
Then there’s the Hyderabad-based SAS Group, or Sri Atma Sakshi Group, which also released an exit poll on the Delhi elections—but it has no website or social media presence.
It predicted 38 to 41 seats for the BJP and 27 to 30 seats for the AAP.
“We don’t give a lot of importance to publicity. I don’t go to any TV channel studios,” Murthy B.V., one of the company’s founders, told ThePrint.
“But we have been conducting surveys for the past 14-15 years. Political parties and politicians approach us for surveys,” he added, while recalling predictions the company got right in the last few years.
Registered last year as Sri Atmasakshi Data Management Services Private, the company was founded by eight IITians, including Murthy.
The company’s Delhi exit poll was funded by political parties and political leaders who approached the group to conduct surveys in a few of the seats. The money received through those surveys then helped fund the entire exercise, Murthy told ThePrint.
Murthy claimed his company took 1,200 to 1,300 samples for each Assembly constituency in Delhi, putting the total sample size for 70 constituencies at over a whopping 84,000.
All interviews are conducted in-person by 190 surveyors or “enumerators”, collectively in a pre-poll survey drive and an exit poll survey drive.
“All our surveyors go house to house, and send the data to Hyderabad online. They record videos, pictures, latitudes, and longitudes related to the survey. Everything is available,” he said.
He said these enumerators were graduates picked through an interview process. Seven analysts then analysed this data after eliminating “problematic samples”. “The quality and variety of the sample is more important than the number of samples,” he explained.
Murthy insists exit polls alone don’t provide correct results, and companies must also conduct pre-poll surveys first.
“We divide the entire state into four zones, and we conduct exit polls in a certain number of constituencies… Exit polls can only be conducted within a duration of 8-10 hours. Each Assembly constituency has around 300 polling booths. How is it possible for anybody to cover it all?” Murthy retorted.
“So, people who talk about exit polls, it’s always bogus. Without conducting a pre-poll survey, you cannot come to a conclusion in exit polls,” he claimed.
Journalists-turned-psephologists
Many agencies in the poll forecasting business operate in near anonymity.
Take Mumbai-based Poll Diary, which predicted 42 to 50 seats for the BJP, 18 to 25 for the AAP, and 0 to 2 seats for the Congress. It only has an X account that says, “We cover every poll on the Earth.”
An archived version of its now-inactive website listed Abhijit Bramhanathkar as the company’s editor. Bramhanathkar—who was a journalist for 15 years before he switched to psephology—told ThePrint he established the company in 2018.
He claimed that the Delhi elections exit poll was based on a sample size of 24,000, collected mostly through on-ground surveys by around 250 to 300 people. The organisation has 24 in-house analysts.
While the Delhi exit polls weren’t commissioned, Poll Diary regularly works with political clients as political strategists.
“I use a lot of different types of methods to arrive at my numbers. I track social media and speak to people personally as well, so my predictions are usually closer to the results. And I’ve been spot on in the last three elections—Lok Sabha, Maharashtra and Delhi,” Bramhanathkar told ThePrint.
He isn’t the only journalist who has taken to psephology.
Manoj Kumar Singh was a journalist for 23 years, before he started his own company, Matrize News Communications, in 2016. His first project came in 2017 when he managed the campaigning for the Congress before the Gujarat elections.
For the Delhi Assembly elections this year, Matrize predicted 32-37 seats for AAP and 35 to 40 seats for the BJP, claiming that it had a sample size of 37,000.
Singh told ThePrint that around 140 surveyors worked on the ground over six months to conduct opinion polls in Delhi, and eight surveyors per assembly constituency conducted the exit polls. These surveyors were a mix of college students and call centre employees—all of whom were trained to conduct these surveys.
“When we conduct opinion and exit polls, it’s not necessary that our opinion is only based on the sample of 37,000. We conduct exit polls to see whether there is any variation in our opinion over the last six months… A base is formed through opinion polls and then analysed together with the exit polls,” Singh explained
Eleven people organise this fieldwork, while the editorial team at Matrize comprises five analysts.
The Matrize exit poll was internally funded, Singh told ThePrint.
“We want to stay visible in the media. So, we conduct these polls to register our presence in the industry…This is a business. We work for government policies, with several political parties and companies. But we want publicity too because people need to see how accurate we are,” he explained.
‘Passion projects’
For a few pollsters in the industry, these surveys represent a deep interest in the politics of the country. Their polling agencies, therefore, seem like passion projects.
One such company is People’s Insight, which registered last year, months before the 2024 general elections.
The two founders are cousins—Sandeep Gupta and Nikhil Gupta. While Sandeep has an MA in Political Science, Nikhil’s expertise lies in artificial intelligence and machine learning.
Talking to ThePrint, Sandeep claimed that People’s Insight is the first company to introduce AI and tech into predictive analytics in India.
“All our surveys are done through APIs. There is zero human intervention during the survey process,” he explained.
They claimed that their sample size for the Delhi elections was a whopping 8,09,723, which included the exit poll sample of 3,11,724. They predicted 40-44 seats for the BJP, 25-29 seats for the AAP, and 0-1 seats for the Congress.
The company had 12 surveyors on the ground to validate the data points received through AI-based IVR calling. The company has two analysts and two data scientists who then look at this data.
“The weightage for exit polls is 50 percent—20 percent weightage is for samples collected one week before polling, another 15 percent in the one week after candidate announcement, and the final 15 percent is from the month before election announcement,” he explained.
The brothers are already established businessmen, running software companies.
“I was always interested in politics, so this is more of a passion for us,” he said.
Another organisation that sees polling as a passion project is Peoples Pulse which collaborated with another company called Codemo, and predicted the most seats for the BJP at 51-60.
Its website doesn’t list the names of the founders or the team members.
“We maintain a low profile. We believe that only our organisation should be in focus, not any individual,” Ravichand, founder of Peoples Pulse, told ThePrint.
The organisation was established in 2012. “We were into student politics during our college days…We began this during the Anna Hazare movement.”
While their core team in Hyderabad is only five or six people, they take the help of local political science students in that state, whenever conducting a survey.
For the Delhi elections, Peoples Pulse claimed to have a sample size of approximately 6,500 to 7,000, with around 30 people on the ground and around 35 people conducting telephone interviews, according to Ravichand.
He told ThePrint that Codemo provided logistical and technical support for its telephonic survey.
He said that while media houses provided some funding for the survey, the company released such data before every election to stay relevant in the market.
However, polling isn’t the bread and butter for Ravichand, who has a law degree, and his colleagues at Peoples Pulse. The organisation is registered as a society, and its members belong to professions such as law and journalism.
“We are doing this only for passion. We are not doing it as a business,” said Ravichand, while acknowledging that Peoples Pulse has conducted exit polls for media houses in the past.
Third parties and corporate trainers
Another company that operates at the intersection of polling and strategy is WeePreside.
The Pune-based company predicted that AAP was likely to get 46-52 seats, the BJP 18-23 and the Congress 0-1.
WeePreside’s founder Sujit told ThePrint his company has been acting as a third-party research agency to give vote share numbers to other organisations that are commissioned by news channels for exit polls.
“We also work with political parties and politicians, but they don’t contact us directly. So, for example, parties hand over campaign work to a company, which then gives us market research work to develop their strategy,” he explained. Additionally, the company also works with other market research agencies, for research.
While its website is just a page saying “Launching Soon”, the company conducted its first exit poll in 2021 for the West Bengal and Kerala elections.
For the Delhi Assembly election exit polls, WeePreside had around 400 surveyors waiting near polling booths. They claim to have a total sample size of 55,000 and a weighted sample size of 29,000, collected by 470 people on the ground. A team of eight analysts based out of Pune then looked at the sample.
A data analyst, Sujit said that he worked with Prashant Kishor’s I-PAC as an intern in 2017 during the Uttar Pradesh assembly elections, which then led him to develop an interest in political research.
Apart from journalists, data analysts, IITians and software engineers, there’s also a corporate trainer and author in this mix of pollsters in India.
JVC, headed by J.V.C. Sreeram whose X profile describes him as a psephologist who “got LS 2019 99.66 percent accuracy”, predicted 39-45 seats for the BJP in the Delhi election.
He told ThePrint he is a corporate trainer and an author who conducts training sessions for corporates. He claims he used his network of agricultural institutions and chartered accountants to predict the 2019 Lok Sabha results.
The Lok Sabha prediction he made on X is still his pinned post on the social media platform.
“In 2019, we just started with a network of people across the country, who used to give me local feedback. Based on that, I arrived at a figure and gave my numbers two hours before exit polls… It almost came exactly the same. That is how my journey started,” Sreeram recalled.
He told ThePrint that while he has people speaking to voters on the ground and recording their responses on Google Sheets through an app, he primarily does “qualitative research”.
“My team does quantitative research, while I do qualitative… Whatever we release in the press is all dipstick. It’s a small sample size, a small questionnaire, just asking them who they voted for… But when I do it for politicians, we have a bigger sample size, because we have more resources,” he explained.
The origin story
While election polls aren’t new, exit polls seem to have gained prominence in the 1980s and 1990s.
“Pre-poll surveys have a long history, but exit polls got into prominence because of the TV news, and satellite TV news coming up in the late-90s,” Deshmukh told ThePrint.
One of the earliest documented exit polls in India is attributed to Dr Prannoy Roy and Ashok Lahiri in the 1980s. In 1996, another exit poll was conducted by Doordarshan, backed by research by the Centre for the Study of Developing Societies (CSDS).
CVoter, set up in 1994, was another player. Its founder, Deshmukh, a graduate of the Indian Institute of Mass Communication in Delhi, recalls conducting exit polls in the mid-90s when the primary clients used to be the print media and newspapers.
(Edited by Sugita Katyal)
Also read: Kejriwal era of ideology-free politics ends. Contradiction of being in INDIA & targeting Congress base
