By Tathagata Bandyopadhyay
The Union government’s recent announcement relating to a framework of standards to curb fake reviews posted on online platforms is a step in the right direction given that it has become a menace for online shoppers worldwide. The framework, IS 19000:2022, has been prepared by the Bureau of Indian Standards in consultation with platform owners. The objective is to ensure that the published reviews are legitimate, the process of collecting, moderating and publishing them is accurate, and they are not misleading.
The question is, can such intervention stop the menace?
Goldman Sachs predicts that the Indian e-commerce market will touch $99 billion by 2024. Online purchase decisions, especially for those with no first-hand experience, depend heavily on feedback of others. E-commerce consulting firm Pattern finds that “an increase of just one star in a rating on Amazon correlates with a 26 per cent increase in sales”. Another recent research finds, “An extra half-star causes restaurants to sell out 19 percentage points more frequently.” So genuine feedback is essential for the e-commerce market to work efficiently.
A paper published in Marketing Science (February 25, 2022) reports that there is a large and thriving market for fake reviews. A large number of sellers, usually selling products with weak branding, low ratings, and inferior quality are manipulating the reviews and ratings for short-term profit. They procure these reviews via private Facebook groups. An estimated 4.5 million sellers sourced fake reviews to post on Amazon via these Facebook groups in 2019. Incidentally, the vast majority of these sellers are located in or around Shenzhen, China. As far as short-term impact is concerned, the data is clear: Fake reviews are extremely effective. In the first few weeks after fake reviews are posted, products enjoy high ratings. Once these products begin to build a positive reputation, real consumers start buying and leave genuine reviews causing the ratings to fall, and sales to fade out within a month or two.
It’s not just the buyers, the online platform also takes a hit. Fake reviews corrode consumer trust. In 2019 Amazon spent more than $500 million and employed more than 8,000 people to reduce fraud and abuse.
The study found that Amazon was deleting around 40% of the fake reviews — but it took them more than 100 days on an average to remove one after it was posted. The time window was more than enough for these sellers to hoodwink buyers.
What’s the way forward? Online platforms must create barriers for sellers who post fake reviews, and, if posted, speed up the process of deleting them. Most platforms are using machine learning algorithms to detect fake reviews, but their performance depends crucially on the training data or the data on which the algorithm is trained. If the test data, the data on which it is applied, is different from the training data, the algorithms perform poorly. The sellers are dynamically finding new ways to create fake reviews to change the test data.
Seems like a never-ending game, doesn’t it?
The author is a Distinguished Professor at the Dhirubhai Ambani Institute of Information and Communication Technology, Gandhinagar