Consider this: The potential size of the online retail market in India, according to media reports, was $225 billion in FY 2021-22 which is about 27 per cent of the estimated size of India’s retail sector and is expected to grow at 32 per cent year on year over the next few years.
India’s retail sector, on the other hand, is likely to grow at 10 per cent. With dynamic pricing, inventory management and optimisation of delivery cost as its pillars, the Open Network for Digital Commerce (ONDC) seeks to transform digital commerce and democratise large scale participation of businesses.
This can potentially increase e-commerce penetration in the country from the current levels of 8 per cent to 25 per cent in the next two years. So, how does the growth in online retail sales prevail over that of brick and mortar even though the later offers several advantages such as look and feel, inspection of quality, testing before buying, easy search, instant delivery and faster refunds?
The answer lies in several features of online shopping such as shopping at any time of the day, eliminating the need to drive to a store, customer feedback for products and sellers, search powered by Natural Language Processing, instant contextual help through virtual assistants and checking prices from multiple websites before buying.
Of the above features, the most important, in our opinion, is customer feedback and rating of sellers. Rating provides shoppers with a better idea of a seller’s reputation necessary for an informed purchase decision. Sellers can analyse customer feedback data, supplement their product descriptions with additional information and react to customer preferences. ONDC, which is a network of networks, will have a reputation ledger to record and track the reputation of sellers.
Questions, therefore, naturally arise about the extent to which ratings of sellers are objective, particularly when the majority of consumers rely on product reviews and ratings for e-commerce purchases. Stars ranging from one to five are often used as symbols for rating products, sellers, movies, rides, hotels and restaurants. No break-up of the components constituting the overall rating is provided in B2B and B2C e-commerce.
Better rating metrics
One can argue that a star rating by a customer represents a true, overall consolidated summary of his or her entire online shopping experience and this should be the sole driver of the rating of a seller to the exclusion of any other metric. But a survey by LocalCircles has pointed out that almost 65 per cent find product ratings to be positively biased, likely indicating that sellers may be influencing opinion for their products.
The same survey also points out that 58 per cent of consumers said e-commerce websites don’t publish their negative ratings and reviews. Also, since reviews are typically a significant factor in search rank algorithms and thus have a big impact on product visibility and sales, these systems often also create powerful incentives for sellers to manipulate their products’ rankings through fake reviews.
So which reviews should a buyer believe in? Should he/she rely solely on intuition and discretion to arrive at a decision? Or can the rating system for a seller be made more data-driven and, therefore, more objective?
One possible approach is to identify the components of rating which are data-driven and therefore objective. Metrics for the following three components can be obtained directly from the transaction level data on the platform:
(i) Timely delivery: Actual time of delivery vis a vis the scheduled or “promised” time of delivery;
(ii) Reliability: Orders accepted by a seller as a ratio of total orders placed on the same seller;
(iii) Quality of order fulfilment: Quantity accepted by buyers on delivery as a ratio of the total quantity delivered by that seller.
The other two components of rating, namely,
(iv) Buyer’s feedback and
(v) End- users’ feedback are subjective assessments.
End-users’ feedback will be relevant in cases where end-users are distinct from buyers. Buyer’s feedback and end-users’ feedback can capture the holistic experience with other non-quantifiable components such as packaging, return/replacement policies, product description and support provided by the seller during pre and post-purchase orders in addition to the quantifiable parameters.
It turns out that by deploying advanced statistical techniques and ML algorithms, the weights of all the above components constituting a rating can be learnt objectively from data thus obviating the normally prevalent tendency of assigning weights to components subjectively. It also turns out that all these five parameters are necessary to adequately describe a seller’s reputation. Moreover, the breakup of each rating component along with the overall rating can be provided unlike popular e-commerce sites where only the overall rating is provided.
Aiding MSMEs
This will enable a seller to focus on each component of the rating so as to improve and catch up with the competition. The number of transactions and the cumulative value of transactions till date can be displayed along with the number of ratings and customer reviews. All these make the rating system more data-driven, transparent and objective as three out of five components are obtained directly from transaction-level data on the portal.
The system of rating can be improved further by working through the following four key levers:
One, routine, standardised products of day-to-day use such as pencils may be associated with a weighting scheme different from customised high-value items such as laptops.
Two, the system of ratings must factor in both the value and frequency of transactions as a seller might ensure high quality/delivery/reliability in low-value products associated with a high frequency of transactions while, at the same time, getting away with poor performance in a few high-value transactions.
Three, the rating system must provide a level playing field by being size agnostic so that sellers with smaller size balance sheets and lower turnover have an equal opportunity to grow alongside the larger ones.
Four, making recent transactions the basis for deriving weights will provide a more informative picture of the latest trends in a seller’s rating.
A fair, transparent and robust rating system has several advantages: first, it enables the e-commerce platform to grow while protecting the interests of buyers; second, it allows the rating of a seller to be used as a filter to select a smaller set of sellers fairly and impartially; third, it incentivises sellers to continuously improve their operations; and fourth, it provides equal opportunities to MSMEs to grow alongside the larger sellers.
Kumar is Member (Finance), Space Commission and former CEO, Government eMarketplace. Das is with Indian Institute of Foreign Trade, Kolkata. Views expressed are personal.