The Multicast Difference: Why an ONDC Catalog Entry Isn't Competing for a Search Slot the Way an Amazon Listing Is

Two different contests for the same word, "discoverability"

When a business owner hears "get discovered online," the mental model is almost always Amazon's or Google's: a search happens, an algorithm ranks candidates, and a position on the results page is won or lost. That model is accurate for those platforms, and it explains why so much of small-business marketing effort goes into reverse-engineering a ranking signal nobody outside the company can see in full.

ONDC, the Open Network for Digital Commerce, is often described to merchants using the same vocabulary, "get discovered," "improve your ranking," "show up in search." The vocabulary is borrowed, but the mechanism underneath it is different in a specific, documentable way. Understanding that difference matters less as trivia and more as a practical question: what should a small business actually do to be found on a network that does not work the way Amazon does.

What a private index does

A marketplace like Amazon operates a single search index that it owns and controls end to end. When a buyer searches, Amazon's ranking system, popularly known by shorthand terms like A9 and its successor, scores every matching listing against relevance and performance signals and returns a ranked list. Only sellers who clear that internal bar, however it is weighted in a given moment, appear where a buyer is likely to look.

The stakes of that control are not hypothetical. In September 2023 the Federal Trade Commission and seventeen state attorneys general sued Amazon, alleging the company illegally maintained monopoly power in the online superstore and online seller-services markets. Among the specific practices named in the complaint was an internal system the FTC called the Featured Offer Disqualification algorithm, which, according to the FTC, could make a seller ineligible to win the "Buy Box," the default purchase option on a product page, if that seller offered a lower price somewhere else on the web, effectively suppressing their visibility on Amazon regardless of relevance to the buyer's query. Amazon has disputed the FTC's characterization and said it discontinued the tool years earlier. The case remains in litigation. The relevant point for a small seller is structural, not a verdict on guilt: on a private-index marketplace, the operator sets and can change the rules of visibility unilaterally, and a seller has no independent path to a buyer that bypasses that index.

What the ONDC protocol actually specifies

ONDC is not a marketplace. It is a set of open, government-backed protocol specifications, built on the Beckn Protocol, that let independently owned buyer apps and seller apps interoperate over a shared network, the way email lets Gmail and Outlook exchange messages without either company owning the other's users. The mechanics of search on that network are written into ONDC's own governance documents, not left to a proprietary black box.

ONDC's Network Policy defines the role of a "Gateway" in exact terms: a Gateway is "the technology provider that will ensure discoverability of all Sellers in the ONDC Network by multicasting the search request received from Buyer App to all Seller App and vice-versa," matched against criteria such as location, availability, and other buyer preferences. Multicasting, in networking terms, means the request goes out to every matching recipient at once, not to a single winner selected from a ranked list. The same policy framework requires Network Participants to "conduct in a fair, reasonable, and non-discriminatory behavior and not provide preferential treatment to one participant over others in the same situation," a rule that has no equivalent obligation on a private marketplace operator.

The Beckn Protocol's own technical specification describes the same flow from the engineering side. When a buyer app, the "BAP," does not already know which seller apps, the "BPPs," to query, it sends the search to a Beckn Gateway, which looks up matching providers in a registry and "may then broadcast this request to multiple BPPs that match the context of the request." Each seller app responds independently with its own results, which the gateway routes back to the buyer app. Nothing in that flow scores one seller's response against another's to decide who gets broadcast to in the first place; the broadcast happens on the basis of matching criteria, not competitive bidding for inclusion.

Why this is reaching real MSMEs, at real scale

This is not a paper architecture. As of a written reply to the Lok Sabha on 9 December 2025, the Ministry of Commerce and Industry stated that more than 1.16 lakh (116,000-plus) retail sellers were live on ONDC across more than 630 cities and towns. A year and a half earlier, a written reply to the Rajya Sabha in July 2024 put total sellers and service providers on the network at 5.7 lakh (570,000-plus) across 71 seller apps, 22 buyer apps, and 16 logistics providers, and described monthly transactions growing from just over 1,000 in January 2023 to more than 9.9 million by June 2024. ONDC's own blog, citing an Equirius Securities report, put the network past 10 million monthly transactions that same June, with a single-day peak of 374,000 orders during the 2024 T20 World Cup final.

The government has also built a dedicated onboarding push for small enterprises specifically. The Ministry of MSME's Trade Enablement and Marketing, or TEAM, initiative, running FY 2024-25 through FY 2026-27 and implemented through the National Small Industries Corporation, targets five lakh (500,000) micro and small enterprises onto ONDC, with half reserved for women-owned businesses, funding catalogue preparation, account management, and logistics support through seller network participants. As of 31 December 2025, the Minister of State for MSMEs reported around 2.06 lakh merchants had completed at least one transaction under the scheme, with roughly 6 lakh service providers onboarded and about 3,127 MSMEs each recording over one lakh transactions in the current financial year. ONDC's own published case material shows what this looks like at the smallest scale: a farmer-producer organization in Jammu and Kashmir, Baba Sankari FPO, reported over 100 orders from across India within two months of onboarding, and a Manipur food business, Kanglei Foods, described moving from local sales to regular monthly orders from multiple states after gaining ONDC-linked logistics support.

None of this is a claim that every seller on ONDC is thriving, or that multicast discovery is a substitute for a good product and reliable fulfillment. It is evidence that the protocol's structural difference from a marketplace index is operating at a scale involving hundreds of thousands of real small sellers, not a demonstration project.

Where the limits sit

Three limits deserve to be stated plainly rather than glossed over. First, multicast inclusion is not the same as being chosen. A buyer app still sorts and displays the responses it gets back, and a seller with a thin or poorly structured catalog can be technically included in a broadcast and still be effectively invisible in how a given buyer app renders the results. Second, the network's growth figures come from government written replies and ONDC's own reporting; they are credible primary and official sources, but they are not the product of an independent third-party audit, and different replies to Parliament have used different scopes, "retail sellers" versus "sellers and service providers," that are not directly comparable across dates. Third, the fairness the Network Policy requires is a governance commitment enforced through ONDC's own compliance and grievance mechanisms, not a technical impossibility of favoritism; a policy promise and an architectural guarantee are not identical, even when the architecture makes the promise easier to keep than it would be on a black-box index.

The useful comparison, then, is not "ONDC is fair and Amazon is not." It is that the two systems put the contested step in different places. On a marketplace, the contest is for a rank inside one company's index, adjudicated by criteria the company sets and can change. On ONDC, the contest that matters most is whether a seller's catalog data is structured well enough to be correctly matched and returned at all, a legibility problem more than a competition problem.