What is ONDC and how does it work? A dummy's guide
UPI moment for e-commerce?
Open Network for Digital Commerce (ONDC) is supposed to revolutionise e-commerce and level the playing field for small sellers. Is this the UPI moment for e-commerce?
ONDC is a gigantic topic where each section can be a separate ebook in itself. Here, we’ll keep things as accessible as possible, focusing on the broader picture rather than diving into technical knitty gritties.
The article is split into two parts:
Part 1 covers the problems ONDC is supposed to solve.
Part 2 dives into ONDC itself, what it is, how it works, and the challenges it faces as it evolves.
Note that when we talk about “commerce,” it includes all digital services from food delivery (like Zomato and Swiggy) to quick commerce (Instamart, Blinkit) to service delivery platforms (Urban Company).
Let’s start!
Part 1: Problems ONDC is supposed to solve
During covid, the govt. conducted research to understand the impact of the pandemic on MSMEs. The research showed that (data for the year 2020):
~80% of kirana (near and local) stores were digitally absent.
More than 90% of the supply chain (mfd., logistics, warehousing etc.) was digitally absent.
Gross merchandise value of digital retail commerce was only 4.3% of the country’s total GMV. In comparison, it was 25% in China, 26% in Korea, 23% in UK.
On the buyer side, only 20% of internet users purchased online. This number is increasing though.
Existing e-commerce players provide integrated solutions covering the entire supply chain from logistics to warehousing to last mile delivery & payments. Now as a small seller, I cannot afford to handle all the moving parts on my own coz they are capital intensive. So, I will register myself as a seller on an e-commerce platform and they will do all the grunt work for me. This is a good thing for the seller but it comes with its own risks.
Platforms play the role of kingmaker because of the sheer scale at which they operate. Case in point: Boat. Most of its sales come from e-commerce platforms. This was flagged as a business risk in its IPO filings. A change in the platform’s algorithm can affect the visibility and sale of their products.
Concentration of risk: Consolidating all digital commerce under one platform creates a single point of failure. Eg. During the initial years of big festive sales, these platforms couldn’t handle high traffic and would often crash, thus affecting the sale of thousands of sellers on the platform.
There are services like Shopify, Shiprocket and other startups which can help you host your own digital store but you will never get the kind of distribution and visibility these platforms offer.Promotes discretionary behaviour: Platform dictates the terms of engagement. As a small seller I don’t have any bargaining power.
Portability of trust: As a seller, if I want to move from one platform to another/sell on multiple platforms, I cannot transfer my ratings between platforms. I need to start from scratch on the new platform. Since I don’t have any reviews on the new platform, my sales won’t be that great.
Sellers need to maintain separate inventory for each platform. Each platform has its own terms and conditions.
Discoverability of choice: Buyer and seller need to be on the same platform to discover each other.
Detour #1
In e-commerce, there are two types of models:
Inventory based: Company purchases large quantities of goods at cheap prices and sells it directly to the end user.
Marketplace model: Platform based model where individual sellers sell their products to buyers. Platform is just an intermediary where a commission is charged per sale.
Which model is more profitable for the platform?
Let's take the example of a book seller who makes a book at a cost of INR 100.
Marketplace model: The book seller sells them for INR 400 on an e-commerce site. Let’s assume a commission of 10%, then the platform earns INR 40 on the sale of 1 book. Since the platform is incurring the cost of pick up, logistics and warehousing, there is very little scope for profit.
Inventory model: The platform buys 10,000 books from the same book seller at a price of INR 150 per book. Since the ticket size is large, the book seller would be willing to sell at a discounted rate. Seller makes 50% profit i.e INR 5,00,000 in one go. Now, if the book were to be sold at INR 400, then the platform would make INR 250 per book. There would be a fixed cost of storage but this leaves room for a lot more profit as the volume grows.
Clearly, the inventory based model is more profitable for e-commerce platforms. This is how Amazon operates in the US. If an e-commerce platform wants to make money, it has to buy bulk products at cheap prices to increase its margin.
Detour#2: Indian e-commerce regulatory landscape
The following e-commerce rules exist in India. The rules have evolved over the years.
100% Foreign direct investment (FDI) allowed in marketplace model. No FDI investment allowed in inventory model. This means that e-comm platforms cannot operate on inventory based models in India.
There was an existing rule which stated that a seller’s sale cannot be more than 25% of the total sale on the marketplace in an year. So if INR 100 worth of goods are sold on a platform in an year, a single seller’s sale cannot account for more than INR 25. This was done to prevent the platform from promoting certain sellers over others.
This was replaced by a new rule in 2018, which stated that a seller cannot procure more than 25% of its inventory from the platform or any of its group companies.
So, if I am selling 100 pens on a platform, then I cannot procure more than 25 pens from the platform’s B2B vertical. This was done to ensure that the platform does not have indirect control over inventory. The B2B arm of the platform used to sell in bulk to the vendors (sellers) on its platforms who in turn sold to the customer.An entity cannot sell on the marketplace if the platform or any of its group companies has a stake in it. This was a problem because some platforms had invested in brands which sold on its marketplace (Amazon had invested in Shoppers stop).
Merchants were prohibited from selling products exclusively on a single platform. For eg. a specific phone only being sold on one platform.
A platform has to provide all services like fulfilment, logistics, payments to all sellers in an equitable manner. It does not matter whether you are a small seller operating from a garage or a large seller who owns a gigantic warehouse. The same set of services will be provided to both (in an ideal world).
Given all this, the following questions come to mind:
Q1. How are these platforms surviving if inventory based model is not profitable (given they have FDI)?
A1. Platforms like Amazon funnel money from other profitable ventures (like cloud services), others are heavily funded. Apart from this, both Amazon and Flipkart have separate B2B arms where they sell goods on wholesale which is relatively more profitable.
Note that the 25% inventory rule (point 3) mentioned above is only applicable if you are a seller on the platform. If not, you can procure more than 25% goods from the B2B arm (based on my understanding of the laws).
Q2. How do the B2B verticals and dark stores (think quick commerce i.e zepto, blinkit, instamart) operate if inventory model is not permitted for companies which have FDI?
A2. The B2B arm does procure goods and raw material in bulk. However, these B2B arms are listed as separate entities registered in India without any FDI (via complex corporate structures). So, they can run inventory based models. Similarly the dark stores (aka mini warehouses) are owned by separate Indian entities, legally controlled by Indian residents. This kind of structure falls a bit in the grey area. The FMCG association has filled a complaint with the competition commission of India for the same.
Fun fact: Hyperpure, which is Zomato’s B2B vertical, contributes ~27% to its total revenue.
Q3. How does Amazon run brands like “Amazon basics” and “Solimo” if the marketplace is not supposed to own brands?
A3. For this, I will be directly quoting the response of the ex-country head (Manish Tiwary). Make of it what you will.
He said that “Because of FDI rules we are not supposed to have our own brands, they are either licensed or we are developing them together and are sold by a third-party seller”
“If a producer comes to us and says we want to sell on your platform, we do the due diligence and say that we will work together. There is a commercial agreement and that's how we launch the brand. We also have brands which Amazon may own worldwide which we license to a third-party manufacturer. For instance, we are giving AmazonBasics licensing to manufacturers who are great at that product and they are selling it on their own.”
Q4. How are some mobile phones only available on certain platforms?
A4. Platforms have just rejigged their contracts and changed the word “exclusive” to “preferred”. Unfortunately this kind of preferential arrangement still persists. For eg. Flipkart is the preferred partner of Nothing and Pixel phones.
Detour #3: The Reuters report
Reuters released this investigative report on Amazon India. I would highly encourage you to read the entire report. However, if you are short on time, I am sharing a condensed version here (note that the following are findings of the investigative report and does not reflect my personal opinion):
Internal documents revealed that Amazon gave preferential treatment to certain sellers. They exerted significant control over the inventory. It helped these sellers get exclusive deals with brands. 33% of the sellers accounted for over 66% of the sale on the platform. They treated “Appario” and “Cloudtail” as special merchants. Both were given privileged services. Pressure from retailers and increased scrutiny from govt. authorities led both to halt operations.
Amazon was accused of creating knock off labels where they copied best selling products, partnered with the original manufacturer to match quality, and sold them at discounted rates. For instance, Amazon’s in-house label, Xessentia, replicated the dimensions of John Miller shirts down to the exact inch.
You may have noticed this practice while shopping on Amazon, where an in-house label’s replica product is recommended at a discounted rate.
Sellers were essentially seen as guinea pigs. This is something that was mentioned in the book “The everything store” as well. Whenever Amazon wanted to venture into new verticals it would approach existing specialist retailers and let them sell on their platform. It would learn everything just by observing and then start selling the products on its own at cheaper cost.
Amazon promoted its own labels in search results via badges like “Amazon’s Choice” and placing them at the top of the search results. The in house labels were supposed to contribute 20-40% of sale in all product categories.
While individual sellers do not have access to aggregate data, Amazon used that data to replicate products and promote its own labels. Similarly, food delivery platforms can use aggregate data to promote their own in house cloud kitchen labels and optimise their offerings based on historical demand.
So, we saw how the incentives of the seller and platform are not aligned. This concludes part 1! At this point you should be able to understand what are the problems ONDC is trying to solve.
Part 2: ONDC itself
Let’s look at some basic terms we will be using in the rest of the article.
Buyer app: The app which handles the buyer side operations i.e the demand side of the transaction. This is the consumer facing app on which products can be ordered/services can be availed. Eg. PhonePe’s Pincode, Paytm, Magicpin.
How does a buyer buy from ONDC?
As a buyer, as long as you are shopping from a buyer app (like Magicpin or PhonePe’s Pincode) which is part of ONDC, you will automatically see products from all sellers on the network.Seller app: The app which handles the seller side operations i.e the supply side of the transaction. The seller can manage their orders, inventory, logistics and other services on it. Here is an example.
How does a seller become part of the ONDC network?
A seller can directly join ONDC OR they can register themself on a marketplace like Magicpin or Pincode which is already a part of the network. As soon as the seller joins the platform, they will automatically become a part of ONDC.
Types of seller appsa. Marketplace seller nodes (MSN): These are aggregators who do not hold any inventory themselves. They function as a marketplace where sellers (who are themselves not part of the network yet) can come and sell.
b. Inventory seller nodes (ISN): These are participants who joined the network as a seller and hold their own inventory.
Network: A network is a collection computers. In case of ONDC, a network is a collection of all the sellers side apps, buyer side apps, and other software services.
Node participant (NP): A node is just an entity which is a part of the network. Apart from the buyer and seller side apps discussed earlier, there can be other entities who are are part of the network. For eg. an app provides cataloging as a service where the seller needs to upload photos of their products and a catalog will be generated automatically. Similarly, there can be separate entities for dispute resolution, payment settlement on the network, each of these is a node. We will explore this in greater detail below.
Registry: A registry is a knowledge base and an address book of sorts. It contains the list of all network participants, the kind of service they are providing/products they are selling, network policies etc.
Gateway: A gateway is a piece of software that will route your request to the relevant service. Imagine it to be a trust worthy guard at a hospital whom you can ask for directions. This guard will ensure that patients go to the correct department and there is no crowd in the corridor.
Next, we will be looking at some fundamental features of ONDC.
Universal search
Traditionally, when you search for a product, you will only see sellers who are listed on the platform. With ONDC, the results show products from all the sellers who are part of the network. It is not necessary for the buyer and seller to be on the same platform.
Hypothetically, if Amazon and Flipkart are a part of the ONDC network, and a user searches for heaters on Amazon, then the search results will show heaters from sellers who are listed on either of Amazon or Flipkart. A seller does not have to register themselves with different platforms. As long as the seller is part of the ONDC network, their products will be shown in the search results irrespective of the buyer app being used.
Unbundling and Interoperability
Unbundling refers to breaking down a complex system into smaller, modular services, each of which can be independently performed by different actors. When these services are completed, they collectively orchestrate the entire workflow.
A typical e-commerce transaction flow looks like this:
Search & discovery>order>payment>order confirmation>pickup from seller>packaging, warehousing, transportation> last mile delivery
Traditionally, the entire flow was handled by the platform. With ONDC, different entities can perform different parts of this flow. For eg. you place an order on Magicpin and logistics are handled by Porter.
These micro services are interoperable with each other. Any buyer app can talk to any seller app.
If I were to give an analogy, email is interoperable i.e the recipient and the sender need not use the same client to talk via email. An email sent from Gmail can be delivered and read on Outlook. This is possible because of SMTP & IMAP/POP3 protocols. They are a set of rules which the email clients use to send and receive emails.
Similarly, beckn protocol powers ONDC’s interoperability. This is the same protocol which powers the auto hailing service namma yatri.
Note that a single entity can perform multiple steps of the afore mentioned e-commerce flow as well (assuming they have enough capital and expertise)!
We will now be looking at each step of the e-commerce value chain to understand how ONDC is different from traditional platform based model.
Search and discovery
The user searches for a product or clicks on a featured product on the buyer app.
The buyer app sends this request to the gateway (the trustworthy guard), who will check the registry to identify seller apps (aka seller nodes) who have sellers that sell the product or service matching the search criteria.
The gateway will then send this search request to all the seller apps identified in the previous step.
The seller app will query all the sellers on the platform to check if the searched product is in stock. The responses from all the sellers will be collated and shown in the search results to the buyer.
It is mandatory for all requests to pass through the gateway. The gateway must send the request to all sellers and return all the responses without any filtering.
Entities who operate the gateway are forbidden from operating a buyer app or seller app.
The buyer app can decide which all results will be shown to the end user. For eg. the product must have 3 high resolution images or the seller rating should be greater than 3. Based on this custom criteria, search results can be filtered. It is mandatory for the buyer app to publish such filtering standards so that sellers and seller apps can maintain their catalog accordingly.
Similarly, buyer app can apply proprietary algorithms to sort the sellers in the response listing. This algorithm also has to be published along with best practices for sellers to rank higher.
This ensures that all sellers are treated equally.
Any search results which are sponsored or artificially boosted need to be clearly marked.
🧠 Gateway ensures that no seller is given preferential treatment.
Placing an order
Once the search results are displayed, the buyer clicks on one of the products and receives a quotation from the seller (the quotation request originates from buyer app, goes to seller app, who relays it to the seller and the response traverses the same path in reverse). This quotation contains the product price, description, offers, discounts, minimum order quantity, logistics, delivery time, return/refund/cancellation terms etc.
After checking the quotation, the buyer adds the product to cart. The buyer will select a payment option (either online or COD) and confirm the order. Once the order has been confirmed, a transaction level contract will be created between the buyer and seller app which contains the rights and obligations of both the network participants (settlement terms, commission etc). This contract ensures trust is maintained between the two entities and the order is completed.
The response to the search listing (as seen in search and discovery section) contains a limited amount of information. The quotation contains all the details which will guide the purchase decision, hence it is the responsibility of the sellers and the seller app to maintain it accurately.
Fulfilment
Note: Whenever I am using the term “buyer”, I am referring to the person who is purchasing the product on a buyer app. If an entity is availing some service on ONDC, then an appropriate prefix will be used like “logistics buyer”.
The fulfilment stage overlaps with the ordering stage. The delivery can be handled by either the buyer app or the seller app or the seller themselves. Ideally, it should be handled by the seller app.
The seller app (or the seller or the buyer app) can either deploy their own fleet or avail logistics service from ONDC.
If logistics service is availed on ONDC, then the logic of the transaction will be identical to what we discussed in search and discovery section. There is going to be a logistics service buyer and logistics provider (seller). The provider will be a separate node in the network.
So, the logistics buyer (either the seller app or buyer app or the seller) searches for logistics providers, the logistics buyer will receive quotations (containing the price, turn around time etc), which will be sent to the buyer app who in turn will display those to the buyer (who is purchasing the product). The responses can be filtered and sorted as mentioned earlier. The buyer can then select one of the delivery options as per their convenience.
Once the order has been confirmed by the buyer, the logistics buyer simultaneously confirms the logistics order and a transaction level contract is created between the logistics buyer and logistics seller. The two orders are linked with a common reference id and necessary information like delivery address, product details etc. are passed between the two orders.
If the buyer had chosen cash on delivery payment option, then the logistics provider has to collect the payment and the terms of settlement will be mentioned in the contract (more on payment settlement later).
This logistics mechanism gives the seller apps (buyer apps and sellers) freedom to decide how they want to manage fulfilment. They can conduct business on their own terms. Even small players like a courier service provider who can only deliver within a 10km radius can also become a part of ONDC.
The logistics buyer has to ensure that delivery charges are mentioned as a separate line item in the bill provided to the buyer.
The node who is responsible for the delivery has to give the buyer the ability to check the delivery status online.
🧠 This model is much more inclusive and enables small sellers to avail logistics services without committing to high volume contracts.
Payment and settlement
The entity who collects the payment is called the collector NP (node participant). This collector NP can be the buyer app, the seller app or the logistics provider.
Now, ONDC needs two entities to handle payments: Reconciliation service provider (RSP) and Settlement agencies. Either existing entities can take up these roles themselves, or they can avail specialised services. These specialised services are standalone nodes who provide settlement services.
The role of the RSP is to prepare a settlement advice stating how much amount is due to each entity while the role of the settlement agency is to actually disburse those funds.
Once an order is placed, there are multiple participants. One of these participants is the collector i.e the entity collecting the payment while the rest of the participants are receivers (who are owed an amount). After the buyer completes the payment, the RSP will receive the settlement terms from the transaction level contracts signed between the participants. On the basis of the amount received by the collector and the amount owed to other entities (as defined in the contract) it will issue a settlement advice stating how much amount is owed to whom.
This settlement advice will be forwarded to the collector’s settlement agency, who will initiate the settlement.
To ensure the collected funds are not misused and disbursed timely, a special type of bank account is used which ensures funds collected are only paid out according to specific system generated triggers. The collector cannot use these funds for their own operational purposes. All the network participants are mandated to maintain this kind of special account.
The settlement has to be completed within 24 hrs of the settlement event being triggered.
If the seller app is collecting payments on behalf of the sellers, then the sellers will receive their payments as per their terms of engagement.
Returns and refunds
Returns and refunds will be accepted or rejected basis the terms of sale as mentioned in the quotation and transaction level contract.
The logistics for return work in a similar fashion to fulfilment, but the flow will be inverted. The seller/seller app will place the logistics order for reverse pickup if availing on network services else they will handle it on their own.
To ensure transparency, the seller has to disclose the terms of returns, refunds and cancellations as a part of the quotation. These terms are also coded in transaction level contract. The buyer will only transact with the seller if the terms of sale are to their mutual liking.
Issue and grievance management
ONDC has 3 levels of managing complaints and queries.
Any filed complaint starts as an issue. In the first level of dispute resolution, the network participants are expected to solve the issue internally within 24hrs of the issue being filed.
If the network participants are unable to reach a satisfactory conclusion, the issue is escalated to a grievance. Each network participant is supposed to have a designated grievance redressal officer (GRO). The GRO of the network participants are supposed to evaluate the grievance, establish liability and reach a conclusion within 7 days of the issue being filed.
If the GROs are not able to reach a conclusion, the grievance is escalated to a dispute. The dispute is handled by separate network participants who provide online dispute resolution (ODR) services like mediation, conciliation, arbitration.
If the dispute still remains unresolved, the affected parties can approach a court of law.
The network participants are required to provide a mechanism/interface to file and track issues.
Some notes
All the back and forth communication between different entities will happen in the backend. The customer shopping for a product will have a smooth experience on the frontend and they won’t have to switch between multiple apps.
All the nodes (buyer apps, seller apps, reconciliation & dispute resolution etc) are free to charge a fee for the services they are providing. The pricing structure has to be published and is included as a part of the transaction level contract.
Other network wide initiatives
Common taxonomy
Product taxonomy is a system to classify products. A common taxonomy built on commonly accepted standards will ensure the buyer intent is captured accurately and relevant products from all sellers are shown. This will also enable sellers to migrate to a different seller app seamlessly, thus increasing the bargaining power of individual sellers.
Cataloguing as a service
Cataloguing is a repository of all SKUs along with all the relevant details including images, description, variants etc. When a search request is broadcasted, the seller responds with data from its catalogue. A high quality catalog improves discoverability as well as reinforces buying intent. Conversely, a bad catalog might make the buyer lose trust in the seller.
For small sellers, maintaining a high quality catalog can be an arduous task. Cataloging as a service will be provided by tech partners who will be responsible for maintaining and updating the catalog.
Scoring and badging system
A lot of online purchases are driven by seller and product rating. The scoring and badging system will rate sellers, seller apps, buyer apps and logistic providers. This scoring will be done by an agency (similar to a credit rating agency) using a standard formula accounting for their past scores and dispute resolution history. The score will be an indicator of trustworthiness of the network participant.
This scoring system will be network wide and portable. If a seller wants to migrate from one seller app to another, then they can do that without losing its ratings.
Onboarding to ONDC
A participant must adhere to all the network policies. Technical compliance will be ensured through a certification process during onboarding. Participants will be subject to audits and recertification when any updates are made.
Community governance
ONDC will make certain data sets public like the quality of dispute redressal mechanism to incentivise participants to perform better and seek feedback from the community. ONDC will be governed by a user council consisting of node participant representatives who will review policies, conduct audits and shape the future of the platform.
Data policy
Transaction data, user’s personally identifiable information will only reside with the buyer and seller app and not be visible to ONDC. It won’t be storing any transaction data itself.
These two slides from ONDCs consultation and strategy paper do a great job of summarising what is ONDC.
Challenges
Big e-commerce players will continue to give deep discounts (by virtue of funding or funnelling money from profitable arms) which a small retailer can still not compete with. Govt. won’t extend incentives to sellers on ONDC indefinitely.
If enough e-comm players do no join ONDC (since they see it as competition) the the size of the marketplace will naturally reduce, thus reducing seller incentive.
There are too many moving parts which may lead to increased complexity. In case of returns, refunds and complaints there might be lack of clarity and accountability. Whom does the buck stop on? The platform based model might be better than a decentralised network for dispute resolution.
The network requests between multiple entities will have to be optimised to provide a seamless user flow. Even a lag of few seconds can lead to significant drop off rates. A platform based model would be able to provide a superior user experience.
And that is the end of the article! If you have made it this far, thanks a ton!
It took time and effort to write this article. If you want to support my writing, you can buy me a coffee here.
Sources
This article is the result of extensive research from the following sources:
https://www.reuters.com/article/india-ecommerce-explainer-idINKCN1PP1XS
https://www.reuters.com/investigates/special-report/amazon-india-rigging/