Indian online retail has a structural problem that no advertising tool built in San Francisco was designed around: a large share of your orders are not orders yet. They are intentions, and some meaningful fraction of them will be refused at the door. Every return-on-spend number you look at is inflated by that fraction, and the platforms cannot see it. Everything else in this article follows from that one fact.
Why cash on delivery breaks your reporting
Your purchase event fires at checkout. At that moment you know somebody clicked confirm. You do not know whether they will be home, whether they will still want it, or whether they ordered from three sellers to compare and will take whichever arrives first.
Meta and Google receive that event and optimise toward people who behave like the person who sent it. So the system learns to find you more people who place COD orders — which is a genuinely different population from people who pay for them.
What you see
A healthy return on ad spend in the dashboard, campaign after campaign, with a clear winner emerging.
What you have
A bank balance that does not match, because the difference walked out as return-to-origin. The platform never learns this happened.
What gets worse over time
The optimisation compounds in the wrong direction. Each week it gets better at finding the audience that orders and refuses.
💡 This is the one thing to fix before evaluating any tool. Send a second event back to the platform when an order is actually delivered and paid for, and optimise toward that instead of toward checkout. You will see your reported return fall sharply. That is not the campaign getting worse — it is the number becoming true for the first time.
The honest limitation: this needs volume. Delivery confirmation arrives days after the click, and below a certain number of confirmed orders a month there is not enough signal for the platform to optimise on. If you are small, treat RTO as an operations problem — prepaid discounts, order confirmation calls, address validation — rather than something the ad platform can solve.
RTO is partly a targeting problem
Worth saying because most sellers treat it purely as logistics. Certain audiences refuse deliveries at much higher rates than others, and you are choosing those audiences with your ad targeting.
- 1Impulse-heavy creative produces impulse orders, and impulse reverses. A discount-led ad to a broad audience will show excellent conversion and poor delivery.
- 2Tier-3 and rural pin codes typically carry higher refusal rates. That is not a reason to exclude them — it is a reason to measure them separately rather than averaging them into one number.
- 3Prepaid orders from the same campaign behave completely differently. If your store offers both, look at your prepaid share by audience segment. It is one of the most useful signals available to an Indian seller and almost nobody looks at it.
Your own store or the marketplaces?
A question Western tool comparisons never raise, and it changes what advertising is even for.
Marketplace-only
Amazon and Flipkart have their own ad systems and their own logic. Meta and Google advertising pointed at a marketplace listing is hard to attribute and usually disappointing — you are paying to send traffic into someone else's conversion funnel.
Own store, marketplace secondary
This is where Meta and Google advertising earns its place. You own the pixel, the customer data and the margin, and every tool in this category is built for it.
Both, seriously
Keep the advertising separate and judge separately. Blending marketplace and own-store revenue into one return figure hides which channel actually pays.
The festive quarter changes the rules
From late September through Diwali, auction prices rise and consumer behaviour shifts. Three practical consequences:
- 1Your September cost-per-acquisition target is not the right target in October. Holding to it keeps you out of the auction during the weeks that decide your year.
- 2Learning phases are expensive in a hot auction. Build and warm your campaigns in August and early September so they enter the festive period already optimised rather than learning at peak prices.
- 3RTO typically worsens during heavy discounting, because discount-driven orders reverse more. Watch delivered revenue, not booked revenue, or you will scale into your own losses.
Language on the product page, not just the ad
Regional-language creative is increasingly standard among Indian advertisers. Regional-language product pages are not, and that mismatch is expensive.
An ad in Tamil landing on an English product page loses a share of the people it just persuaded — and they are the people who responded to the ad most strongly. If you cannot produce translated product pages, at minimum ensure the price, delivery time and return policy are legible, because those are the three things a hesitant buyer looks for before committing.
Which tools fit an Indian D2C seller
Catalogue automation
Madgicx, AdScale. Built around clean revenue data, which COD complicates. Genuinely powerful once you are sending delivered-order events rather than checkout events — and misleading before that.
Creative at volume
Predis.ai is built in India with real Indian-language coverage, which matters for the script rendering problem more than most international options.
Campaign builders
Zeely, Adyft and similar. The right starting point if you are not already running structured campaigns, and the right answer for a store with a small catalogue where a product feed adds little.
If you run thousands of SKUs, feed management becomes its own discipline and a general campaign builder is the wrong tool — that is worth saying plainly rather than pretending otherwise.
The order to do this in
First
Send a delivered-and-paid event back to the platform. Until you do, every other number is describing orders that partly do not exist.
Second
Split reporting by prepaid and COD. Treat them as two businesses, because in terms of unit economics they are.
Third
One well-funded campaign on your best-margin product, left alone for a fortnight.
Fourth
Only now consider catalogue automation, and only once it has clean delivered-revenue data to work from.
Campaigns built for how Indian customers actually buy
Adyft builds Meta and Google campaigns into your own ad account, in rupees, with carousels that send each card to its own product. From ₹1,999/month, 14-day free trial.
Start free — no card needed