What an app report can tell you, and what it cannot
We tried the Blinkit ads page, the Blinkit MediaMax page, the Blinkit home page and the Instamart brand portal. The first and the last opened with only a title. The other two refused our fetch tool. So this guide states no Blinkit or Instamart tool, format, fee or report. The test works without them, because it runs on questions you ask and numbers you keep.
The one app that explains its ads in public is Zepto, in its engineering blog. According to a Zepto post of 27 July 2026, advertiser dashboards read from day-level tables. A spend job runs every three minutes. An analytics job runs hourly and looks back four hours to catch late events. The post does not say which orders are counted as ad sales. It gives no counting window and no ROAS formula.
That gap matters. A ROAS in a report is a ratio of two things, and you cannot judge it until you know what sits in each. This is Zepto's account of its own system. We do not say Blinkit or Instamart work the same way. Ask each app.
For how ads, stock and brand-name search work inside Zepto, see the hub. For the ad choices on each app, see our quick commerce marketing guide. The way in to each app is in our guide on how to sell on Blinkit, Zepto and Instamart. Our field note on the Blinkit listing check lists what to fix before you apply.
Before week 1: work out your break-even ROAS
ROAS is return on ad spend: the sales credited to ads divided by what the ads cost. Our guide on ROAS meaning and formula shows the basic sum. On a quick commerce app it has one extra twist. The shopper pays the shelf price, and you keep only part of it.
Break-even ROAS = 1 ÷ what the pack leaves after the app's cut and your costs, as a share of the shelf price
This is our own method. It assumes the app's report counts sales at the shelf price. Whether it does is the first question to ask.
Say a packaged snack brand sells a pack for ₹120 on the app's shelf. The terms in this example are made up. They are not the terms of any app. Say the app pays the brand ₹78 for the pack. Making it, packing it, carrying it to the hub and setting aside something for expiry and damage cost ₹54. So the pack leaves ₹24 before ads. To keep the sums short we leave out tax. Ask your chartered accountant how tax enters your own sums.
The share left is ₹24 ÷ ₹120 = 0.20, or 20%. Break-even ROAS is 1 ÷ 0.20 = 5. Each ₹1 of ads must bring ₹5 of shelf-price sales before the ads pay for themselves.
Now change one thing. Say the report counted sales at the ₹78 the app pays you, not at the shelf price. Then break-even is ₹78 ÷ ₹24 = 3.25. Same pack, same ads, and a different line. If you compare a report counted one way with a margin worked out the other way, you will reach the wrong answer.
If you are weighing which places to sell in and what each order leaves, our guide on D2C marketing cost and metrics has the wider sums.
Weeks 1 and 2: plain words in one area
Start with the words a new shopper would type: the kind of product, the size, the flavour. Leave your own brand name out for now. According to Zepto's post of 6 October 2025, a campaign has keywords, and each keyword has a match type: exact, phrase or broad. Exact means the same words. Broad means similar meaning, as the post describes it. The post says a brand can set one bid for the whole campaign, which Zepto tunes towards ROAS, or set a bid for each keyword.
Our reading: begin narrow, so you can see which words earn. Ask whether the app gives a search terms report, which lists what shoppers actually typed. If it does, trim words that spend and sell nothing at the end of week 1. If it does not, say so in your notes, because you will be less able to tell what worked.
Keep a daily spend limit you can afford to lose. According to the Zepto post of 27 July 2026, its system must stop serving ads when a campaign's spend limit is reached. Check that your own app does the same, and what happens to the day's spend at midnight.
Change one thing at a time. If you change the words, the price and the pack photo in the same week, you will not know which one moved the sales.
Weeks 3 and 4: are you paying for buyers who would have come anyway?
A shopper who types your brand name already knows you. An ad on that search may bring a sale you would have had anyway. It may also stop a rival taking the shopper. From the report alone you cannot tell which. The test can.
According to Zepto's post of 6 October 2025, its system sorts each keyword as branded or generic. The post calls it brand defence when a campaign bids only on the brand's own product name. That is Zepto's view of its own tool. For our method, the question is simpler: if the brand words go off, do total sales fall?
In week 3 add your brand name words in area A. In week 4 switch them off again and keep the plain words running. Area B stays without ads throughout. Then compare total packs sold in A in the two weeks, after you allow for any change in B.
Say the report credits 300 packs to the brand name words in week 3. In week 4, with those words off, area A sells 60 packs fewer once B's change is taken out. Then the brand words brought about 60 extra packs, a fifth of what the report credits. The rest would likely have sold anyway. All figures here are made up.
Two cautions. Switching the words off may let a rival ad sit on your brand name. Look yourself: search your brand on a phone in area A, and note the date and what shows. And one pack in one area for one week is a small test. Treat the answer as a reading for that pack. It is no rule for every pack.
Run the same test the other way for plain words later. Switch them off for a week in one area and watch the total. Plain words are often where new shoppers come from, but your own numbers should say so.
After week 4: read the app's number next to your own
Now put the two sources side by side. Take these from the app's report, and treat them as the app's count.
- The spend, which should match the invoice.
- The sales credited to ads, and the words and places they came from.
- Views and clicks, if the report gives them.
Take these from your own sheet, which is the one you keep.
- All packs sold in area A and area B, week by week, from before the ads and through the test.
- Days the pack was in stock in each area.
- What the app pays you per pack, your costs, and an allowance for expiry and damage.
Then work out three numbers yourself. This is the made-up example from above, carried through.
Say the test spent ₹30,000 on ads in area A over four weeks. The report credits ads with ₹1,80,000 of sales at the shelf price. That is a ROAS of 6, above the break-even of 5. It looks like a win.
Now your sheet. Before the ads, area A sold 400 packs a week, or 1,600 in four weeks. Area B, with no ads, sold 5% more in the test weeks than in the four weeks before. So area A would have sold about 1,680 packs without ads (1,600 × 1.05). With ads it sold 2,400. The extra is 720 packs.
The report credits ads with 1,500 packs (₹1,80,000 ÷ ₹120). Your sheet says about 720 were extra. About half the packs in the report would have sold anyway.
The extra packs leave 720 × ₹24 = ₹17,280. The ads cost ₹30,000. So the test lost ₹12,720. Each extra pack cost about ₹42 of ads (₹30,000 ÷ 720) and left ₹24.
The idea is borrowed from a comparison group. According to Google Ads Help, October 2026 (no date on the page), a Google conversion lift study compares users who saw ads with users held back from seeing them. It defines incremental cost per action as total ad spend divided by incremental conversions. That tool belongs to Google Ads and does not run on a quick commerce app. Our sum is the same shape, done by hand with packs.
These figures are made up to show the sums. They do not show how any app's ads perform. Two limits apply. Area B is never an exact twin of area A. And the Zepto post of 6 October 2025 describes sales from ads raising a product in its ordinary results, so some of a first-month loss may buy later sales. Decide before you start how long you will wait for that, and write it in the stop rule.
| Number | Where it comes from | Why |
|---|---|---|
| Spend | The app's report, checked against the invoice | It is what you were charged. |
| Sales credited to ads | The app's report | It is the app's count. Treat it as a claim to check. |
| Total packs sold in the area | Your own sheet, from the app's sales report if it gives one | Only total sales show what really changed. |
| Days in stock | Your own sheet | The app may not report days when you had no pack. |
| Margin after the app's cut | Your own sheet and your terms | Nobody else knows your costs. |
| Break-even ROAS | You work it out | It needs your margin and the app's counting rule. |
| Extra packs from ads | You work it out | Packs sold with ads, less what you expect without them. |
Stock and ad spend belong in one sheet
According to Zepto's post of 11 June 2026, a search is answered from what is available right now at the delivery hub that serves the shopper's address. Out-of-stock products stay in the search index for a set number of days and can show as unavailable. The post does not say whether a sponsored listing keeps running for a pack that is unavailable. Ask your app.
Our rule: do not spend where the pack is not on the shelf. Keep a daily stock log for the test pack in areas A and B. Ask whether you can pause ads for one pack in one area.
Say area A was out of stock on 4 of the 28 days, and spend ran evenly. Then one seventh of the ₹30,000, about ₹4,300, went on days with no pack to sell. Drop those days from your reading, or note them, before you judge the ROAS. The hub has a measure for this, which it calls days in stock.
If stock breaks often, fix that first. The stock talk is a job for the app's category team, with your forecast in hand.
What to ask the app's team, in writing
Answers in a call are forgotten. Send these questions by email and keep the replies with your sheet. If the answer is that the app does not share something, write that down too.
- What does the sales figure in your ROAS count: the shelf price, with or without tax, delivery fees and discounts? The whole basket or only the advertised pack?
- How long after a click or a view is an order counted as sales from the ad?
- Are cancelled and returned orders taken out of the report? How soon?
- How am I charged: per click, per thousand views or another way? Is there a minimum spend, and how are the terms changed?
- Can I choose areas for a campaign, and pause one pack in one area?
- What happens to a sponsored listing when my pack is out of stock at the nearest store?
- Which reports do I get: by day, by search term, by place on the screen, by area? How old is the data?
- Do ads show on searches for my brand name, mine and my rivals'? Can I see the search terms?
- Who is my contact for questions, and how do I get an invoice I can match to the report?
We found no page we could read that gives a fee, a minimum spend or a rate for any app. Rate cards on agency sites are not official. Get the terms from the app's team, and put them in your sheet.
Mistakes that waste a test
- Judging the test by the app's ROAS alone, without total packs sold.
- Comparing a ROAS counted at the shelf price with a margin worked out on what the app pays you.
- Testing three packs in five cities at once, so nobody can tell what moved.
- Changing words, budget, price and photo in the same week.
- Reading a week before late orders have landed.
- Running ads in areas where the pack is often out of stock.
- Copying another brand's ROAS target. Their margin is not yours.
- Ending the test early after one good day or one bad day.
- Letting a test run on because the last week looked better. The stop rule is there for that.
Keep, fix or stop
Write the decision rule before week 1, so that the result cannot change it. Here is one we suggest. It is our own method.
- Keep: the extra margin from the extra packs is above the ad spend with room to spare, and stock held. Raise the spend in small steps, and run a second test on the next pack.
- Fix: the extra margin is close to the spend. Change one thing, such as the weak words or the area, and run another four weeks.
- Stop: the extra margin is well below the spend and stock held. Stop the ads, keep the learning, and look at price, pack size and listing before you try again.
Quick commerce is one place to sell among several. For where it sits in a D2C plan, read our guide on D2C marketing strategy for Indian brands. If tracking and reports are the weak part, our page on analytics and tracking describes what we set up.
