TL;DR
- Fixed splits (60/20/20 and friends) answer the wrong question. Allocation follows from per-platform contribution per rupee — and because take rates differ, the same SKU has a different break-even ROAS on each platform.
- Blinkit leads on GMV share (~46%), Instamart runs strongest in the South, Zepto skews younger and metro — but your category's reality on each platform beats every generalisation. Two weeks of your own data outranks any published comparison.
- Rebalance monthly on margin-adjusted marginal return, move 10–15% of budget per step, and mind cross-platform cannibalisation: three apps, considerable overlap in the same pin codes.
Every multi-platform brand eventually asks for the magic percentages. The honest answer is that any fixed split is folklore — the right allocation falls out of arithmetic you can do with your own settlement reports, and it shifts as your data comes in. Here's the arithmetic.
What actually differs between the three
| Axis | Blinkit | Swiggy Instamart | Zepto |
|---|---|---|---|
| GMV share (2026, approx.) | ~46% | ~27% | ~21% |
| Geographic centre of gravity | Delhi-NCR, Mumbai, Bengaluru, Pune | Strongest in South India | Mumbai, Bengaluru, Hyderabad; younger skew |
| All-in take rate (typical range) | 22–32% | 26–36% | 20–28% |
| Ad platform maturity | Most formats, deepest tooling | Improving; leans on Swiggy ecosystem | Aggressive pricing to win ad share |
Ranges are indicative — rates vary by category and negotiation, and the platforms revise them often. The point survives the imprecision: the widest spread between platforms is the take rate, not the CPC — and take rate moves your margin on every order, while CPC only moves the cost of a click.
Same SKU, three break-evens
Per-platform break-evenbreak-even ROAS (platform) = 1 ÷ contribution margin % after that platform's take
| ₹450 SKU, 40% COGS | Take rate 22% | Take rate 28% | Take rate 34% |
|---|---|---|---|
| Contribution before ads | ~38% | ~32% | ~26% |
| Break-even ROAS | 2.6× | 3.1× | 3.8× |
| Reported ROAS needed at a 33% attribution haircut | ≈ 3.9× | ≈ 4.7× | ≈ 5.7× |
Read the last row twice. A campaign reporting 4.5× is comfortably profitable on the cheap-take platform and underwater on the expensive one — identical reported performance, opposite verdicts. Any allocation method that compares platforms on reported ROAS without normalising for take rate is comparing apples to invoices.
The allocation loop
- Step 1 — floor each platform honestly. Compute the break-even table above with your real category rates. A platform whose break-even exceeds what your category plausibly achieves shouldn't get a "fair share" — it should get a small test budget or nothing.
- Step 2 — seed by fit, not by share. National GMV share says little about your shelf: a South-heavy brand may find the ~27% platform its best market. Start where your category and cities overlap the platform's strength; use published comparisons only to pick the starting line.
- Step 3 — measure contribution per rupee, monthly. (settlement GMV × contribution margin after take) ÷ ad spend, per platform. Rank. This number already absorbs CPC differences, conversion differences and take-rate differences — it's the whole scoreboard in one line.
- Step 4 — rebalance 10–15% per month toward the leader. Big swings destroy the campaign learning and the week-on-week comparability you need for step 3. Slow money compounds; fast money thrashes.
- Step 5 — watch the cannibalisation tell. Consumers multi-home across these apps in the same pin codes. If platform A's GMV rises exactly as B's falls while total holds flat, you're paying ads to move your own sales between shelves. The check is the total line across all three, monthly.
When concentration beats diversification
Under ~₹1.5–2L/month of total q-commerce ad budget, splitting three ways buys you three statistically unreadable experiments. Concentrate on the single platform with the best break-even for your category, prove contribution, then expand with evidence. Diversification is a luxury of budgets big enough to learn on.
FAQ
Shouldn't I be everywhere for discoverability?
Listing everywhere is cheap and usually right. Advertising everywhere is expensive and usually wrong before you have per-platform contribution data. Separate the two decisions.
Festive season — shift budget to the biggest platform?
Festive demand lifts all three; CPCs rise everywhere too. The break-even math doesn't pause for Diwali — if anything, recompute it with festive CPCs before authorising the bump everyone requests in October.
How do I get city-level performance?
Platform reporting varies — where city or dark-store splits exist, use them; where they don't, your settlement reports plus delivery pin codes reconstruct a usable approximation. Imperfect geography beats no geography.
A platform is offering ad credits to shift budget. Take them?
Free money is fine — just book the uplift honestly. Credits change this quarter's arithmetic, not the platform's structural take rate, and the allocation loop should resume its verdict the month the credits stop.