TL;DR
- The launch decisions that matter happen before day zero: contribution per unit after the take rate, break-even ROAS, ad ceiling, and the kill thresholds — written down while you're still objective.
- Weeks 1–4: narrow exact-match launch, search-term mining, controlled expansion, then a margin-based verdict. Boring, and it front-runs the two classic failures — scaling on reported ROAS, and quitting on noise.
- Pre-commit the day-30 rules: real contribution ROAS ≥ break-even scales, within 20% below holds another 30 days, further below kills. Deciding after seeing the numbers is how sunk cost writes your media plan.
Every platform onboarding deck tells you how to start advertising. Almost nothing in the ecosystem tells you what result should make you stop — which is not an oversight, given who writes the decks. Here's a first month designed around the decision at the end of it, with the exit criteria set before the first rupee moves, while you're still capable of objectivity.
Day 0 — the gates (no spend yet)
- Gate 1: unit economics. From your signed rate card: all-in take on your category and ticket, contribution per unit after COGS. If contribution before ads is under ~25% of MRP, fix price point or pack size before advertising — ads amplify economics; they don't repair them.
- Gate 2: break-even ROAS, written down. 1 ÷ contribution margin. This number goes in the tracking sheet next to every reported figure for the next 30 days.
- Gate 3: listings ready. Stock in the dark stores you'll advertise to, fill rate healthy, images and titles complete, ratings not broken. Paying for traffic to a listing that can't convert is the one waste with no redeeming discovery value.
- Gate 4: the exit rules, pre-committed. Write the day-30 scale/hold/kill thresholds now (last section). Sunk cost is the strongest bidder in month two.
The four weeks
| Week | Do | Don't |
|---|---|---|
| 1 · Narrow launch | 3–5 hero SKUs with the best unit economics; exact-match on product and top category terms; modest budget (₹500–1,000/day is a common floor); log daily spend, orders, and stock-outs | No broad match, no reach formats, no festive-week launches — you need a readable baseline |
| 2 · Mine | Pull the search-term report; kill-rule the zero-order terms (3× contribution-per-order, no orders → pause); note converting queries you didn't bid; check price-point mismatches | Don't judge overall ROAS yet — one week of a new account is noise wearing a trend costume |
| 3 · Expand deliberately | Promote proven queries to exact; a small discovery budget (~10–15%) on close variants; add next SKU tier only if week-1 SKUs hold; first look at city-level splits | Don't raise total budget past the ad ceiling from Gate 1, however good reported ROAS looks |
| 4 · Verdict prep | Settlement-report GMV (not dashboard) × contribution margin ÷ total spend = real contribution ROAS; compare to Gate 2; apply the attribution haircut from your branded/total-line sanity checks | Don't let the platform's quarter-end account manager set your month-two budget |
Day 30 — the pre-committed verdict
The month-two rulereal contribution ROAS ≥ break-even → scale +25–50% · within 20% below → hold & fix · further below → kill or restructure
- Scale: real (settlement-based) contribution ROAS clears break-even → raise budget 25–50%, never past the ad ceiling, and re-verdict monthly.
- Hold: within 20% of break-even → one more 30-day cycle focused on the specific leak (usually keyword mix or a wrong-ticket SKU), same budget, no expansion. One cycle — a hold that renews itself monthly is a slow kill you're paying for.
- Kill or restructure: more than 20% under break-even after a clean month → stop, and diagnose at the unit-economics level, not the keyword level. Wrong ticket size, wrong category-commission fit, or a platform whose take rate your margin can't carry. Re-enter later with different SKUs, or don't — a channel that fails the arithmetic isn't a growth channel, whatever its GMV chart says.
The two classic month-one deaths
Death by enthusiasm: scaling in week 2 because reported ROAS hit 5× — before settlement data, before the haircut, past the ceiling. Death by impatience: killing in week 2 because reported ROAS was 1.8× on eleven orders — a sample that decides nothing. Both are cured by the same medicine: verdicts only on pre-committed thresholds, only at day 30, only on settlement numbers.
FAQ
Which platform should I launch on first?
The one where your category's break-even ROAS is lowest — usually the platform with the friendliest take rate for your category, adjusted for where your cities are. One platform for the first 30 days; the comparison loop comes after you have one clean baseline.
Is ₹500–1,000/day enough to learn anything?
At typical q-commerce CPCs it buys a few thousand clicks a month — enough to run the kill rule on your top terms and get a directional contribution read. What it can't do is support week-2 verdicts; thin budgets need the full 30 days.
Should I launch during a sale event for the extra traffic?
No — festive CPCs and shifted shopper behaviour contaminate your baseline, and month two inherits targets set in a distorted month. Launch in a boring fortnight; join the festivals once you know your normal.
What do I actually log daily?
Five columns: spend, attributed orders, attributed GMV, stock-outs on advertised SKUs, anything you changed. Settlement lines join weekly. The sheet takes four minutes a day and is the difference between a day-30 verdict and a day-30 vibe.
The account manager offered a launch package with reach banners. Take it?
Reach formats (banners, spotlights) are brand spend judged on different math — fine for launches with budgets that can afford unmeasurable outcomes. In a measurement-first first month they add spend you can't attribute and noise you can't remove. Performance placements first; buy awareness once the unit is proven.