ROAS-focused media planning: from plan to weekly execution

Most media plans are written to spend a budget. A ROAS-focused plan is written to earn a return, and then to spend more wherever that return holds.

That sounds obvious, but in practice it changes almost every decision: how you split the budget, which campaigns get protected, what you report, and how often you are willing to move money. It also changes the conversation with finance, because every line in the plan can be traced back to the margin it has to earn.

This is the system I've refined across roles at Google, as a CMO and as a growth leader, and in my consulting work today. It's how Fresh Greens ran at 12:1 ROAS while building €2.8M in recurring revenue on €750K–€1.2M of annual ad spend, and how Del Monte's ROAS moved from 3.4x to 8.3x in about six months.

Step 1: start from unit economics, not channel habits

Before any channel mix, I need three numbers: contribution margin per order (or per qualified lead), the break-even ROAS that margin implies, and the target ROAS that leaves room for profit and growth. If a product carries a 40% contribution margin, break-even ROAS is 2.5x; anything below that loses money on the first order, however good the platform dashboard looks.

• Break-even ROAS = 1 / contribution margin.
• Target ROAS = the level that funds your growth goals after fixed costs, agreed with finance.
• For subscription or repeat-purchase businesses, plan on first-order ROAS and on payback period, so you can accept a lower first-order return where repeat revenue justifies it.

Step 2: split the budget by intent, not by platform

A blended ROAS hides everything that matters. I split every plan into at least three intent layers and give each its own target:

• Brand: people already searching for you. Usually your highest-ROAS spend, and the spine you protect.
• Non-brand: category and problem searches. Lower ROAS, but this is where new customers come from.
• Competitor: bidding on rivals' names. Often the leak, and it has to earn its place with numbers.

On Blinkit for Priya Foods, this split was the whole story: brand keywords ran at 33–46x ROAS, non-brand around 5x and competitor terms at 0.7–1.0x. The blended April result was 3.74x on ₹5.07L of spend, which hides both the strength of brand and the cost of competitor bidding. Once the split was visible, the plan wrote itself: protect brand, keep non-brand for new-to-brand volume, and cut competitor spend that couldn't pay for itself.

Step 3: build the plan around tests, not certainties

Every plan has a core and an edge. The core is spend with proven returns. The edge is a fixed share of budget, typically 10–20%, reserved for structured tests with a clear hypothesis, success metric and end date. That keeps the account learning without risking the whole return on unproven ideas.

1. Write the hypothesis: "Shifting 15% of non-brand to a new audience will hold ROAS above target while adding new customers."
2. Define the metric and threshold before launch.
3. Run for long enough to exit the learning phase and collect meaningful conversions.
4. Decide: graduate to core, iterate once, or kill.

Step 4: weekly execution, the cut-or-scale rhythm

Plans fail in execution, usually because nobody moves money fast enough. I run a fixed weekly rhythm:

• Monday: one table with spend, revenue and ROAS by intent layer and platform, against target.
• Cut list: lines below break-even for two consecutive weeks with enough data to be sure.
• Scale list: lines above target with headroom (impression share, audience size, inventory).
• Creative and landing-page flags: when media is efficient but conversion is the bottleneck, say so early.
• Budget moves executed the same day, logged with the reason.

The same discipline works at platform level. In quick commerce, for example, I compare platforms separately rather than blending them: for Priya Foods, Blinkit ran around 7.1x overall and BigBasket at 5.86x, while Zepto needed a fix (3.1x in one week, 1.9x the next). That's a rebalancing decision you can only make when each platform has its own line.

Step 5: scale without breaking ROAS

Scaling is where ROAS usually collapses, because the cheapest conversions are bought first. A few rules keep it under control:

• Raise budgets in steps (for example 15–20% at a time) and let algorithms re-stabilise between steps.
• Watch marginal ROAS, not average ROAS: what did the last block of spend return?
• Expand reach before raising bids: new keywords, audiences, placements or platforms.
• Look for growth that doesn't need more money. At Digital Upshot, we scaled affiliate and programmatic-native acquisition by 60% without extra spend, while delivering up to 5x ROAS.

Step 6: measure what the business actually earns

A ROAS plan is only as good as its numbers. Platforms over-report, attribution windows differ, and refunds or unqualified leads never show up in the ad account. I reconcile platform data with backend revenue weekly, and for lead generation I optimise towards qualified outcomes, not raw leads.

For PNB MetLife, rebuilding campaigns around conversion quality took lead-to-sale conversion from 0.58% to 3.67% and cut CPA by 77.48%. For Margadarsi, focusing on lead quality kept junk at 12.9% and lifted hot-lead-to-sale conversion from 2.11% to 5.00%.

Under the hood that means server-side tracking, a single source of truth where spend and revenue meet, and automated anomaly flags, so decisions are based on the same numbers the CXO sees.

Common mistakes I see in ROAS plans

• Judging every campaign against one blended target, so brand subsidises waste and non-brand gets cut for not matching brand.
• Reporting platform ROAS to leadership without reconciling it against backend revenue.
• Changing bids, budgets and creatives all in the same week, so nobody knows what caused the result.
• Treating the plan as fixed for the quarter instead of a starting point that the weekly review keeps correcting.

A simple ROAS plan template

1. Unit economics: margin, break-even ROAS, target ROAS, payback assumptions.
2. Budget by intent layer (brand, non-brand, competitor) and by platform, each with a target.
3. Core vs test allocation, with the test backlog and success thresholds.
4. Weekly rhythm: owner, report, cut and scale rules.
5. Measurement: source of truth, reconciliation, attribution rules agreed with finance.

Where to start

If your blended ROAS looks fine but growth has stalled, or if it looks poor and nobody can say why, the answer is almost always in the split. My Complimentary ROAS Audit breaks your spend down by intent and platform, maps where it leaks, and gives you a clear scale-or-cut list.