The systems that manage ₹150 crore of ad spend

What the number actually means

First, honesty about the number: ₹150 Cr+ is cumulative managed spend across 160+ brands over 6 years, not one year's billing. Against it sits ₹450 Cr+ in attributed revenue and a 3.8× average ROAS. We're documenting the system here for a simple reason: at small spend, talent covers for process; past a few lakh a day across dozens of accounts, only process survives. Every practice below exists because its absence once cost a brand real money.

If you're evaluating agencies for serious budgets, use this as a probe: ask any shortlisted agency to show you their version of each system in this post. The ones who scale spend safely have written answers — the ones who don't will improvise on your money. (Our take on who handles big budgets well is in the enterprise agencies roundup.)

Account architecture: boring on purpose

Every account we run looks unfashionably similar: consolidated campaign structures instead of fifty micro-campaigns, strict naming conventions a stranger could parse in five minutes, and budget tiers — roughly 70% core proven structures, 20% scaling bets, 10% pure tests. Fragmentation is the silent killer at scale: fifty tiny ad sets each learn nothing, while five consolidated ones feed the algorithm enough signal to actually optimise. Consolidation is also what makes a ₹40L month manageable by humans.

One source of truth per brand — a single sheet that reconciles platform spend, backend revenue and the current test slate. When the dashboard and the bank account disagree, the sheet is where we find out, not the monthly review.

The other unglamorous rule: nothing structural changes on a Friday, and no more than one structural change per account per week. Platforms need stability to learn; so do the humans reading the results. Half the "algorithm broke" stories we inherit from other agencies are just three overlapping changes nobody logged.

The creative pipeline is the real media plan

In 2026, targeting is mostly the algorithm's job; creative is the lever you actually control. So creative runs as a pipeline, not a request queue: fixed weekly testing slots per account, an angle matrix (proof, problem, occasion, offer — every live account keeps all four lanes populated), a deliberate mix of UGC, statics and video, and kill rules agreed in advance so losers die on data, not on debate.

The sharpest proof this works under constraint: Kalories, an intimacy-adjacent chocolate brand Meta kept rejecting. We rewrote the website and every creative line-by-line for policy and grew it 10x in 8 months — the pipeline just kept shipping compliant angles until the winners emerged. Creative volume without a system is noise; volume with kill rules is compounding.

Measurement: blended first, platform second

Platform ROAS is directional; the blended number is the truth. Every account runs on MER (total revenue over total spend) with platform metrics as diagnostics underneath — and for COD-heavy brands, on COD- and returns-adjusted ROAS, because a dashboard 4× can be a real 2.1× after RTO does its damage. The full arithmetic is in our COD and returns ROAS post; it changes more scaling decisions than any attribution tool we've used.

The rhythm that keeps 160+ brands honest: Monday is numbers day (blended, by brand, against plan), Wednesday is creative review (what shipped, what died, what scales), Friday is pacing (budget vs month plan, festive pre-loads, platform anomalies). Weekly beats real-time: real-time dashboards invite panic edits that reset learning phases; a weekly cadence with daily guardrails invites decisions.

Attribution tooling matters less than people think at this level. Pick a reasonable stack, keep it stable, and spend the saved energy on holdout tests twice a year — a clean geo-holdout answers the incrementality question that no dashboard, however expensive, ever will.

Risk systems: assume something breaks

At this spend level, the question is never whether something breaks — an account ban, a policy sweep, a platform outage, a payment failure at 11pm on Diwali week — but whether it takes the month down with it. The standing controls: compliance pre-checks before launch for every claim-sensitive category (nutra, ayurvedic, organic), verified business structures with backup ad accounts kept warm, pacing alarms that flag anomalies within hours, and no brand left dependent on a single channel once it crosses meaningful scale — Meta plus Google plus marketplace or quick-commerce legs, weighted by category. Each control earns its place the same way: it must have prevented, or would have prevented, a real loss on a real account.

Two case studies show the system in both modes. Triage: Barosi, a dairy brand at 0.6× ROAS, rebuilt to 3.8× in 2 months — that speed comes from having a diagnostic checklist, not inspiration. Compounding: Pro Nature, certified organic, 1.2× to 8× over 10 months of the same weekly rhythm, no single hero moment anywhere in the run. Same checklists, opposite tempos.

What breaks past ₹50 lakh a month

Somewhere past ₹50L/month of spend, three new walls appear: creative velocity (audiences exhaust faster than most teams can produce), incrementality (the marginal rupee stops behaving like the average rupee, and you need holdouts or geo-tests to see it), and operations (finance, inventory and CX become marketing constraints). We've written that stage up separately in scaling past ₹50 lakh a month.

Who this post is for: founders about to hand an agency a serious budget, and operators building their own in-house system. Steal all of it — none of this is secret, it is just consistently done. The gap between knowing and doing weekly, for years, is the actual moat. If you take one thing from this post, take the Monday–Wednesday–Friday rhythm: most accounts we audit don't have a measurement problem or a creative problem first — they have a cadence problem, and everything else is downstream of it.