Break the ratio worship first. ROI (or ROAS) divides return by spend — a fraction that peaks where spending is smallest, because the first rupees always find the lowest-hanging demand: the in-market buyers, the brand searchers, the easy retargets. Each additional rupee reaches slightly colder prospects, so marginal return falls. Diminishing returns isn't a market failure — it's what successfully working down the demand curve looks like. The profit-maximising budget therefore accepts declining ROI on purpose, spending until the marginal rupee just clears its cost of capital. A CMO maximising ROI and a CFO maximising profit are optimising different shapes — the ratio and the area — and the confusion between them quietly sets marketing budgets across the economy. (The economists formalised this in 1954 — Dorfman and Steiner — which makes the ongoing confusion less forgivable.)
Marginal thinking then solves allocation, the question budget meetings actually fight about. Every channel has a response curve — sales as a function of spend, rising and bending (the curves come from mix models and experiments — the measurement page is the instrument shop). The equilibrium rule: allocate until marginal returns equalise across channels. If the next rupee in search earns ₹1.20 and the next in video earns ₹2.10, money moves to video regardless of their average ROAS rankings, because averages describe the past position on the curve, not the next decision. This one rule dissolves most allocation politics: the argument stops being 'whose channel is better' (a question about averages and egos) and becomes 'whose curve is steepest at current spend' (a question with an answer).
Budget-setting methods now grade themselves. Percentage-of-sales: circular — sales cause budget, so growth ambitions inherit last year's demand (and downturns cut spend exactly when share is cheapest). Competitive parity: outsources your strategy to rivals' averages. Affordability: marketing as the residual — whatever's left after everything real. Objective-and-task: build the budget from goals and the tasks that achieve them — the defensible method, and the only one that can justify a number the anchoring ratchet didn't produce. Overlay ESOV as the strategic check: the long-running finding (Jones, refined by the IPA school) that brands whose share of voice exceeds their share of market tend to grow share, roughly in proportion to the excess — voice as territory purchase. Honest label: correlational, category-variable, and noisier in digital-first categories — but as a sanity check on whether a budget is sized to grow, hold, or donate share, it has no equal one-liner.
The horizon split is where sophisticated budgets most visibly differ from naive ones. Activation — performance media, promotions — converts existing demand: effects arrive in days and decay in weeks, and dashboards see them perfectly. Brand building — reach, distinctive assets, memory structure — creates future demand: effects compound over quarters and years, arriving partly as 'cheaper activation later' (higher base conversion, stronger branded search, price tolerance), which short-term instruments structurally misattribute or miss (the metrics page's warning, now with a budget line). Binet and Field's synthesis of effectiveness cases put the average optimum near 60:40 brand-to-activation — swinging by category, price point, and lifecycle, so treat it as a prior, not scripture. The governance failure it prevents is predictable and universal: brand spend is the line whose absence doesn't hurt this quarter, so it's cut first and mourned later — at the bottom of downturns, where the recovering evidence (from Depression-era cereal wars to modern IPA case data) keeps suggesting the brands that held or raised voice while rivals cut bought share at clearance prices.