Windows

Month 1 runs from a brand's onboarding date to day 29; month 2 runs day 30 to day 59. Windows are anchored to each brand's own onboarding date, so they do not align to calendar months and no two brands share a window. That removes seasonality as a shared confounder but means a festive period lands inside different windows for different brands. Every brand is compared only against itself.

How each objective was tested

• Scaled. Revenue up on month 1, with blended ROAS no more than 15% below month 1. A tighter 10% floor would qualify 23 brands rather than 23.
• Optimised. Blended ROAS up 10% or more, with revenue no more than 10% below month 1. The revenue condition is what separates optimising from retreating: a brand that halves its media spend and loses 40% of its revenue posts a higher ROAS without having optimised anything, and does not qualify here.
• Health, held stable. At least one of AOV up, new-customer share up, cancellation rate down or RTO rate down, with revenue and ROAS both inside ±10%. Cancellations and RTO are measured as rates, not counts, so a brand cannot qualify by simply selling less. The ±10% band is deliberate: at ±5% only 4 brands clear it and all of them already qualify on another objective, which would make the third test redundant.

Anonymity

Brands are identified by sequential number only. No category, vertical, product type, region or other descriptor is published, and participating brands are not named individually or collectively. Within each group brands are ordered by scale, so a number indicates relative size within that group and nothing further.

Revenue, ad spend and average order value are published as indices against each brand's own month 1 rather than as currency amounts. An absolute AOV or revenue figure is close to a fingerprint in a market this well-observed, and none of the three objectives depends on the level — only on the direction and size of the change between the two windows. Cohort figures are reported as sums across all 35 brands, where no individual contribution is recoverable.

Definitions and data

Figures come from each brand's own BooleanMaths account via get_dashboard_metrics. Revenue is gross booked revenue including cancelled and RTO orders, so revenue, orders, AOV and RTO rate share a single denominator and RTO is reported as visible leakage rather than netted away silently. Blended ROAS is total revenue over total ad spend across all channels. Attribution is linear touch at order time on stitched journeys, with Asia/Kolkata day boundaries. All figures are in Indian rupees and no currency conversion is applied anywhere in this study.

Limits worth stating

• Zero RTO generally means no shipping integration rather than flawless delivery. Where that is the case the RTO signal is silent, not positive, and the brand can only qualify on another health metric.
• 3 brands have no month 1 cancellation or RTO baseline to compare against. Those cells are shown as unavailable rather than assumed to be zero.
• Percentage growth off a very small base overstates the result. Two brands in the scaled group grew from under 100 orders in month 1. Both qualify on the stated test, but the percentage is not comparable to a brand transacting thousands of orders a month.
• New-customer share fell across most of the cohort between the two windows as repeat purchasing began to compound. That flatters month 1 blended ROAS and makes some efficiency gains look smaller than they are. A new-customer ROAS cut would separate genuine efficiency movement from a shift in customer mix.
• This measures correlation over a brand's first two months on the platform, not causation. Onboarding coincides with other changes inside a business, and no counterfactual is available.