Myntra Data Scraping Services

Cash on delivery is a purchase mechanic, and a data caveat

COD is mainstream in Indian ecommerce, and it does two things that a Western fashion dataset never has to account for.

It affects conversion independently of price

• COD availability differs by pincode and sometimes by product or price band.
• A cheaper item without COD can convert worse than a more expensive one with it.
• COD availability is a merchandising decision, so changes in it are competitive activity that price monitoring cannot see.

It changes what availability data means

COD orders carry materially higher refusal and return rates than prepaid ones. That means stock returning to availability on a COD-heavy product is even less likely to indicate genuine demand than on a Western platform.

We capture cod_available per pincode where it varies, and we state the caveat explicitly rather than letting availability changes be read as demand. On a market with both high returns and high COD refusal, an availability transition is a weak demand signal and we say so.

There is no common size scale across Indian brands

Western fashion has imperfect but broadly usable size conventions. Indian apparel does not: brands publish their own size charts with their own measurements, and the same label can mean materially different garments across brands.

Why forcing a common scale is wrong here

• An M from one brand and an M from another can differ by several centimetres in the same measurement.
• Chart availability varies — some brands publish full measurements, others publish only labels.
• A normalised scale would imply comparability that does not exist, and would then be used for size-curve analysis across brands.

We retain the published size label exactly, capture the brand's own size chart measurements where published, and deliver size_chart_available so you know which records support measurement-based comparison. We do not map to a common scale.

Where you need cross-brand comparison, measurements are the honest basis and labels are not. That means the comparison is only possible on the subset of records with published charts, and we report that subset size rather than filling the gap with a mapping we cannot defend.

Size availability oscillates, and sale events dominate

The same returns-driven pattern we handle on Zalando and Otto applies here with additional force, because return rates in Indian fashion are high and COD refusals add to them.

• Daily size-level collection is necessary; weekly sampling reports noise as signal.
• Restoration timestamps separate returned stock from genuine reorders.
• Core-size stockout flags remain the strongest demand proxy available.

Sale events add a second layer. Indian fashion runs heavy multi-day sale events where discount depth against MRP reaches levels that make MRP-based depth figures meaningless. We record sale_event and sale_phase as context, and recommend indexing against observed competitor prices rather than against MRP — MRP is a printed maximum, not a prevailing price.