Manyavar says algorithms take 70-75% of its retail decisions

Manyavar has moved its retail decision making onto data. Inc42 reports that the ethnic wear brand no longer relies on traditional garment naming. Each product is broken into granular subcategories covering colour, texture, fabric, design type and numeric codes. That structure feeds an allocation engine which decides what each store receives, based on local demand patterns.

An executive surnamed Modi told Inc42 that “70-75% of decisions are taken by algorithms, and the remaining edge cases are taken by AI.” The company says about 850 million historical data points, collected over two decades, sit on a Model Context Protocol server behind the system.

The allocation is not static. When a design underperforms in one store and shows potential in another, the system moves stock between the two automatically. Manyavar claims this redistribution clears an extra 7% of inventory. It also claims revenue above Rs 1,000 Cr, margins above 65%, a 90% sell-through target, and dead stock of about 3% which it donates rather than discounts. Those are company claims, not audited accounts.

What operators should take from this

Ignore the model and look at the assortment number. Manyavar says only about 35% of inventory by design is common between its South Delhi and Karol Bagh stores. Those two stores sit roughly 12 km apart in the same city. If a third of the mix is all that two neighbouring stores share, one national assortment is indefensible for most brands.

You do not need 850 million data points to start. Pull twelve months of sell-through by SKU by store, or by pincode if you sell online. Rank each SKU within each location, not nationally. Any SKU sitting in the bottom quartile at one location and the top quartile at another is a transfer candidate. Run that list monthly and actually move the stock. Then change the buy. Stop ordering one common assortment and start ordering against the location ranks.