Akshat Dua
CTO,
Showroom B2B
The fast-fashion model was built around a simple promise: bringing trends to market faster. The ability to spot a trend, secure production and get a garment onto the shelf ahead of competitors is what separated fast fashion from the traditional design-to-retail cycle. That advantage still matters. But across a decade of watching buying calendars, open-to-buy decisions and vendor lead times move between retailers and manufacturers, one pattern has become hard to ignore: nearly every serious player in the value chain has gotten faster. Speed has become table stakes rather than a point of differentiation. The businesses pulling ahead now are not the ones with the shortest lead times but the ones whose merchandising and sourcing decisions are grounded in the most accurate reading of demand.
The challenge has become more difficult to solve than it was five years ago. Consumer preference in India varies sharply by region, climate and local buying calendar, so a category performing well in one market can underperform in a neighbouring one with a similar demographic profile. Social platforms compress the life of a trend further, pulling a silhouette into demand and out of it within weeks rather than a full season. Yet the sourcing calls behind most products – fabric booked, factory capacity blocked, dye lots confirmed – are typically locked in eight to twelve weeks ahead of the season, based on forecasts anchored largely to the previous year’s sell-through. Demand has not just grown more volatile; it has fragmented across geographies and channels faster than most planning cycles were designed to handle.
The Shift from Forecasting to Demand Intelligence
Forecasting in apparel has traditionally worked backward: take last season’s sell-through by SKU, adjust for growth targets, and assume the market will repeat itself with minor variation. This held up reasonably well when consumer behaviour was more uniform and retail was concentrated in fewer, larger formats. It breaks down quickly in a market as layered as India’s, where a single retail chain may be planning assortments across hundreds of micro-markets, each with its own fabric preference, price band and festival-linked buying pattern.
Demand intelligence is built to close that gap. Instead of relying on a single input, historical sales, it draws together live sell-through, store-level inventory movement, regional buying patterns, weather data, price elasticity, and digital demand signals into one continuously updated view of the market. The difference is not cosmetic; it changes what the data is used for. In a conventional planning model, data explains what has already sold. In a demand-intelligence model, the same data becomes an input into the next production and allocation call, while there is still runway to act on it, rather than a postmortem on the last one.
This is already visible across manufacturer-retailer networks operating at scale. A supplier able to see, in near real time, that a category is overperforming in one region while underperforming in another is not refining a forecast for the next cycle; it is reallocating fabric, production capacity and floor space within the current one. That is a materially different capability from forecasting as the industry has practised it, and it is the capability the next phase of fast fashion will be organised around.
The Cost of Late Information
The scale of what is at stake deserves to be stated plainly. Industry estimates routinely place unsold and excess apparel inventory at 20-30% of global production, and India’s fragmented, multi-tier retail structure adds to that complexity rather than easing it. Crucially, this waste does not start with unsold stock sitting in a warehouse or a season-end markdown rack. It starts three months earlier, at the point a fabric order is placed, a production run is sized, and a buy is committed on incomplete information. By the time a garment is discounted or written off, the resources behind it, cotton, water, dyeing capacity, labour, and freight, have already been spent. Improving demand visibility upstream, at the sourcing and production planning stage, is one of the few levers that reduces waste at its origin rather than managing it after the fact. This is as much a margin conversation as a sustainability one, and in a category running on thin retail margins and high SKU growth, that distinction matters more than it might elsewhere.
India’s Position in This Shift
Much of the current conversation around AI-driven forecasting in apparel is framed around large Western retailers and their supply chain technology stacks. That framing overlooks where the more consequential version of this problem actually sits. India’s apparel sector combines significant manufacturing depth, much of it in small and mid-scale units, with one of the most regionally fragmented consumer bases anywhere. A national chain sourcing across Tier II and Tier III markets is, in practical merchandising terms, managing several distinct assortment strategies under one balance sheet, shaped by differences in climate, income levels and regional style preference.
This fragmentation has typically been treated as a cost of operating in India, something to be absorbed through buffer stock and wider size-colour-style matrices. It is, in fact, a meaningful advantage for building demand intelligence: the more variation there is in regional buying behaviour, once properly captured, the more precise and transferable the resulting demand model becomes across categories and seasons.
The infrastructure gap, however, is real. Retail sell-through data, manufacturer capacity data and supplier lead times still largely sit in separate systems on separate update cycles, where they are digitised at all. A retailer’s point-of-sale signal from last week may not reach the factory floor until the next planning meeting, if it reaches it at all. Closing that gap, not by adding another reporting layer on top, but by getting sell-through, inventory and production data to move on a shared clock, is the less visible infrastructure work that demand intelligence actually depends on. AI and machine learning add value here, but their contribution is in recognising patterns once the data is flowing consistently, not in compensating for data that is missing, delayed, or fragmented across the chain.
What Comes Next for the Industry
None of this replaces the judgment that makes fashion buying what it is: a merchandiser’s read on an emerging trend, a founder’s conviction in a new silhouette, the willingness to bet ahead of the data occasionally and be wrong. Demand intelligence does not remove that uncertainty, and it should not try to. What it does is compress the distance between the moment demand shifts on the ground and the moment sourcing, production and inventory decisions respond to it, a distance that has historically run to months rather than weeks.
For an industry that has spent two decades optimising how fast it can produce, the more useful question now is how early it can know what to produce. The real advantage will not come from predicting every trend correctly. It will come from making smaller initial commitments, identifying winners earlier, scaling them faster and limiting exposure when demand does not materialise. Manufacturers and retailers who build that visibility into their sourcing decisions, rather than treating it as a reporting add-on after the fact, will be the ones setting the pace for Indian fashion’s next phase of growth, not by outrunning demand, but by staying ahead of it.
