The structural flaw in supply-push agriculture
India's fresh fruit and vegetable (F&V) chain runs in the wrong direction, and it always has. A farmer plants what did well last season. A commission agent aggregates whatever shows up at the mandi that morning. A trader moves the pile downstream to anyone who will take it. Price gets settled dead last, after the crop is already harvested, graded, packed, and trucked. That is supply-push. Demand isn't a starting input; it's something you go hunting for once the produce already exists.
The problem is that fresh produce rots. A crate of tomatoes isn't a shelf-stable SKU you can hold for a better week; it loses value by the hour. Push that crate toward demand nobody confirmed and the gap turns into waste, distress sales, or a flat rejection at the door. Post-harvest losses in Indian horticulture run in the mid-to-high double digits across the chain, by most industry estimates. That's not an accounting rounding error. It's value that lands on someone's margin.
Quick commerce made the whole thing impossible to ignore. Dark stores run tight assortments and they're strict about quality. They don't want whatever the mandi sent. They want a specific grade, a specific pack size, at a specific store, inside a specific window, tomorrow and the day after and the day after that. Supply-push can't hold up to that, for one plain reason: it never knew the requirement before the crop was committed.
What 'demand-led' actually means
Demand-led flips the order. You capture the demand signal first, before you lock any sourcing, and you let it run upstream into what you buy, how you grade, and what you process. The chain gets pulled by orders you actually have, not pushed by a harvest you're hoping to place.
Those signals reach us through a few doors. Quick-commerce and modern-trade buyers publish their input requirements: SKUs, grades, volumes, delivery windows, usually a few days out. Our white-label direct-to-consumer app picks up advance orders and household-level consumption. Institutional buyers commit to standing volumes. None of it is a forecast in the horoscope sense. It's forward information, and once you aggregate it you know what to source before you source it.
This changes the day-to-day completely. A demand-led operator sizes each morning's buy against a confirmed order book instead of a hunch. We know the grade split we have to hit, so farm-gate sorting gets aimed at that split. We know which dark stores need which pack, so we pre-allocate at the hub instead of untangling it afterward. In every case the information shows up before the produce does, which is the whole point.
- Quick-commerce and modern-trade requirements, published before the delivery window
- Our D2C app, which captures advance orders and what households actually consume
- Standing institutional commitments that set a baseline we can count on
- Years of demand curves, so the prediction models have something to forecast against
Dark-store rejection: the hidden P&L killer
Rejection is the most brutal line in fresh fulfillment, and you don't really see it until you're running volume. A dark store turns away a delivery, because the grade is off or the pack is wrong or the produce showed up already halfway through its shelf life, and the supplier eats all of it. You've already paid to source it, move it, process it, and position it for last mile. Now you either dump the rejected lot into a secondary channel for pennies or you write it off.
Supply-push keeps rejection rates high by design. You're matching a speculative crop to a hard requirement after the fact. The grade split almost never lines up. The freshness window is tight. And there's no lever upstream to fix the mismatch once it exists. Every rejection is a margin hit, and it repeats daily.
Demand-led kills rejection at the source. Because we know the spec before we buy, produce gets graded to it at the farm gate, graded again at the hub, and only the conforming units go into the fulfillment stream. The order book says how much of each grade goes where, which is exactly the mismatch that drives rejection in the first place. This isn't a warehouse-efficiency win. It's just knowing the requirement before you commit to the crop.
Waste and shrinkage: turning a wasting asset into a managed one
Shrinkage is the loss of saleable weight and value between the farm and the plate, and it's the fresh sector's standing tax. It comes from over-buying, mis-grading, handling damage, and plain time. Supply-push makes all four worse because it starts from not knowing what demand actually is.
A demand-led chain chips at shrinkage from several sides at once. Sizing the buy to a real order book cuts over-procurement. Grading early sends the non-conforming produce into value-added processing, cut fruits and vegetables mostly, instead of leaving it to rot as unsold whole stock. Lot-wise yield tracking shows how much saleable output each incoming lot actually threw off, so the next buy is smarter than the last.
The compounding is where it matters. Every point of shrinkage you avoid drops almost straight to gross margin, because you already paid for the produce. In a business where net margins are thin to begin with, shrinkage control is the line between scaling profitably and growing the top line while the cash quietly leaks out.
Why investors assign higher multiples to demand-led systems
Nobody pays a premium multiple for revenue itself. They pay for how good that revenue is: how predictable it is, how defensible. On both counts a demand-led system reads better than a supply-push trader.
Start with predictability. An operator with a visible forward order book can forecast revenue instead of praying for it. That turns a weather-and-mandi-exposed trading P&L into something closer to contracted, recurring revenue, which is the kind of profile that lets you scale without torching the balance sheet. Mulyam's numbers show it: 3,000+ farmers across 9 states, 120+ commodities, 130+ metric tons moving daily. Growth that compounds, because the demand was locked before the cost was spent.
Then defensibility. A demand-led system leaves an exhaust trail (order histories, grade-split patterns, yield curves by region and season) that a pure trader never gets to accumulate. Over time it makes sourcing sharper, grading tighter, and rejection rarer, and the lead widens instead of closing. It's a data moat sitting on top of a physical one.
Last, margin structure. Because demand-led fulfillment cuts rejection and shrinkage and opens up value-added processing, the same produce throws off more gross margin per rupee of revenue. Investors read that as operating leverage: scale the volume and the unit economics get better, not worse. A supply-push trader sees the reverse: margins thin as it chases volume. That split, improving unit economics versus eroding ones, is the real reason demand-led earns the higher multiple.
The takeaway for allocators
The fresh F&V market in India is huge, but the market matters less than the model. You can grow revenue fast in a supply-push posture and still be one bad season, one rejection wave, or one shrinkage spike away from a loss. Demand-led builds the opposite: forecastable revenue, margin that improves with scale, and a data asset that keeps compounding into a moat.
That's what Mulyam is built on. Capturing demand up front (through quick-commerce requirements, the D2C app, and institutional commitments) and letting it drive sourcing, grading, and processing isn't an efficiency tweak around the edges. It's a different kind of business, and it's the one we'd want to own.