// prepared for Meesho · operations

Meesho's whole model is a bet on cost per delivered order.

At 1.8 billion orders a year, every rupee of cost-to-serve is the margin.

The lever you've been pulling by hand

Between FY23 and FY25 you cut logistics cost per delivered order from ₹70.5 to ₹50.8 — nearly 30% in two years¹. Valmo runs about 5% under third-party logistics and you're targeting another 5-7% over the next 12-18 months². That discipline is why FY25 free cash flow turned positive at ₹1,032 crore, from negative ₹2,336 crore the year before³. On a low-AOV, high-volume model, the next few rupees are the hardest — and the most valuable.

$ meesho unit-econ --metric cost_per_delivered_order
FY23      ₹70.5
FY25      ₹50.8   ▼ 29.7%
engine    Valmo  (~5% under 3PL · 20-22% of volume)
next      -5 to -7%  target 12-18 mo

Where a focused AI build adds the next point

The remaining cost sits in decisions made thousands of times a day — routing, sortation, returns triage, support. Meridian AI is a boutique applied-AI studio that ships production-grade LLM and agent systems on your existing stack: retrieval over your own order and logistics data, function-calling agents wired into Valmo and your current services, evals and guardrails in CI. No model training, no rip-and-replace, no new infrastructure to run, and a human-review path wherever a decision touches a shipment or a seller.

Underwritten on one number

Every engagement targets one measurable operating metric per quarter — cost per delivered order, return-to-origin rate, deflected support contacts — and we don't scale the retainer until that metric moves. You prove the point before you fund the next one.