One synthetic company again. ThaliGo (synthetic) is an Indian food-delivery marketplace. Figures are FY26, in ₹ crore unless stated.
The raw data. GMV is 8,000. There were 20 crore orders (200 million) at an AOV of ₹400. Annual transacting buyers number 2.5 crore, and there are 2.4 lakh active restaurant partners. Each order earns a restaurant commission of ₹64 (16% of AOV), customer delivery + platform fees of ₹18 and advertising of ₹10, for revenue per order of ₹92. Per-order variable costs are rider payout ₹52, discounts ₹14, payment processing ₹5, plus support and refunds ₹6, for ₹77 in all. Fixed costs are technology + HQ at 210 and brand marketing at 140. Three years ago (FY23), discounts ran ₹32/order and delivery fees were ₹8.
How do you break down GMV and take rate?
Scale and its decomposition. Always tear GMV into its three drivers, because each is a different lever with a different cost:
GMV = Active buyers × Order frequency × AOV
8,000 cr = 2.5 cr buyers × 8 orders/yr × ₹400Advertisement
Growth from buyers costs CAC. Growth from frequency is the cheap, high-quality kind. Habit formation is the metric that separates food delivery, ordered ~weekly, from furniture, bought ~biennially. Growth from AOV is mix and inflation. Nothing wrong with that, but check it isn't masking falling frequency. On the seller side, 2.4 lakh restaurants means GMV per restaurant ≈ ₹3.3 lakh/yr, which is the supply side's health check. A partner earning too little churns. Seller churn destroys selection, and lost selection destroys buyer frequency. That's the flywheel in reverse.
Monetization. Take rate = 92/400 = 23% all-in. Net revenue = 20 cr × ₹92 = ₹1,840 cr (= 8,000 × 23%). Look at the composition: 16 points of commission, 4.5 points of customer fees, 2.5 points of ads. The ads line is the one with room to grow without touching the restaurant relationship directly.
How do you measure liquidity and cohort retention?
Liquidity metrics. The panel differs by category, but the question is constant: how reliably and quickly does a match happen?
| Marketplace type | Core liquidity metrics |
|---|---|
| Food delivery / quick commerce | Order fill rate (ThaliGo: 97.5%), median/p90 delivery time (33/48 min), rider utilization, restaurant online-hours |
| Goods (Etsy/Meesho-style) | Sell-through rate, time-to-first-sale for new sellers, search-to-purchase conversion |
| Services/labor (Urban-Company-style) | Job fill rate, time-to-assign, pro utilization %, repeat-match rate |
| Ride-hailing | Match rate, ETA/time-to-fill, surge frequency, driver idle % |
| Lodging (Airbnb-style) | Search-to-book conversion, occupancy of listings, booking lead time |
Liquidity metrics rarely appear in filings. You'll find them in investor decks, S-1/DRHPs, and your own scuttlebutt. Order the product at 11 pm in a tier-2 city: that's primary research, and Phase 8 will formalize it. When a platform stops disclosing a liquidity metric it used to trumpet, treat the disappearance as data.
Cohort GMV retention is the marketplace analog of NRR. Take the buyers acquired in year N and measure their total GMV in year N+1 ÷ year N. ThaliGo's FY25 cohort spent ₹1,200 cr in its first year and ₹1,080 cr in FY26, which is 90% cohort GMV retention. Break that down. Only 55% of the cohort's buyers are still active. The survivors' frequency and AOV, though, rose enough to recover most of the value (0.55 × higher spend ≈ 0.90). Older cohorts (FY23) are at 105%, because the habit deepens with age. Above 100%, the platform grows even with zero new buyers. Think of it as the marketplace version of NRR >100%, every bit as decisive for the multiple. Below ~70%, growth is a treadmill of paid acquisition.
What is contribution margin per order?
Contribution margin per transaction: the unit economics floor. Per order:
CM/order = Revenue/order − Variable costs/order = 92 − 77 = ₹15 (3.75% of AOV)
Total contribution = 20 cr × ₹15 = ₹300 cr → EBITDA = 300 − 350 fixed = −₹50 crThaliGo loses money at the EBITDA line while making ₹15 on every marginal order. Call it the marketplace version of the SaaS freeze-growth column. The leverage math falls straight out. Breakeven needs contribution of 350, i.e. 350/15 = 23.3 cr orders. So ThaliGo needs 17% more volume at constant per-order economics, or the same volume with CM/order of ₹17.50. Marketplace analysis, then, is contribution-margin analysis. Consolidated EBITDA tells you where the company is on the volume curve. CM per order and its trend tell you whether the curve leads anywhere. Indian platform companies now disclose contribution margin by segment in their shareholder letters, a practice Zomato/Eternal made standard. These are non-Ind-AS metrics, so reconcile them to the audited P&L before you trust them. Apply the same non-GAAP policing you learned on the statements.
How does a subsidy exit show whether GMV is real?
The FY23-to-FY26 swing: anatomy of a subsidy exit. Three years ago, ThaliGo's revenue was ₹74/order (commission ₹64 + delivery fees ₹8 + ads ₹2). Discounts of ₹32 sat inside total variable costs of ₹95, so CM was −₹21/order. Every order destroyed ₹21, and "growth" meant paying the public to transact. How did it get to +₹15? Discounts were cut from 32 to 14. Delivery and platform fees went up from 8 to 18. The platform-fee innovation, a flat ₹2-then-₹10-per-order charge that Indian food delivery introduced from 2023, is a pure take-rate lever. Ads were built from 2 to 10. That's the standard subsidy-exit playbook. Whether the GMV was real or bought shows in what volume did next. ThaliGo's frequency dipped from 8.0 to 7.4 orders/yr in the two quarters after the discount cuts, then recovered to 8.0 as habit held. GMV that survives subsidy withdrawal was real demand. GMV that evaporates was rented. Run this test on every cash-burning platform. What happens to cohort activity when the incentives stop? The marketplace killers turn it into a named flag.