The Marketplace That Pays Before It Sells

Executive Summary

Core claim: A new Indian fashion marketplace does not fail merely because first-order CAC is high. It fails when the acquired customer leaves behind too little retained contribution, repeat behaviour, and marketplace memory to make later orders cheaper.

Five findings define the analysis:

1. Mature marketplaces monetize more than commission. Myntra’s FY25 operating revenue included ₹2,051.8 crore from marketplace services, ₹2,918.9 crore from logistics, and ₹914.5 crore from advertising. [S1]
2. CAC is not publicly disclosed. Myntra’s advertising and promotion expense reached ₹2,105.3 crore, while Nykaa Fashion spent ₹382 crore on marketing and selling/distribution and still reported negative ₹37 crore FY26 EBITDA. [S1][S2]
3. Frequency changes recovery. Meesho reached 10.1 annual transactions per consumer, yet growth spend remained ₹990 crore and marketplace contribution margin was 3.5% of NMV. [S3]
4. Attribution is not incrementality. Project NIRV’s post-view model must be interpreted as a stress test, not a market benchmark.
5. The same 150,000 monthly orders can require ₹3.15 crore or ₹52.5 lakh of direct paid acquisition, depending on new-customer share, paid share, and CAC.

The 150,000-order target, 3× PVC-LTV/CAC hurdle, 12% haircut, 70% viewability, and 60% incrementality are explicit Project NIRV assumptions.

Figure note: The existing 29 visuals are retained unchanged and moved to the appendix so they do not interrupt the 600-word argument. Figures based on external data are mapped to source IDs. All unlabelled scenario values are Project NIRV model outputs, not market-wide benchmarks.

1. The marketplace must remember the customer

A new fashion marketplace does not fail simply because first-order CAC is high. It fails when the first order leaves no durable state. Marketplace memory is the residual after a transaction: known size, saved address, trusted return process, brand preference, app habit, and higher repeat probability. The customer becomes economically valuable only when this state makes later orders cheaper to create.

Myntra’s FY25 consolidated statements show what maturity can produce. It reported ₹6,042.7 crore in operating revenue: ₹2,051.8 crore from marketplace services, ₹2,918.9 crore from logistics, and ₹914.5 crore from advertising. Flipkart Internet reported ₹6,317 crore from advertisement services in FY25. [S1][S4] These revenue layers are not available at launch; they emerge after the platform has repeat, intent-rich traffic.

2. CAC is a system diagnostic

Public filings do not disclose exact CAC for these companies. What they disclose is surrounding pressure. Myntra spent ₹2,105.3 crore on advertising and promotions in FY25, 34.8% of operating revenue. Nykaa Fashion generated ₹4,954 crore GMV and ₹1,447 crore NSV in FY26, but spent ₹382 crore on marketing and selling/distribution, ₹162 crore on fulfilment, and reported EBITDA of negative ₹37 crore. [S1][S2] This does not prove a universal CAC benchmark. It shows why acquisition must be evaluated after fulfilment, returns, discounts, and repeat behaviour.

Meesho illustrates the frequency counterweight. FY26 NMV was ₹41,560 crore, annual frequency reached 10.1 transactions per consumer, and marketplace contribution margin was 3.5% of NMV. Growth spend still reached ₹990 crore. [S3] Repeat spreads acquisition cost; it does not remove logistics or growth pressure.

These disclosures are not directly comparable: Myntra’s spend is measured against operating revenue, Nykaa’s against NSV, and Meesho’s against NMV. They should be read as operating signals, not as a ranking of marketing efficiency.

3. Attribution is not economics

Google defines a view-through conversion as a later conversion following an impression without interaction, and notes that longer windows usually record more of them. [S6] The IAB/MRC standard therefore treats incrementality as causal lift beyond baseline and recommends test-control methods. [S7]

Project NIRV models incremental post-view CAC as CPM divided by 1,000 × viewability × post-view conversion × incrementality. At ₹80 CPM, 70% viewability, and 60% incrementality, 0.05% of viewable impressions converting implies ₹381 CAC; 0.08% implies ₹238. These are stress-test outputs, not market benchmarks.

4. Contribution must carry acquisition

PVC-LTV is contribution per retained order × contribution-bearing orders × 0.88, using a 12% internal haircut. The model applies a 3× PVC-LTV/CAC hurdle as a safety margin, not a universal law. A ₹250 CAC therefore requires ₹750 of present contribution; ₹500 requires ₹1,500.

A retained order survives cancellation, failed delivery, return, refund, and platform-funded discount strongly enough to produce contribution. GMV can rise while this economic base contracts. That distinction governs the model.

Fashion makes this difficult because contribution can disappear after the checkout. An India-focused peer-reviewed study identified 34 return factors; fit and size variation, defects, wrong delivery, lenient return policies, and value for money ranked among the crucial drivers. [S10] Product architecture therefore sets the CAC ceiling before media buying does.

5. Scale only when dependence falls

At the modelled target of 150,000 monthly orders, two marketplaces can look identical and have opposite economics. With 60% new-customer share, 70% paid share, and ₹500 paid CAC, direct acquisition spend is ₹3.15 crore monthly. At 35%, 40%, and ₹250 respectively, it is ₹52.5 lakh. The difference is marketplace memory.

The recommended first engine remains curated mid-market supply, creator-led discovery, paid amplification only after cohort proof, CRM, and category-level contribution tracking. Scale is credible only when paid CAC falls, repeat/CRM/direct share rises, returns remain controlled, and contribution per retained order deepens. The decisive question is not whether the first order can be bought, but whether the second costs less to create.

Visual Appendix

Company evidence

Evidence: [S1].

Evidence: [S2].

Evidence: [S3].

Evidence: [S1], [S4], and [S5].

Evidence: [S1], [S2], and [S3]. Denominators differ—operating revenue, NSV, and NMV—so this is not a like-for-like ranking.

Acquisition, contribution, and scale models

Project NIRV model.

Project NIRV model using attribution definitions from [S6] and measurement principles from [S7].

Project NIRV model.

Project NIRV model.

Project NIRV model.

Project NIRV model; 3× is an internal hurdle.

Project NIRV model; timelines are scenarios, not forecasts.

Project NIRV model.

Project NIRV model.

Project NIRV model.

Project NIRV model.

Marketing and AI

Project NIRV model; channel ranges are not public company disclosures.

Sensitivity model using the 5%-15% range from [S8].

Product architecture

Project NIRV model.

Evidence and derived anchors: [S2], [S3], and [S9]. Meesho’s derived value is not an official AOV.

Project NIRV model.

Project NIRV model; [S10] supports the return mechanisms, not the modelled rupee costs.

Validation and scenario tests

Project NIRV model.

Project NIRV model.

Project NIRV model.

Project NIRV model.

Project NIRV model.

Project NIRV model.

Project NIRV model.