How to Analyse a New-Age Internet Stock (GMV, Take Rate)

In May 2025, an analyst congratulated Blinkit's leadership on a clever-looking quarter: contribution margin had held steady while the headline loss widened, because incentive spends had been moved to the "adjusted" line below it. The reply was unusually honest. Management refused to guide on where contribution margin would go next — "there is a lot of variability, which we are also not certain about and hence are unable to give guidance" (Eternal Q4 FY25 concall, 1 May 2025) — and admitted, plainly, that the loss was running at "-2% margin today in terms of Adjusted EBITDA as a percentage of NOV" (same call). (Illustration of how to read the numbers, not a view on the stock.)

Read that last sentence again, because the trap is hidden in three letters. The loss is quoted as a percentage of NOV — net order value — not GOV, gross order value. On the very same call, management confirmed the two differ "around 22% odd" at the Blinkit level, and in some categories "70% - 80%" (Eternal Q4 FY25 concall). A loss that looks like -2% on the smaller NOV base would look meaningfully worse on the larger GOV base. Nobody is lying. But the denominator is doing quiet work, and a reader who doesn't notice which one is in play will compare two internet companies as if their numbers mean the same thing when they don't.

That is the whole game in this sector. A new-age internet company — quick commerce, online beauty, payments, an insurance or B2B marketplace — does not make money the way a factory does. It runs a take on transactions it doesn't own, and almost every important number is a ratio over a base that the company itself defines. Learn to read the base, the take, and the slope from burning cash to making it, and you can tell a genuine path-to-profit from an accounting one. Miss it, and the adjusted-EBITDA slideware will read it for you.

A boundary first: you will not model these businesses to the rupee from outside, and many of the metrics that decide the outcome are not in the income statement at all — they live in investor decks and concall Q&A. What you can do is read the direction of a handful of operating numbers and check whether management's account of them survives the questions.

What actually drives the economics here?

Strip a new-age internet company down and it is a toll booth on other people's transactions. It does not manufacture the toothpaste or underwrite the insurance; it owns the demand, the data, and the rails, and it skims a slice off every order that crosses the platform. The slice is the take rate; the total value crossing is GMV (gross merchandise value, called GOV or NMV by some). Revenue is roughly GMV times take rate — which is why two companies with identical revenue can have wildly different economics depending on how big the base underneath is. The listed Indian universe here spans the whole spectrum — from quick commerce to payments and online insurance — with names like Swiggy and PB Fintech sitting alongside the platforms dissected below.

That framing fixes the two errors beginners make. First, GMV growth on its own means almost nothing — a platform can buy ₹100 of GMV by handing out ₹15 of discounts and free delivery, booking the growth now and the losses later. Second, a reported "profit" can be an accounting choice — the difference between contribution margin (real, per-order economics) and adjusted EBITDA (after management decides what to "adjust" out) is where the optimism gets parked. The question is never "is it growing?" It is: is the take rate real, do the unit economics work without subsidy, and is there a credible, self-funded path from burning cash to making it?

The homely version: this is a tollgate on a new highway. Early on the owner waives the toll to get traffic onto the road, and pays for the asphalt out of pocket. The only questions that matter are whether drivers keep coming once the toll goes back up, and whether the toll, fully charged, covers the asphalt before the cash runs out. Everything below is a way of answering those two.

The metrics that matter — and where they hide

These are sector-specific. Generic ROE-and-P/E reading will mislead you here, because most of these companies are deliberately suppressing reported profit to buy growth.

GMV / NMV — the size of the toll road

GMV (gross merchandise value; "GOV" at Eternal, "NMV" at some marketplaces) is the total transaction value flowing across the platform. It is the base everything else is a percentage of, and it is the number management most wants you to anchor on, because it grows fastest.

Where to find it: investor presentations and the shareholders' letter, almost never the income statement. What good looks like: durable growth with improving economics, not growth bought with widening losses. Real number: Nykaa's consolidated GMV was ₹5,795 crore in Q3 FY26, up 28% year-on-year, with the beauty vertical at ₹4,302 crore growing 27% (Nykaa Q3 FY26 concall, Inve data). Eternal's Blinkit hit a ~$5 billion annualised GOV run-rate by Q1 FY26 (Inve data, Q1 FY26).

The catch — watch the denominator. Eternal switched its headline base from GOV to NOV precisely because, as it grows into unbranded grocery, "incremental GOV will start giving an incorrect picture on the overall size of the business" (Eternal Q4 FY25 concall). That is a defensible reporting choice — and also exactly the kind of base-redefinition that makes year-on-year and cross-company comparisons treacherous. When a company changes how it counts the biggest number on the page, slow down.

Take rate — the toll, as a percentage

Take rate is revenue as a share of GMV — the slice the platform keeps. It is the single best test of pricing power: a platform that can raise its take without losing volume owns its demand; one that can't is a price-taker dressed as a tech company.

Where to find it: rarely stated directly — you usually compute it (segment revenue ÷ segment GMV), or assemble it from concall Q&A. What good looks like: a take rate that rises gently over time as ancillary, high-margin streams (advertising, subscriptions) grow faster than the base. Real number: at Blinkit, ad income alone is "north of 4% of GOV today" — and, crucially, it sits outside the GOV definition and "directly goes to our revenue" (Eternal Q4 FY25 concall). So a meaningful chunk of monetisation never appears in the headline base at all, which both helps the economics and complicates any take-rate you compute from reported numbers. This is why comparing two platforms' take rates without reading each one's definitions is a category error.

Contribution margin — the only honest per-order number

Contribution margin is revenue minus the variable costs of fulfilling that revenue — delivery, payment, packaging, the customer-facing discounts — before any corporate overhead. It is the closest thing to truth in this sector, because it cannot be flattered by moving costs to an "adjusted" line. If contribution margin is negative, the company loses money on every order it ships and growth makes the hole deeper.

Where to find it: investor decks and concall, again — not the P&L. What good looks like: positive and expanding, with a stated long-term target management is held to. Real number: Paytm's contribution margin was about 57% in Q3 FY26, with management guiding it toward "mid-50s" as the business mix shifts (Paytm Q3 FY26 concall). Blinkit's contribution margin on a NOV basis was 3.9% in Q1 FY26, with a long-term target of "5-6%" reiterated in Q4 FY26 (Inve data). Hold management to that target across quarters; a contribution-margin goal that keeps getting pushed out is the tell.

Adjusted EBITDA — read it adversarially

Adjusted EBITDA is EBITDA after management strips out whatever it deems non-recurring or non-cash — most often ESOP cost and "investment" spends. It is the number every deck leads with, and the one to trust least, because the company chooses what to adjust.

Where to find it: front page of the deck (which is the problem). What good looks like: the gap between contribution margin and adjusted EBITDA narrowing as the business scales overhead — and ESOP being a genuinely small, declining share. Real number: Blinkit ran at "-2% of NOV" on adjusted EBITDA in Q4 FY25 while contribution margin held — because, as the analyst noted, "the loss expansion is more to the adjusted levels" (Eternal Q4 FY25 concall). The honest read: contribution economics were fine; the reported loss widened on choices about where to book incentive spend, on a base (NOV) ~22% smaller than GOV. Always re-anchor adjusted EBITDA back to contribution margin and to actual cash.

Monthly transacting users (MTU) and AOV — the volume engine

MTU (or MTC, monthly transacting customers; AUTC, annual unique transacting customers) counts the people actually paying, not "registered users". AOV — average order value — is what each order is worth. Revenue is, ultimately, users × frequency × AOV × take rate, so these are the levers underneath GMV.

Where to find it: decks and concall. What good looks like: MTU growing and AOV stable or rising (so growth isn't being bought by chasing low-value orders), with strong cohort retention. Real numbers: Blinkit had 17 million MTUs and Eternal's food delivery 23 million MTCs in Q1 FY26 (Inve data). Nykaa reached 16.5 million annual unique transacting customers in Q1 FY26, up 26%, with AOV "increased 4% to over INR 2,000" (Nykaa Q1 FY26 concall). Rising users with a rising AOV is the healthy combination; rising users with falling AOV usually means discounting is doing the recruiting.

Cash burn and runway — the clock everything runs against

Every metric above is a race against the bank balance. Cash burn is operating cash outflow per quarter; runway is cash on hand divided by burn. A company that needs to raise equity before it turns cash-positive hands its existing owners a dilution bill — and the timing (forced, in a bad market, versus opportunistic) tells you a lot.

Where to find it: the cash-flow statement plus the balance-sheet cash line; the qualitative colour is in the concall. What good looks like: burn shrinking quarter on quarter while GMV grows — the burn-down that precedes breakeven. Real number: Paytm is the cleanest live example of the clock being beaten. It posted a roughly ₹545 crore net loss in Q4 FY25, then printed four straight profitable quarters through FY26 — about ₹123, ₹21, ₹225 and ₹183 crore — and full-year FY26 EBITDA of "500 crores" (Inve data; Paytm Q4 FY26 concall). The loss-to-profit turn, funded from its own improving contribution margin rather than a fresh raise, is exactly the slope you are looking for.

How do you value a business that chooses its own profit?

You cannot use a P/E when there are no stable earnings, and you should distrust an EV/adjusted-EBITDA multiple because the denominator is a management choice. So this sector is read on EV/sales and, more honestly, EV/GMV — but only ever alongside a credible path to profit. A multiple without a path is just hope with a decimal point.

The discipline is to make the company show its own arithmetic. Take the stated long-term contribution-margin target, apply it to a defensible future GMV, subtract a realistic overhead, and ask whether the resulting steady-state profit justifies today's enterprise value. For Blinkit, that means taking the "5-6%" long-term NOV margin (Inve data, Q4 FY26) against a future order base and asking what GOV the business must reach for the maths to work — and how many years and how much cash that takes. For Paytm, where the turn has already happened, EV/sales can be cross-checked against an emerging EBITDA the way you would a normal company, because the path stopped being hypothetical.

The same instinct that separates reported profit from real cash earnings applies here, sharpened: in this sector the company is telling you which profit to look at, so your job is to look at the other one.

A worked case: the denominator that does the talking

Put the Blinkit Q4 FY25 call together as one picture, because three forces converge on a single quarter — and that convergence, not any one number, is the lesson (illustration, not a view on the stock; figures from the Eternal Q4 FY25 concall, 1 May 2025, and Inve data).

Here is the uncomfortable part, and the reason this teaches more than a fraud would: nothing here is dishonest. The contribution margin really did hold. The NOV base really is a more sensible denominator as the mix shifts to unbranded grocery. Ad income really does flow to revenue. Each choice is defensible alone. Together, they mean that the headline loss, the base it's measured against, and the monetisation that doesn't show up in that base are all moving at once — and an investor anchoring on any single reported figure is reading a number the company has, with complete propriety, arranged. The craft is to find the picture uncomfortable precisely while every individual line checks out.

This is also why the concall matters more here than in almost any other sector: the deck gives you the adjusted number; the Q&A is where someone forces management to say what it's adjusted against. Inve's concall summaries pull every such exchange — the question, the speaker, the quote — into one place per quarter, so the denominator switch and the dropped guidance don't slip past on slide 4.

The red flags specific to this sector

• GMV growing faster than revenue, quarter after quarter. The take rate is silently falling — the platform is buying volume it can't monetise.
• The base gets redefined. GOV→NOV, "core" GMV, a new "contribution profit" line. Sometimes legitimate (Blinkit's was arguable), but every redefinition resets the comparison clock and deserves a hard question.
• Adjusted EBITDA improving while contribution margin stalls. The progress is overhead-shuffling and ESOP exclusions, not better per-order economics. Re-anchor to contribution margin and cash every time.
• Rising MTU with falling AOV and falling retention. Growth bought with discounts that recruit one-time bargain hunters, not customers.
• Burn flat or rising as GMV grows, with the cash clock ticking toward a raise. The path to profit isn't a slope, it's a wish — and the dilution bill lands on today's owners.
• Management that won't give a contribution-margin or breakeven number it can be held to. Sometimes genuine uncertainty (as Blinkit said); often, the absence of a number you can mark against next quarter is itself the answer.

A management that talks about its contribution margin, its cash runway, and its path to breakeven unprompted is telling you it has done the maths. One that keeps the conversation on GMV and adjusted EBITDA is telling you where it would rather you not look.

Where this read can be wrong

The strongest case against everything above is that in a genuine land-grab, suppressed profit is the correct strategy, and an investor who insists on near-term contribution margins will sell the best compounders too early. Amazon ran on thin or negative reported profit for two decades while building an unassailable position; a quick-commerce platform deliberately taking a -2% adjusted loss to win the only three cities that will ever matter may be making exactly the right call. The metrics here can tell you whether the economics can work; they cannot tell you whether a winner-take-most market is still being decided, and that question — not the contribution margin — often determines the outcome.

So the honest claim is narrower than it looks. Reading GMV, take rate, contribution margin, and burn against the company's own guidance tells you whether management's path-to-profit story survives its own numbers, and whether the optics flatter the reality. It does not tell you whether the market is worth winning, or who will win it — and in this sector, that is frequently the whole game. We are confident about the method; on which platform compounds, far less so. We have also been wrong on the timing of these turns before — a path that looked two years out can arrive in three quarters, as Paytm's did, or stall for years, as several have.

A hard limit worth restating: judging these companies means judging competitive dynamics in markets that are still forming, which is genuinely hard. A clean concall and an improving contribution margin do not immunise a platform against a better-funded rival willing to subsidise longer — the FY25–26 quick-commerce price war is precisely that, and it is why Blinkit's margin "could have been profitable absent competition" (Inve data, Q4 FY25).

Frequently asked questions

The discipline comes down to refusing to read the number the company hands you. The business is the toll road, and it speaks through the base, the take, the per-order economics, and the cash clock — not the adjusted line on slide one. So invert the question you bring to these results. Don't ask "is this growing?" Ask: if management were quietly arranging the optics — picking the friendlier denominator, parking the spend below the line, headlining the base that grows fastest — what would the numbers look like, and does this disclosure rule that out? A loss measured against a base that just got 22% smaller does not rule it out; it is the pattern itself.

And the owner's question, the one to sit with before you buy a single share of any platform: what must I believe about the steady state — the take rate fully charged, the subsidies gone, the market settled — for this toll road to earn its keep, and will the company reach it on its own cash before it has to come back to me for more? If the honest answer leans on the adjusted slide rather than the contribution margin and the cash balance, you've read the deck, not the business.

Related reading

• How to Analyse a Bank Stock (NIM, CASA, GNPA, ROA)
• How to Analyse a Building Products Company (Pipes, Tiles)
• How to Analyse a Cables and Wires Stock