पाठशाला Pathshala · हिसाब Hisāb, Unit economics · Lesson 29 · Scale
Benchmarks: what good looks like for Indian startups, by model
The published ranges for SaaS, D2C, marketplaces and lending, with Indian listed companies as reference points, and how to read your own margins, payback and retention against them without fooling yourself.
Pathshala, The Founder Library · 11 October 2026 · 7 min read
Every founder is eventually told that their numbers are below benchmark, usually by someone holding a range from a different kind of company in a different country at a different stage. A benchmark is useful when it is the right one, read the right way. Then it tells a founder which of three or four numbers to work on first, and how far there is to go.
This lesson sets out what a benchmark is and is not, gives the published ranges for four common models, SaaS, D2C, marketplaces and lending, with Indian listed companies as reference points where they publish the number, and ends with a quarterly page that keeps the comparison honest.
What a benchmark is and is not
Most published ranges come from venture portfolios and public markets in the United States. They describe companies that raised money, survived and reported, which is not a random sample. They are computed with definitions that may not match yours: gross margin with or without support cost, payback on gross profit or on revenue, retention by logo or by rupee. And they are stage-specific: a company at ₹5 crore of revenue is not meant to look like one at ₹500 crore.
So use benchmarks for direction, not as a pass mark. Three rules help. Match the definition before comparing the number; the [lesson on COGS](/library/what-belongs-in-cogs) and the [LTV lesson](/library/ltv-computed-honestly) set out the definitions investors use. Compare with your own model and segment, not with the best-known one. And put more weight on the trend in your own number than on its distance from someone else’s: a gross margin rising three points a year below the band is a better story than one falling inside it.
Three Indian adjustments are worth making before any comparison. Prices are lower, so a payback target built on US contract sizes is harder to meet at Indian ones and is lost quickly to a slightly higher sales cost. Customers often pay later, so payback computed on invoices flatters payback computed on cash. And GST comes out of both revenue and cost before any margin is compared, because tax collected is not revenue and input credit claimed back is not cost.
SaaS: margin, payback and net retention
Bessemer’s Scaling to $100 Million gives the cleanest set. Average gross margin for a cloud company is 65 to 70 per cent regardless of stage, and the middle half sits between roughly 60 and 80 per cent. On CAC payback it gives targets by segment: under 12 months for SMB-focused companies, under 18 for mid-market and under 24 for enterprise, against an average of 15 months for companies between $1 million and $10 million of ARR. On retention it reports average net retention of 140 per cent between $1 million and $10 million of ARR, falling to 120 per cent above, with only the bottom quartile below 100.
Those retention figures describe a strong portfolio. Lenny Rachitsky’s What is good retention?, drawn from operators and investors, gives more usable ranges by segment: good net revenue retention of about 90 per cent and great of about 110 for land-and-expand SMB and mid-market software, and good of about 110 and great of about 130 for enterprise. Many Indian SaaS companies sell to SMBs at lower prices than their US peers, so the SMB row is usually the right one, and payback should be held to the SMB target even when the deal sizes feel larger.
D2C: margin, repeat and the marketing line
Consumer brands publish fewer standard ranges, so the useful references are listed Indian companies. Honasa Consumer, the parent of Mamaearth, reported a gross margin of 70.7 per cent and an EBITDA margin of 5.1 per cent for the quarter to March 2025. A beauty or personal-care brand with a gross margin far below that is either buying more expensively, discounting more heavily or selling more through channels that take a larger cut.
The retention reference comes from Rachitsky’s consumer transactional row: about 30 per cent of customers still buying after six months is good and 50 per cent is great. A D2C brand below 30 per cent is paying to acquire most of its revenue every year, which is why the [D2C lesson](/library/d2c-unit-economics-order-that-must-make-money) insists that the first order make money on its own. Watch the gap between gross margin and EBITDA margin: in a brand it is mostly marketing, and it closes only when repeat customers carry a growing share of revenue.
Marketplaces: take rate and contribution per order
Take rates vary by category and by how much the marketplace does. Bill Gurley’s A Rake Too Far reported eBay just under 10 per cent of gross merchandise sales and Amazon’s marketplace fees at 6 to 15 per cent by category, against 30 per cent for the app stores, and argued that the most dangerous strategy for a platform is to price too high. A marketplace that takes more than the range should be able to say what it does that the others do not: logistics, payments, credit, demand the supplier could not find alone.
The take rate is half the answer; the other half is what survives delivery and discounts. Swiggy’s Q4 FY2025 shareholder letter reports food-delivery contribution margin of 7.8 per cent of gross order value and Instamart at minus 5.6 per cent, after delivery charges, platform-funded discounts and other variable costs. Those two numbers from one company show the range in a single model: a mature line well positive per order, a fast-growing one still negative. A marketplace founder should know which of the two they resemble and how many quarters it takes to move. The [marketplace lesson](/library/marketplace-unit-economics-take-rate-gmv) works through the arithmetic.
Lending: credit cost before growth
A lender’s margin is the yield less the cost of funds, operating cost and credit cost, and the last of those decides most outcomes. The listed reference is Bajaj Finance, whose investor presentation for FY2026 reports gross NPAs of 1.01 per cent at 31 March 2026, loan losses and provisions of 2.09 per cent of average assets under finance, a cost of funds of 7.54 per cent and a return on assets of 4.3 per cent. It is one of the best-run lenders in India and the wrong target for a two-year-old fintech; it is the right reference for how far a book has to mature.
For a young lender the test is simpler: credit cost on each monthly cohort, at the same age, falling or flat, and comfortably below the spread between yield and cost of funds. A book growing fast enough that new loans hide the losses on old ones will show a low NPA ratio until growth slows. The [lending lesson](/library/lending-unit-economics-yield-credit-cost) shows how to read the book by vintage.
Reading your numbers against the ranges
Pick the model, set your three numbers and read where each sits. The shaded bands are the published ranges; the ticks are reported figures for named Indian companies. The figure starts with a SaaS company at 68 per cent gross margin, 20 months of payback and 95 per cent net revenue retention: the margin is in the band, payback inside the mid-market target but well beyond the SMB one, and retention in the good range but short of great.
Then ask which number to work on first. The rule of thumb is to fix the one furthest from range that the others depend on. In SaaS, retention usually comes first, because it raises lifetime value, shortens effective payback and lifts the margin investors will credit. In D2C, repeat purchase comes first, for the same reason. In a marketplace, contribution per order comes before take rate. In lending, credit cost comes before everything.
A benchmark tells you where to look, not whether you have passed. The trend in your own number is worth more than its distance from someone else’s.
The quarterly benchmark page
Once a quarter, after the books close, put one page together: the three numbers that matter for the model, computed on the definitions the benchmark uses, beside the published range and the nearest Indian reference, with the last four quarters for each. Write one line per number on why it moved. Choose the one number the company will work on next quarter and the lever that will move it.
Keep the sources and their dates on the page, and refresh them once a year; benchmarks drift as markets change. When an investor quotes a range, ask which companies and which definition it comes from, and add it to the page if it is the right comparison. A founder who can say where every number sits, against which range and why, has already answered half of the diligence questions before they are asked.
Nothing here is legal, tax or investment advice. Ranges and company figures are stated as their sources state them and were checked on 11 October 2026; published benchmarks describe portfolios and listed companies, not random samples of Indian startups.
Sources
- Mary D’Onofrio, Ethan Ding and Atlas editors, Scaling to $100 Million, Bessemer Venture Partners, September 2021 — Cloud gross margin 65–70 per cent, middle half 60–80; CAC payback targets under 12, 18 and 24 months by segment; average net retention 140 and 120 per cent.
- Lenny Rachitsky, What is good retention?, Lenny’s Newsletter, June 2020 — Six-month user retention and twelve-month net revenue retention, good and great, by category.
- Bill Gurley, A Rake Too Far: Optimal Platform Pricing Strategy, Above the Crowd, April 2013 — eBay just under 10 per cent; Amazon marketplace 6–15 per cent; app stores 30 per cent.
- Business Standard, Honasa Consumer shares fly 14% on Q4 results, May 2025 — Gross margin 70.7 per cent and EBITDA margin 5.1 per cent for the quarter to March 2025.
- Swiggy Limited, Q4 FY2025 Shareholder Letter, May 2025 — Food delivery contribution margin 7.8 per cent of GOV; Instamart minus 5.6 per cent.
- Bajaj Finance Limited, Investor Presentation for the quarter and year ended 31 March 2026, April 2026 — GNPA 1.01 per cent; loan losses and provisions 2.09 per cent of average AUF; cost of funds 7.54 per cent; RoA 4.3 per cent for FY26.