Startup Business Models and Unit Economics Explained

A startup's business model is how it makes money; its unit economics show whether it makes money on each customer. Unit economics largely come down to five numbers: customer acquisition cost (CAC), lifetime value (LTV), CAC payback, gross margin and contribution margin. If a customer is not worth clearly more than they cost to win and serve, growth tends to make the problem bigger.

Our view is that most founders are better off copying a proven business model (by Y Combinator's count, nearly every billion-dollar company uses one of nine) and putting their originality into the product.

Be original in what you build. Be boring in how you get paid.

Definition: Unit economics measure the revenue and variable cost attached to a single unit of the business, usually one customer, so you can tell whether each new customer adds or destroys value.

This guide covers the model choice, the formulas, an illustrative worked example, dated benchmarks, the levers we would look at first and a checklist for your next board meeting. For a tour of revenue model types, see startup revenue models explained; this page is about the per-customer math behind whichever model you choose.

Choose a proven startup business model before you innovate on anything else

In most cases we wouldn't invent a new way to make money. When founders can't get funded and the business won't grow, a common hidden cause is an unproven model: revenue that sits too far from the payment, scales with headcount or depends on someone else's platform. Nine models account for nearly every billion-dollar company, and each has its own primary metrics:

In our view the evidence for sticking to these is strong. Y Combinator's own analysis of its 100 most valuable companies (from a talk dating to around 2022) found SaaS behind 31, transactional businesses behind 22 and marketplaces behind 14, so those three models covered 67 percent of the list. Marketplaces punched above their weight: 14 percent of the companies but about 30 percent of the value, with five of the top ten (Airbnb, Instacart, DoorDash, OpenSea and Faire). Transactional companies were 22 percent of the list and 29 percent of the value. Advertising was only 3 percent. Our rule of thumb: be wary of ads as your primary model unless you genuinely expect to become one of the ten biggest sites on the internet.

What that list leaves out is just as instructive. It contained no services or consulting businesses, no affiliate businesses, no hardware businesses and none built on top of another company's platform. Each omission can be read as a unit economics story: consulting scales with people and earns low margins, affiliates sit too far from the payment, hardware eats capital, and a platform owner can take the revenue for itself. The models that tend to win share five traits: recurring revenue, high retention, a defensible moat, growth that scales with software rather than people, and a position as close to the transaction as possible.

The five unit economics numbers and how to calculate them

Compute these five numbers every month, with definitions you keep stable. Quietly redefining a metric to make it look better fools one person first: you.

Gross margin %          = (revenue - cost of goods sold) / revenue
Contribution margin %   = (revenue - COGS - other variable costs) / revenue
CAC (paid)              = paid sales and marketing spend / customers won through paid channels
CAC (blended)           = all sales and marketing spend / all new customers
CAC payback (months)    = CAC / (monthly revenue per customer x gross margin %)
Customer lifetime       = 1 / monthly churn rate   (months)
LTV                     = monthly revenue per customer x margin % x lifetime
LTV:CAC                 = LTV / CAC

How we would sort costs into each bucket:

• Cost of goods sold (COGS) is any cost that rises with each additional customer. For an AI product, the model API bill is COGS. Free cloud or model credits don't remove that cost. In our view they hide it until the credits run out. Pure software once ran near 95 percent gross margin, while operationally heavy businesses can sit at 5 to 15 percent.
• Contribution margin subtracts every other variable cost as well: payment processing, support time that scales with usage, fraud, shipping, card issuing. A consumer bank such as Monzo nets per-customer revenue against the cost of cards, support contacts, fraud and transaction fees. Fixed costs such as engineering salaries and rent stay out.
• CAC is most useful when it counts customers who stick. We suggest measuring it to an active, paying, retained user rather than a sign-up, because early drop-off in consumer products can reach 80 to 90 percent. It helps to report paid CAC separately: many investors care about it most, because blended CAC flatters the paid channel with free organic customers.
• LTV in our view should use margin rather than revenue. One approach is to build it from the gross or contribution margin a customer generates, including support, installation and servicing costs. The strictest version, and one many investors have in mind, is the present value of the future net profit from that customer.

Worked example: unit economics for a B2B SaaS startup

Consider an illustrative startup selling an AI support tool to mid-sized companies. All figures are per customer per month unless noted, and every number below was computed in Python.

• Price: $2,000 a month ($24,000 a year)
• COGS: $440 (model API usage, hosting, onboarding support)
• Other variable costs: $120 (payment processing, customer success time)
• Last quarter: $270,000 of sales and marketing spend won 15 new customers
• Monthly logo churn: 2 percent

The unit economics:

On the classic rules of thumb this passes: LTV of about 3x CAC and CAC recovered in under 12 months.

Now pressure-test it the way an investor would. A model that only works on its best assumptions isn't a model yet.

• Cap the lifetime. A 50-month average life assumes churn stays at 2 percent forever. Counting only the first 60 months of a declining cohort, LTV falls to about $50,600 and LTV:CAC to 2.8x. Conservative investors often look at it this way.
• Double churn to 4 percent. LTV halves to $36,000 and LTV:CAC drops to 2.0x. Retention is often the most sensitive input, which is why our cohort retention guide is the natural next read.
• Raise the price 25 percent to $2,500, holding dollar costs flat (a simplification). Monthly contribution rises to $1,940, payback falls to about 9.3 months on contribution, and LTV:CAC climbs to 5.4x without winning a single extra customer.
• Add ten organic customers. Blended CAC drops to $10,800, but paid CAC is still $18,000. We would report both, and be wary of letting the blended number justify more ad spend.
• Assume the ad platform understates CAC by 20 percent. True CAC becomes $22,500 ($18,000 / 0.8), payback stretches from 11.5 to about 14.4 months and contribution LTV:CAC falls from 4.0x to 3.2x. The model still passes, with far less room.

Finally, check the sales model. A common rule of thumb (it varies by company and segment) is that a salesperson closes roughly five times their total compensation in new ARR. On that rule, an account executive on $150,000 of total pay would bring in about $750,000 of new ARR, or about 31 deals a year at $24,000. That is workable. At $1,000 a year, the same target would take about 750 deals, more than 60 a month, which likely only a self-serve or inside-sales motion can handle.

Unit economics benchmarks, with dates

Benchmarks vary by segment and year. Read these as ranges, not targets. Our startup KPI framework covers which metrics to watch at each stage; the figures below are the ones specific to unit economics.

• LTV:CAC. David Skok's "Startup Killer" essay (published 2009, updated 2016) set the benchmark most investors still use: about 3x for a viable recurring-revenue business. His "SaaS Metrics 2.0" guide (first published 2013) puts the best SaaS businesses above 3x, sometimes as high as 7x or 8x.
• CAC payback. Skok's line is under 12 months, with the best companies recovering CAC in 5 to 7 months. Bessemer's "Scaling to $100 Million" (September 2021) sets targets by segment: under 12 months for SMB, under 18 for mid-market and under 24 for enterprise, because enterprise customers stay longer.
• Gross margin. Bessemer's 2021 report puts the average cloud business at 65 to 70 percent gross margin, with the middle 50 percent between about 60 and 80 percent. When setting B2B prices, many founders aim for software margins of 80 to 90 percent.
• Net dollar retention. For early-stage B2B SaaS, 125 to 150 percent is a commonly cited bar, with 110 to 120 percent good for mature companies; below 100 percent in enterprise SaaS usually suggests something is wrong.
• Consumer. Ranges we often see cited: 15 percent monthly user growth as good, a net promoter score of +50 as a minimum for a new consumer company, and more than 80 percent organic growth as typical of the strongest consumer companies. Great consumer companies rarely reach scale with more than half their sign-ups coming from paid ads.
• Spend context. SaaS Capital's 2026 spending benchmarks (1,000+ private B2B SaaS companies surveyed in March 2026) put median spend at 15 percent of ARR on sales and 8 percent on marketing, a useful sanity check on what your CAC implies about total go-to-market spend.

How to improve unit economics: six levers we would look at

1. Price on value, not cost. One approach is to write down with the customer's champion what your product saves or earns them, then charge roughly 25 to 50 percent of that value, so the customer keeps the larger share. As an illustration, take a 100-agent support team that costs $10M a year fully loaded: a 20 percent saving is worth $2M, and a price near $700,000 (about a third) is a good deal for both sides. Cost is only a floor. Our pricing guide has a worksheet for this step.
2. Consider raising prices before you cut anything else. Underpricing by an order of magnitude is common at seed; some startups charge a tenth or even a hundredth of what the product is worth. Raising prices is often the easiest way to grow revenue, and higher margins can let you outspend competitors on acquisition.
3. Measure CAC by channel, to a retained customer. Record where every customer came from and keep that record. A channel that delivers cheap sign-ups can still produce negative lifetime value; Monzo shut one down for exactly that reason. Not all revenue is good revenue, a point we make in our take on toxic customers.
4. Move toward committed recurring revenue. In our view committed MRR or ARR tends to beat pure usage pricing because contracts protect revenue in a downturn. A practical bridge: run usage pricing for a month or two, then offer a flat monthly commitment slightly below the customer's average usage in exchange for a 12-month contract.
5. Fix retention before scaling acquisition. At 95 percent monthly retention, 100 customers become 54 after a year; at 90 percent, only 28 remain. A leaky bucket is hard to scale.
6. Fix negative unit economics before growing. Tom Blomfield, a Y Combinator partner, ran Monzo while it lost money on each of its first half million or so customers, then flipped the economics by bringing technology in house, adding fees and launching products customers paid for. His lesson, which we tend to agree with: fix the unit economics first, then scale.

Enforce CAC discipline as you scale

Measuring unit economics honestly is step one. Step two is building rules that stop a growth team from spending past them. The clearest public description of such rules comes from Matteo Franceschetti, co-founder and CEO of the sleep technology company Eight Sleep, in a September 2026 interview. It is one company's practice, not a benchmark, but the mechanism transfers.

Give every channel a CAC cap. Open paid channels one at a time, usually starting with Meta, rather than spreading budget everywhere at once. Each new channel gets a small weekly budget and a maximum acceptable CAC. The team can then scale spend as far as the channel allows (Franceschetti's example ran to $100M) as long as it stays inside the cap, and earns more budget only by proving it can. The reason he gives is behavioral: growth teams tend to want to spend more, and a team that overspends and then pulls back shows no year-over-year growth the following year. By his estimate, Eight Sleep could grow about 50 percent faster by spending more, but its CAC would turn upside down the year after.

Consider building your own CAC model. Ad platforms tend to over-attribute sales to themselves; Franceschetti estimates Meta's reported CAC is probably about 20 percent lower than the true figure, which is the scenario stress-tested in the worked example above. One practice is to run an incrementality test roughly every six months: switch a channel off in one region, leave it on in a comparable one, and measure how much revenue disappears.

Ask for payback on the first sale where the product allows it. Eight Sleep wanted immediate payback and a healthy contribution margin on the first sale, with the subscription compounding on top. That is stricter than a 12-month test and works because a hardware product with an up-front price collects most of its revenue at once; a monthly SaaS product cannot. It is also a useful counterpoint to hardware's absence from the top-100 list: the category is hard, but strict first-order economics are one way to make it work.

Count AI savings in the budget, not in anecdotes. Tying a specific AI tool to a specific saving is often very hard, because freed-up hours fill with other useful work. One option is to capture the gain through operating expenses instead: if you believe AI makes your staff 10 to 30 percent more productive, hold headcount flat next year instead of growing it. Another is to give each leader one combined budget for headcount and compute and let them decide where each dollar earns more. In our view model costs belong in COGS, and the productivity they buy should show up as slower headcount growth. Our guide to AI gross margins and inference costs covers the COGS side in depth.

Where views differ on unit economics

Serious investors tend to agree on the destination and differ on emphasis. Our read on the main splits:

• Which number comes first. At seed, we tend to lead with revenue for B2B, and some investors treat gross merchandise value and gross transaction value as vanity metrics that teams end up optimizing by mistake, while marketplace investors often still want to see them. LTV and CAC frameworks matter more later, because early LTV depends on churn you have not observed for long enough.
• LTV precision. Definitions range from discounted net profit, to gross margin over the customer's lifetime, to plain revenue. The spread between them can be two or three times, so it is worth stating which one you used. We prefer contribution margin with a capped lifetime.
• Payback targets. Bessemer's segment-based targets (12, 18 or 24 months) are looser for enterprise than a blanket 12-month rule. Both benchmarks are older (2021 and 2009 to 2016), and neither was built for AI products with meaningful model costs in COGS, where credits can hide the true margin.
• Pricing below cost. Some argue falling model prices justify pricing slightly lower today; others see it as a risky bet. We lean toward caution, for the reasons below.

"But plenty of giants lost money on every customer"

They did. The 2010 to 2021 era produced famous companies that scaled negative-margin businesses (Uber, ten-minute grocery delivery, scooters) and bet on fixing the economics later. Monzo lost money on each of its first half million or so customers and still turned it around.

But that era ran on cheap capital that is no longer available, and many investors in 2026 share our caution. Monzo's path is instructive for a different reason: it fixed the unit economics and then scaled, not the other way around. Losing money on every customer can be a strategy. More often it's a countdown, and cheap early money tends to hide the ticking (our view on oversized early rounds).

Where we land

Pick a proven model, measure the five numbers honestly, and fix the per-customer math before you pour money on acquisition. We'd rather see a founder with modest growth and clean unit economics than fast growth on a leaking bucket.

That's our view, and it's a default rather than a law. Deep tech, hard tech and some marketplaces may reasonably run negative economics for a while, as long as the path to positive is specific and dated.

A unit economics checklist and common mistakes

One simple check to run before a board meeting or investor update:

UNIT ECONOMICS CHECK: [Month, Year]
Business model (one primary):              _______
Price per customer per month:              $_______
COGS per customer (incl. model/API, hosting, credits at full cost): $_______
Other variable cost per customer:          $_______
Gross margin % / contribution margin %:    ___% / ___%
Paid CAC / blended CAC (to a retained customer): $_______ / $_______
CAC cap per channel / last incrementality test: $_______ / _______
CAC payback (months, gross margin basis):  ___
Monthly logo churn / net dollar retention: ___% / ___%
LTV (state basis: gross or contribution, capped at ___ months): $_______
LTV:CAC:                                   ___x
Best and worst channel by LTV:CAC:         _______ / _______
Last price change and result:              _______

The mistakes we'd steer you away from:

• Counting credits as free. Model and cloud credits are a cash cost that has been deferred.
• Using revenue for LTV. Margin usually gives a truer picture.
• Blending away a bad channel. Organic customers make paid CAC look cheaper than it is.
• Competing on price. A cheaper-than-rivals pitch can give you bad data about whether anyone truly wants the product. Companies such as DoorDash and Airbnb charged a premium.
• Scaling a negative margin. Each new customer can deepen the hole, and your runway math turns against you. Our guide to default alive or default dead shows how quickly that happens.

If your product is validated and you have early traction but your unit economics live in a spreadsheet you rarely open, 1752vc's GTM Accelerator fits that moment. It is a 12-week, hands-on program, remote and self-paced, that teaches founders to sell, recruit, fundraise and build traction, and pricing conversations and sales capacity math like the examples above are often where a founder-led go-to-market motion proves itself or breaks.

The bottom line

A business model is a promise about how money will flow. Unit economics are the proof, one customer at a time.

Growth makes good economics better.

It makes bad economics bigger.

Key takeaways

• In our view it usually pays to copy a proven business model and innovate on the product: SaaS, transactional and marketplace businesses made up 67 percent of Y Combinator's 100 most valuable companies.
• Many founders track five unit economics numbers monthly: CAC, LTV, CAC payback, gross margin and contribution margin, using fixed definitions.
• Consider building LTV from margin, not revenue, and measuring CAC to a retained customer; common targets are LTV about 3x CAC with payback under 12 months, or under 12, 18 or 24 months by segment.
• Price and retention often move unit economics most: a 25 percent price rise in our example lifted LTV:CAC from 4.0x to 5.4x, while doubling churn cut it to 2.0x.
• We would fix negative unit economics before scaling, and treat AI model credits as a real cost.
• One way to enforce CAC discipline is a cap per channel and periodic incrementality tests, because ad platforms tend to understate true CAC.

Frequently asked questions

What are unit economics in a startup?

Unit economics are the revenue and variable costs tied to one unit of a business, usually one customer. They show whether each new customer adds value after the cost of acquiring and serving them. The core numbers are customer acquisition cost, lifetime value, CAC payback, gross margin and contribution margin. Positive unit economics mean growth improves the business rather than deepening losses.

How do you calculate LTV and CAC for a startup?

CAC equals sales and marketing spend in a period divided by the customers won in that period; calculate paid and blended CAC separately. LTV equals monthly revenue per customer, times gross or contribution margin, times expected customer lifetime in months, which is one divided by monthly churn. We suggest using margin, not revenue, and capping the lifetime at three to five years for a conservative figure.

What is a good LTV to CAC ratio?

The widely used rule of thumb, from David Skok, is an LTV of about three times CAC for a viable recurring-revenue business, and the best SaaS companies run higher, sometimes 7x or 8x. In our view the ratio works better as a check than a goal, because early-stage LTV rests on churn estimates. Pair it with CAC payback, ideally under 12 months.

What is the difference between gross margin and contribution margin?

Gross margin subtracts only the cost of goods sold, such as hosting, model API usage and direct delivery costs, from revenue. Contribution margin also subtracts other variable costs that rise with each customer, like payment processing, variable support and shipping. Neither includes fixed costs such as engineering salaries or rent. Contribution margin shows what each customer actually contributes toward those fixed costs.

What is a CAC cap?

A CAC cap is the maximum customer acquisition cost a startup will accept in a channel. The team may scale spend freely while the channel stays under the cap and stops or fixes it when it goes over. Many teams pair the cap with regular incrementality tests, such as switching a channel off in one region, since platform-reported CAC tends to look cheaper than the true cost.

Should a startup scale with negative unit economics?

Usually not, in our view. Many founders fix negative unit economics before scaling, because each new customer adds to losses and investors are now often reluctant to subsidize them. The exception is a deliberate land grab with a credible, specific plan to flip the economics, as Monzo eventually did by bringing technology in house and adding fees. Even then, the plan should name which costs fall and which prices rise.

Sources

• Y Combinator: YC guide to business models (Aaron Epstein)
• Y Combinator: Key Startup Metrics (Tom Blomfield)
• Y Combinator: Consumer Startup Metrics (Tom Blomfield)
• Y Combinator: How To Price For B2B (Tom Blomfield)
• Y Combinator: Most startups are undercharging (Dalton Caldwell)
• Andreessen Horowitz: 16 Startup Metrics
• For Entrepreneurs (David Skok): Startup Killer, the Cost of Customer Acquisition
• For Entrepreneurs (David Skok): SaaS Metrics 2.0, A Guide to Measuring and Improving What Matters
• Bessemer Venture Partners: Scaling to $100 Million
• SaaS Capital: 2026 Spending Benchmarks for Private B2B SaaS Companies
• 20VC: Matteo Franceschetti, Eight Sleep (September 2026)
• 20VC: Andrew Macdonald, Uber (August 2026)