Business Models

A business model is the answer to a single question: how does the company make money from the value it delivers? The answer matters because everything downstream — valuation multiples, growth strategy, hiring patterns, sales motion, financial planning, what counts as a healthy quarter — depends on the model. Two companies in the same industry with different business models behave like different species.

This guide walks through the five business-model archetypes most modern companies fit into (or hybrid across): SaaS, marketplace, e-commerce / DTC, services, and ads-supported. By the end, you should be able to read a company description, classify the model (or models), and know which metrics matter most for each kind.

It is not a guide to designing a new business model. It is a guide to understanding the model any specific business is running — which, for an operator or investor or job candidate, is most of what determines whether the business actually works.

We'll move in six steps. The five archetypes at a glance. SaaS — the dominant tech-industry model. Marketplace — and why two-sided is the hardest cold start. E-commerce and DTC — and the unit-economics trap most fall into. Services — the model the tech press undervalues. And ads-supported — the original internet model, currently under pressure from AI.

§1 The five archetypes at a glance Foundational

Most companies fit one of five primary business-model archetypes. Many companies actually run hybrids — a SaaS company with a services arm, an e-commerce business with an ads-revenue side-line, a marketplace with subscription-tier features. But each component fits one of the five basic shapes.

SaaS (Software-as-a-Service). Customer pays an ongoing subscription (monthly or annual) for access to software the company hosts and maintains. The seller delivers software updates continuously; the customer pays continuously. Revenue is recurring, predictable, and (when growing) compounding. Examples: Salesforce, Slack, Notion, almost every B2B tool you've used since 2015.

Marketplace. The company connects two distinct sides (buyers and sellers, riders and drivers, hosts and guests) and takes a cut of each transaction. The company doesn't own the supply or the demand — it owns the matching, the transaction infrastructure, and the trust layer. Examples: Uber, Airbnb, Etsy, eBay, Upwork, DoorDash.

E-commerce and DTC (Direct-to-Consumer). The company sells physical products directly to customers, usually online, often bypassing traditional retail. Revenue per transaction is one-time (with repeat-purchase patterns varying by category). Examples: Warby Parker, Allbirds, Casper, Glossier, almost any Shopify-powered brand.

Services. The company sells labor or expertise — consulting, agency work, professional services, custom development. Revenue scales with hours billed or projects delivered. Examples: McKinsey, Accenture, every law firm and accounting firm, custom software shops, marketing agencies.

Ads-supported. The company gives away its primary product to end users and monetizes by selling access to those users' attention or data to advertisers. Examples: Google Search, Meta (Facebook/Instagram), YouTube, ad-supported streaming, most news sites.

A handful of high-leverage cross-cutting points:

The valuation multiples differ wildly. SaaS businesses with high net revenue retention (NRR — see §2) trade at the highest multiples (10-20x revenue historically; lower in 2024-26 markets but still richest in the set). Marketplaces at scale trade at high multiples too (Network effects justify it). E-commerce trades at low multiples (1-3x revenue typically) because margins are structurally thinner. Services trade lowest (often <1x revenue) because the work doesn't scale beyond billable hours. Ads-supported varies widely depending on audience moat. The business model determines the multiple bracket; execution determines where in the bracket the company lands.

The unit economics work differently per model. SaaS focuses on CAC, LTV, payback period. Marketplaces focus on take rate and frequency. E-commerce focuses on contribution margin and repeat purchase rate. Services focuses on utilization and bill rate. Ads focuses on CPM/CPC and audience scale. Apply the wrong metric set to the wrong model and the analysis is wrong on contact.

Hybrids are common and underrated. Most SaaS companies have a services revenue line (implementation, professional services). Most marketplaces have advertising revenue (sellers paying for placement). Most e-commerce companies have subscription elements. The pure-play model is the exception; the hybrid is the rule, and reading the model means decomposing each revenue line into which archetype it fits.

A diagram showing the five archetypes side-by-side with revenue mechanic, dominant metric set, and example companies makes the differences land at a glance.

Related glossary: SaaS, marketplace, DTC, ads-supported, business model, valuation multiple.

§2 SaaS — the dominant B2B tech model Building

SaaS — Software-as-a-Service — became the default model for B2B software companies in the 2010s and remains so. The customer pays a recurring fee for access to software the vendor hosts, maintains, and updates. The customer never installs anything, never owns a license, and stops paying when they stop using.

The economics of SaaS are different from previous software-business shapes (perpetual-license + maintenance, on-premise installations) in ways that drive how SaaS companies are built, valued, and run:

Recurring revenue. The customer pays each month or year, and the revenue recurs as long as the customer keeps subscribing. This makes revenue predictable in a way one-time software sales never were. A $1M ARR (annual recurring revenue) customer is worth roughly $1M of revenue per year ongoing — easier to forecast, easier to finance, easier to value.

The customer-lifetime focus. Because revenue is recurring, the value of a customer isn't the initial sale — it's the entire stream of payments over the customer's lifetime. This shifts the operating focus from "close the deal" to "close the deal AND keep the customer." The customer success function (covered in Org & Roles §1) exists primarily because SaaS made retention a board-level metric.

The growth-vs-profit trade-off becomes deliberate. A SaaS company can spend aggressively on sales and marketing knowing that each new customer will generate recurring revenue for years. So the rational play, when growth is healthy, is to spend hard early and accept losses in exchange for share. This is why so many SaaS companies were unprofitable through their high-growth years — and why investors valued them on revenue growth instead of profit.

The metrics that matter for SaaS:

ARR (Annual Recurring Revenue) — the annualized value of all subscription contracts currently active. MRR (Monthly Recurring Revenue) is the same idea expressed monthly. ARR is the single most-quoted SaaS number.

NRR (Net Revenue Retention) — what percentage of last year's customer cohort is still generating revenue this year, adjusted for upsell and downsell. NRR above 100% means existing customers are paying more this year than last (because of expansion or price increases) even after churn. NRR above 120% is excellent. NRR below 90% means even rapid new-customer growth might not offset losses from the existing base.

CAC (Customer Acquisition Cost) — total sales and marketing spend divided by number of new customers acquired. The cost to land a customer.

LTV (Lifetime Value) — total revenue a customer generates over their lifetime, minus the cost to serve them. LTV / CAC ratio of 3x or higher is the conventional health benchmark — meaning each dollar spent on acquisition returns $3 of lifetime value.

Payback period — months until a new customer's revenue covers their CAC. Under 12 months is considered healthy for most SaaS; under 6 months is exceptional. Long payback periods (24+ months) mean the company is running on growth capital and can't survive a funding slowdown.

Rule of 40 — revenue growth rate + EBITDA margin should sum to 40 or more. A company growing 50% with -10% margins (sum: 40) and a company growing 20% with 20% margins (sum: 40) are both considered healthy by this metric. Below 40 signals trouble; above 40 signals strength.

The honest read on the SaaS model in 2026: still the highest-margin software model when it works, but the bar has risen. The 2021 era of "grow at any cost" valuations is gone. Investors now want growth AND a credible path to profitability — the Rule of 40 is the shorthand for that bar. The SaaS companies thriving in 2026 are the ones that hit Rule of 40 without sacrificing NRR; the ones in trouble are the ones whose 60% growth was masking 130% spend.

Related glossary: SaaS, ARR, MRR, NRR, CAC, LTV, payback period, rule of 40, churn.

§3 Marketplace — and the cold-start problem Building

A marketplace business owns the connection between two distinct sides — typically buyers and sellers (Etsy), riders and drivers (Uber), hosts and guests (Airbnb), workers and gigs (Upwork), restaurants and diners (DoorDash). The company itself doesn't supply the demand or the supply — both sides are external participants. The company supplies the matching, the transaction infrastructure, the trust layer, and (when working) the network effects that make both sides want to be there because the other side is there.

The defining metrics:

Take rate. The percentage of each transaction that the marketplace keeps. Take rates range widely — Airbnb takes ~15% per booking; eBay takes ~10%; Uber takes ~25% from riders; Etsy takes ~6.5%; OpenTable takes per-reservation fees that are tiny percentages of meal value. Take rate determines unit economics but is constrained by what either side will tolerate before defecting.

GMV (Gross Merchandise Value) or GTV (Gross Transaction Value) — the total value of transactions flowing through the marketplace, not what the marketplace keeps. GMV is the headline number marketplaces quote because it's larger than revenue (which is GMV × take rate). The distinction matters: a $10B GMV marketplace with a 10% take rate has $1B in revenue. Confusing the two is a common mistake reading marketplace headlines.

Liquidity. How fast a buyer can find what they want and how fast a seller can find a buyer. High liquidity means listings sell quickly and searches return relevant matches. Marketplaces with low liquidity die — buyers don't find what they want, leave, and sellers follow. Liquidity is the marketplace's single most important operational metric and the hardest to manufacture in early stages.

Frequency. How often a typical user transacts. High-frequency marketplaces (Uber, DoorDash — multiple times a week) have better retention and lower per-transaction acquisition costs than low-frequency marketplaces (Airbnb — a few times a year). High-frequency marketplaces also build behavioral habit faster.

The cold-start problem is the defining structural challenge of marketplace businesses. A marketplace is only valuable when both sides are present at scale. But on day one, neither side wants to join — buyers won't come without sellers, sellers won't come without buyers. Solving the cold-start problem is what separates marketplaces that work from the 95%+ that fail in their first two years.

The standard cold-start strategies:

Subsidize one side. Pay one side to show up until the other side organically follows. Uber subsidized drivers heavily in launch markets; the demand side caught up. Expensive but proven.

Single-player mode. Make the product useful for one side even without the other side present. Yelp launched as a restaurant-review site (useful for diners alone, before any restaurants engaged). Etsy launched on top of a community of crafters who already had loyal followings (useful for sellers alone, before any buyers).

Launch hyper-locally. Saturate one neighborhood / one campus / one vertical before expanding. Uber and DoorDash both did this — get density in one place where both sides see liquidity, then expand. Tinder famously launched on a single college campus.

Pre-aggregate one side. Sign up sellers (or buyers) before launching publicly, so day one already has supply (or demand) at scale. This is hard to execute but it's how some B2B marketplaces have launched successfully.

Once a marketplace is over the cold-start hurdle, its economics become extraordinary because of network effects — each new participant on either side makes the marketplace more valuable for participants on the other side. These network effects, when they kick in, create durable moats and high valuation multiples. The reason marketplace valuations are so high at scale is the same reason most marketplaces fail before they get to scale: the network effect either compounds or never starts, with not much in between.

Related glossary: marketplace, take rate, GMV, liquidity, network effects, cold-start problem.

§4 E-commerce and DTC — the unit-economics trap Building

E-commerce is the sale of physical goods online. DTC (Direct-to-Consumer) is a flavor of e-commerce that bypasses traditional retail intermediaries — the brand sells directly to the end customer through its own channels. Warby Parker selling glasses through warbyparker.com instead of through optometrist offices was the canonical DTC move; the model expanded across mattresses (Casper), razors (Harry's), cosmetics (Glossier), and dozens of other categories in the 2010s.

The mechanic is straightforward: customer visits site, places order, brand ships product, customer pays. Revenue per transaction is one-time (with repeat purchases varying by category). The challenge is that the unit economics are unforgiving in ways that aren't obvious until the math is run.

The critical metrics:

Gross margin. Revenue minus the cost of the physical product (COGS). Healthy e-commerce gross margins range from 40% (low-end consumer goods) to 80% (premium beauty, supplements). Mattress DTCs run 60-70%; apparel DTCs 50-65%; food DTCs 30-50% (perishables are brutal).

Contribution margin. Gross margin minus the variable per-order costs — shipping (often the brand pays inbound and outbound), fulfillment, payment processing, packaging, returns. Contribution margin is what the brand actually keeps per order before fixed costs. A 70% gross margin product can have a 30% contribution margin after free shipping, returns processing, and payment fees.

CAC. Cost to acquire a new customer. DTC CAC ballooned in the late 2010s and early 2020s as Facebook and Google ad auctions priced in the dozens of competing DTC brands. CAC of $80-150 became common for what used to be $20 categories.

Repeat purchase rate. What percentage of customers buy again within a year. Subscriptions or replenishment categories (razors, vitamins, pet food) have high repeats; one-time categories (mattresses, eyewear) don't. Repeat rate determines whether CAC is a one-shot cost or an investment that pays back over time.

Contribution-margin payback — the number of orders required to recoup CAC. A DTC brand with $120 CAC and $35 contribution margin per order needs 3.4 orders per customer to break even on acquisition. If the average customer only orders 1.8 times in their lifetime, the unit economics are negative — every customer acquired is a loss, paid for out of growth capital.

The unit-economics trap is what killed many of the 2010s DTC brands. The pitch (sell direct, cut out retail, capture the margin) was structurally sound — but the margin gain from cutting out retail was less than the CAC required to acquire a customer at scale on saturated digital ad channels. Companies that thought they had 70% gross margins discovered they actually had 30% contribution margins, and customers who they thought would repeat-purchase actually didn't. Several high-profile DTC brands raised hundreds of millions, hit nine-figure revenue, and either went private at fire-sale prices or shut down because they could never get the unit economics to work.

The DTC brands that have thrived in 2026 share patterns: they sell high-margin replenishable categories (so repeat purchases compound CAC payback), they own a brand strong enough to sustain organic / referral acquisition (lowering reliance on paid ads), and they expanded into omnichannel retail (Warby Parker is now in malls; Glossier is in Sephora) — accepting some retail margin loss in exchange for a more diversified acquisition funnel. The pure-DTC, pay-Facebook-for-every-customer model is structurally fragile and most companies running it haven't survived.

Related glossary: DTC, CAC, contribution margin, AOV, repeat purchase rate, LTV.

§5 Services — the model the tech press undervalues Building

Services businesses sell labor or expertise — consulting, agency work, custom software development, professional services, accounting, legal, marketing, design. Revenue scales with hours billed (the classic model) or with projects delivered (the productized variant). The output is variable and customized; the unit of production is human time.

Services is the largest business model on earth by employment and one of the smallest by tech-industry attention. McKinsey is a services business. Accenture is a services business. Every law firm and accounting firm is. So is every marketing agency, consulting firm, and dev shop. Services businesses sustain the economy, employ millions, and quietly produce enormous wealth. They get less press than tech because their models scale with people instead of with code — but the businesses themselves are real and durable in ways tech startups often aren't.

The critical metrics:

Bill rate. What the firm charges per hour for a given level of staff (junior, mid, senior, partner). Bill rates range wildly — a junior associate at a regional consulting firm might bill at $150/hour; a senior partner at a major firm might bill at $1,500/hour or more for high-stakes engagements.

Utilization rate. What percentage of a billable employee's available hours are actually billed to clients. 60-75% is typical for healthy services firms — the rest goes to internal work, sales, training, bench time between engagements. Pushing utilization above 80% sustainably is hard; under 60% means underbilled capacity (which is essentially carrying inventory in a services business).

Realization rate. What percentage of billable hours actually got billed at full rate after discounts, write-downs, and bad-debt write-offs. A 90% realization means for every $100 of hours worked, $90 actually became invoiced revenue.

Gross margin (services-style). Revenue minus direct labor cost (the consultants' salaries plus benefits). Healthy services gross margins run 35-50%. Higher-end strategy consulting can reach 60%+; commodity body-shop work runs 15-25%.

Project pipeline coverage. Multiple of the next quarter's revenue target that's in the active pipeline. Services businesses live on the pipeline — without 3-4x coverage, the next quarter's revenue is at risk.

The structural challenge of services is scalability ceiling. Revenue grows roughly linearly with headcount, because revenue requires billable hours and hours require people. A services firm at $50M with 200 consultants can't double revenue without roughly doubling consultants — which means doubling recruiting, real estate, management overhead, and culture-sustaining effort. This is why services businesses don't get SaaS-like valuation multiples — the growth is structurally bounded.

The two paths services firms take to address this:

Productize. Take repeatable engagements and turn them into fixed-scope, fixed-price packaged offerings. A custom website project becomes a 6-week "Website-in-a-Box" with a standard methodology and price. Productization improves margins by eliminating scope creep and turns each project into a more predictable revenue unit. Many "services" firms in 2026 are actually services + productized services hybrids.

Build product. Use services revenue to fund the development of an actual software product, then transition over time from services-dominated to product-dominated. Many of today's enterprise software companies started as services firms (custom-software shops) and built software products from the patterns they kept seeing.

The honest read on services in 2026: it remains the right model for high-judgment, high-context, custom work — the work that AI hasn't displaced (and based on §6 of the AI Literacy guide, won't displace at the high end anytime soon). The bottom layer of commodity services (basic content writing, simple coding, basic data entry) is being squeezed by AI; the top layer of strategic advice and complex custom work is growing because the strategic questions AI surfaces still need humans to answer.

Related glossary: services, utilization rate, bill rate, realization rate, pipeline coverage, productized services.

§6 Ads-supported — the original internet model Building

Ads-supported businesses give away their primary product to end users and monetize by selling access to those users to advertisers. The model is the oldest internet business model — predating SaaS, predating marketplaces — and has built some of the most valuable companies on earth (Google's parent Alphabet and Meta both run on it). It's also the model under the most structural pressure in 2026.

The mechanic: build a product users want, attract a large audience, sell advertisers the ability to reach that audience with targeted messages, take revenue per impression, click, or conversion. The user is the product (in the often-quoted phrase); the customer is the advertiser.

The critical metrics:

CPM (Cost Per Mille). Cost per 1,000 ad impressions. The bedrock unit of digital advertising pricing. Premium content + valuable audience = higher CPMs (a finance website might get $50 CPMs; a generic content site might get $2-5).

CPC (Cost Per Click). Cost per ad click. Used for performance-oriented advertising where the goal is driving traffic. Search ads (Google, Bing) and a chunk of social media advertising run on CPC pricing.

CPL (Cost Per Lead) and CPA (Cost Per Acquisition). Cost per qualifying action — a sign-up, a purchase, an app install. Used when the advertiser only pays for conversions, not just exposures. CPA pricing transfers more risk to the publisher (they only get paid if conversion happens) and is typically priced at a premium per event.

ARPU (Average Revenue Per User). Revenue per user per period. Determines how much value the platform extracts from each user's attention. Mature ads-platforms — Meta, Google — have ARPU in the $200+/year range in developed markets. Smaller platforms struggle to push ARPU above $20-30/year.

Engagement metrics. Daily active users (DAU), monthly active users (MAU), session length, sessions per user. Ads-supported revenue scales with attention, so these metrics drive valuation. The DAU/MAU ratio (proportion of monthly users who also use daily) signals stickiness — Facebook's famously hit 60%+ in its prime, most apps run 10-30%.

The historical advantage of the ads-supported model is massive scalability with low marginal cost. Once the platform exists, serving an additional ad costs essentially nothing. This is why a single search-engine company could generate hundreds of billions in revenue at gross margins above 80% — the marginal cost per ad served is fractions of a penny.

The pressures on the model in 2026:

Privacy regulation and platform-level signal loss. Apple's App Tracking Transparency (2021) cratered the targeting precision that ad-funded apps relied on. GDPR, CCPA, and successor regulations continue to constrain audience-level data collection. The result: lower-quality targeting, lower CPMs for non-platform-owned advertising, and a concentration of revenue at the platforms (Google, Meta) that have first-party data and don't need third-party cookies.

AI-generated content saturation. As generative AI makes content trivially cheap to produce, ad inventory expands faster than ad demand. The supply of "place to put an ad" is growing; the supply of advertisers and ad budgets is roughly flat. The math says CPMs decline. This pressure is hitting open-web publishers hardest.

AI-mediated discovery. Search engines historically sent users to publisher websites, which then served the user ads. AI chatbots increasingly answer the question directly without sending the user anywhere — short-circuiting the publisher-ad-impression chain. This is an existential pressure on news and reference publishers; AI agents in 2026 increasingly answer questions without surfacing the source publisher's brand or generating an ad impression for them.

Walled-garden consolidation. Most ad revenue concentrates at platforms with logged-in users (Google, Meta, Amazon, TikTok). The open-web publisher slice keeps shrinking. Smaller publishers either consolidate, pivot to subscription / membership models, or shut down.

The honest read on ads-supported in 2026: still the dominant model for the consumer-internet giants, but increasingly hard for everyone else. The viable shape for new entrants is one of three things — own a niche audience with high CPM and direct relationships (specialized publishers), be a platform large enough to compete in the walled-garden tier (hard to start), or hybrid with subscription so ad revenue is supplementary rather than load-bearing.

Related glossary: ads-supported, CPM, CPC, CPL, CPA, ARPU, DAU, MAU.

§7 Network effects and platform flywheels Strategic

A network becomes more valuable as more people use it — this is the core claim of network effects, and it's one of the most powerful forces in technology economics. The classic illustration is the telephone: one telephone is useless, two creates one connection, ten creates 45, and a million creates roughly 500 billion potential connections. Value scales with the square of users, which is the intuition behind Metcalfe's Law. The practical implication for business: once a network reaches critical mass, each new user makes the network more attractive for every existing user, which attracts more users, which makes it more attractive — a self-reinforcing loop that's very hard to interrupt from the outside.

Direct network effects operate within a single user group: more users directly improve the product for every other user. Phone networks, messaging apps, and marketplaces for social connection are the canonical cases. Indirect (cross-side) network effects operate across two distinct user groups where each side's growth makes the product more valuable for the other side, not for themselves. The classic example is a ride-share platform: more drivers reduce wait times for riders, which attracts more riders, which creates more earning opportunities for drivers, which attracts more drivers. Drivers don't benefit from more drivers (they compete for the same rides); riders don't directly benefit from more riders. But each side benefits from the other growing. This is the structural dynamic that gives two-sided platforms — marketplaces, app stores, payments networks — their particular competitive durability.

The flywheel is the visualization of this compounding loop in action. Amazon's flywheel, famously sketched by Jeff Bezos, starts with lower prices attracting more customers, more customers attracting more sellers, more sellers expanding selection, better selection reinforcing the lower-cost structure through scale. Each rotation of the wheel makes the next rotation slightly easier and faster. The value of a flywheel is that no single turn is decisive — what matters is that every turn reinforces every other turn. This is why platform businesses that have achieved scale are so difficult to compete with: it's not just that they're bigger, it's that their size is continuously compounding their advantages.

Why platforms are hard to displace once network effects are firmly established comes down to the economics of switching. Switching off a network doesn't just mean accepting a worse product for yourself — it means losing the connections and value that exist because others are on the network. You aren't just switching products; you're coordinating everyone you interact with to switch simultaneously, which is a fundamentally different and much harder problem. This is why CAC for challengers fighting an entrenched network is so high — you aren't just paying to acquire a user, you're paying to compensate them for the network value they're leaving behind.

Flywheel Each turn compounds the next

Related glossary: network effects, platform, marketplace, flywheel, CAC

§8 What to remember

Six things to carry:

1. Five archetypes cover most companies. SaaS, marketplace, e-commerce/DTC, services, ads-supported. Most real businesses are hybrids; decomposing each revenue line into archetype is how you actually read a company.
2. SaaS valuation depends on growth + retention, not just revenue. Rule of 40, NRR, and payback period are the bar. 60% growth burning 130% spend isn't a SaaS success; it's a funding bet.
3. Marketplace economics are extraordinary at scale and brutal at start. The cold-start problem kills most marketplaces. The ones that survive launched hyper-locally, vertically, or by subsidizing one side.
4. DTC has a unit-economics trap. High gross margins don't equal high contribution margins after shipping/returns/CAC. Many 2010s DTC brands never escaped the trap.
5. Services scales with people, which caps growth and multiples. The two escape paths are productization (most common) and building a product (most ambitious). Both are real; both are hard.
6. Ads-supported is under structural pressure. Privacy regulation, AI-mediated discovery, and walled-garden concentration are squeezing everyone except the largest platforms. New entrants in 2026 mostly need a hybrid model.

The lens that ties them together: business models aren't decorations on a strategy — they ARE the strategy. Every decision a company makes (who to hire, what to measure, how to grow, when to raise) is downstream of the model. Reading the model is the first step in reading the company.

§9 Related Glossary terms

Archetypes: SaaS, marketplace, DTC, ads-supported, services, business model.

SaaS metrics: ARR, MRR, NRR, CAC, LTV, payback period, churn, rule of 40.

Marketplace: take rate, GMV, liquidity, network effects, cold-start problem.

E-commerce/DTC: contribution margin, AOV, repeat purchase rate.

Services: utilization rate, bill rate, realization rate, pipeline coverage, productized services.

Ads: CPM, CPC, CPL, CPA, ARPU, DAU, MAU.

Cross-cutting: valuation multiple, unit economics, moat.

See also: Financial Literacy Guide — the P&L, cash flow, and margin mechanics underlying each business model archetype; FinOps & Unit Economics Guide — per-unit economics, CAC payback, and build-vs-buy calculus that determine whether a business model works at scale.