The Internet & Platforms group covers companies whose primary asset is a digital network of users and/or transacting participants rather than physical inventory or factories: search, social media, online advertising, e-commerce marketplaces, ride-hail/delivery, app stores, streaming, and "super-apps." Its defining economic feature is the combination of near-zero marginal cost to serve an additional user with network effects — the product becoming more valuable as more people use it. That combination is what produces the sector's signature financial shape (low gross-margin cost of goods, very high operating leverage at scale, and the potential for durable competitive moats) and its signature investment tension: the same dynamics that can entrench a dominant platform also make it a permanent target for regulators and for the next platform shift. The analytical job in this sector is mostly non-financial — you must read user and engagement data, not just the income statement, because revenue often lags the network.

The business models

Platforms monetize a user base in a handful of recognizable ways, and a single company often blends several:

• Advertising (Google Search/YouTube, Meta) — monetizes attention; revenue ≈ impressions × price per impression, sensitive to the ad cycle and to ad-targeting rules.
• Transaction / marketplace take rate (Amazon 3P, Uber, Etsy, app stores) — the platform skims a percentage ("take rate") of gross transaction value (often called GMV/bookings).
• Subscription (Netflix, Spotify, SaaS platforms) — recurring fee per user; the cleanest, most predictable revenue.
• Direct e-commerce / first-party retail (Amazon 1P) — low-margin retail bolted onto a high-margin platform.

A key structural distinction is one-sided networks (a tool/social graph that improves with more same-side users, e.g. a messenger) versus two-sided marketplaces (buyers and sellers, riders and drivers) that exhibit cross-side network effects and must solve a "chicken-and-egg" cold-start on both sides at once (a16z; Version One social handbook).

The metrics that matter

Because revenue trails the network, analysts track operating KPIs the company reports alongside GAAP results:

• MAU / DAU — monthly and daily active users: reach and stickiness. The DAU/MAU ratio proxies engagement frequency; a commonly cited investor rule of thumb (originating with Sequoia) treats ~20% as the threshold for a "sticky" consumer app, though benchmarks vary widely by category — social/messaging apps often run far higher (frequently cited at 50%+) while e-commerce/fintech run lower (roughly 15–30%) (platform-metrics primers, AppsFlyer/Sequoia). Treat any single band as a rule of thumb, not a hard line.
• ARPU (average revenue per user = revenue ÷ users) — the monetization side. A platform's value is roughly users × ARPU × durability; rising ARPU on a flat user base signals monetization maturing (Stripe).
• GMV / gross bookings and take rate — for marketplaces, the volume flowing through and the slice retained.
• Cohort retention / churn, and the "power-user curve" — whether newer cohorts are as valuable as older ones (a sign network value is compounding, not decaying).
• Rule of 40 — for the SaaS-flavored platforms, revenue-growth % + profit-margin % ≥ 40 is a widely used balance test (popularized by VC Brad Feld in 2015); SaaS firms clearing it have historically commanded materially higher EV/Revenue multiples (Brad Feld; Software Equity Group). The 40 threshold is admittedly arbitrary — it is a heuristic, not a valuation model.

How it's used in practice

Investors value these names primarily on growth and forward cash-flow potential, not trailing earnings — early or reinvesting platforms are often unprofitable by design. Common approaches: EV/Sales and EV/Gross-Profit multiples for high-growth names; value-per-user / user-based comps for under-monetized platforms (value the users, assume they monetize toward mature-peer ARPU over time); and eventually DCF / EV/EBITDA / P/E as the model matures. The central diligence question is moat durability: how strong are the network effects, how high are switching costs, and how much multi-tenanting (users on several competing platforms at once) erodes lock-in (a16z). Operationally, platform stocks are also high-beta, sentiment-driven, and earnings-event-heavy — large gaps on KPI surprises (a MAU miss or ARPU deceleration) are routine, which matters for any timing-sensitive strategy.

Standing & evidence

Network effects are real and economically large — the most frequently cited estimate, from the VC firm NFX, found that of 336 digital companies founded since 1994 that reached >$1B in value, only ~35% had network effects at their core, yet those companies accounted for roughly 70% of the value created in that set (NFX, "70% of Value in Tech is Driven by Network Effects"). Note the scope: this is value among billion-dollar digital startups, not literally all global tech equity — a distinction often lost when the stat is repeated. But two further beliefs deserve skepticism:

1. "Winner-take-all" is overstated. Many platforms are local (ride-hail resets city by city), multi-homed (users run TikTok and Instagram), or vulnerable to a platform shift (desktop→mobile, web→AI). Dominance is conditional, not guaranteed. 2. Network-effect language is often used to dignify ordinary scale economies. True network effects (the product improves for users as the network grows) are distinct from mere supply-side scale (cheaper unit costs). The a16z framework exists precisely because the term is loosely applied.

The sector's largest external risk is now regulatory and antitrust, and it is live, not theoretical: a U.S. federal court (Judge Mehta, D.D.C.) found Google an illegal monopolist in general search on Aug 5, 2024; in a separate case a court (Judge Brinkema, E.D. Va.) found in April 2025 that Google also illegally monopolized parts of the ad-tech stack (publisher ad server + ad exchange), with a remedies trial held in late 2025 in which the DOJ sought an AdX ad-exchange divestiture — the remedy ruling was still pending as of mid-2026; the FTC has ongoing monopolization litigation against Amazon; and the EU's Digital Markets Act designated six "gatekeepers" in Sept 2023 (Alphabet, Amazon, Apple, ByteDance, Meta, Microsoft; Booking added 2024) with presumption-of-harm obligations and multi-billion-euro fines already levied (DOJ/FTC filings; EU DMA gatekeeper portal; case reporting). The direction of remedies — and the disruptive potential of generative AI to the search/advertising funnel — are the dominant unknowns for the group's largest names.

Strengths & limitations

Where the model shines: extreme operating leverage (incremental users cost almost nothing), defensible moats when network effects are genuine and switching costs high, and asset-light compounding of cash flow once the network is built. Where it breaks: cold-start failure (a marketplace that never reaches liquidity), multi-homing and low switching costs (commoditized engagement), regulatory unbundling of a monopoly, and platform shifts that strand an incumbent (the recurring history of the sector). The single most common analytical misuse is treating user growth as value without checking monetization (ARPU) and retention — a fast-growing user base that doesn't monetize or retain is a cost center, not a moat. A second is accepting "network effects" as a stated fact rather than testing for them in cohort and engagement data.

System relevance

This is a sector playbook node. For company-specific investment write-ups, the Delvantic stock-detail "coin" system (/stocks/companies/?t=TICKER) is where a particular platform's thesis lives; this node supplies the shared lens (KPIs, monetization model, moat/regulation framing) the analysis pipeline can apply across tech-sector names. For the Augustus trade-setup agent, the load-bearing caveat is behavioral, not fundamental: platform stocks are high-beta and KPI-event-driven, so binary outcomes around earnings (MAU/ARPU surprises) and antitrust headlines can dominate technical setups — these names need an event-risk check before any timing-based entry. Valuation siblings (EV/Sales, Rule of 40) and the broader Technology node carry the cross-sector mechanics; this node should not duplicate them.

Sources

• Andreessen Horowitz — "16 Ways to Measure Network Effects" (network-effect types, multi-tenanting, engagement metrics)
• Stripe — "What Is Average Revenue Per User (ARPU)?" (ARPU definition and investor use)
• AppsFlyer / Sequoia Capital — DAU/MAU stickiness benchmarks (~20% threshold; category variation)
• NFX — "70% of Value in Tech is Driven by Network Effects" (study of 336 >$1B digital companies founded since 1994; ~35% had network effects, accounting for ~70% of value created — scope is billion-dollar digital startups, not all global tech equity)
• Brad Feld — "The Rule of 40% for Healthy SaaS Companies" (2015 origin); Software Equity Group — Rule of 40 correlation to EV/Revenue multiples
• Version One — Social Handbook (one-sided vs two-sided networks, cross-side effects)
• DOJ — U.S. v. Google (search, D.D.C., Aug 2024 liability finding) and U.S. v. Google (ad tech, E.D. Va., Apr 2025 liability finding); FTC v. Amazon; EU Digital Markets Act gatekeeper portal (six gatekeepers designated Sept 2023) — note: ad-tech remedies and other outcomes remained unresolved/contested as of mid-2026