Two frameworks for building consumer and marketplace startups

Executive overview

Most consumer founders optimise for MAUs; most marketplace founders optimise for GMV. Both metrics are vanity — they don't reveal whether you're building something that endures.

Sarah Tavel's two frameworks fix this. The hierarchy of engagement gives consumer founders three ordered priorities: drive the core user action, make the product retentive, then make it self-perpetuating. The hierarchy of marketplaces gives marketplace founders the same structure: focus obsessively on a constrained market, tip that market, then dominate it.

The sequence matters. Each level is a prerequisite for the next.

Skipping levels feels like speed — it is actually the fastest way to build something that won't last.

Hierarchy of engagement — the three levels

• Level 1 — core action: Every successful consumer product has one action that signals a user truly understands the product and is likely to return. For Pinterest it was pinning, YouTube it was subscribing, Snapchat it was sending a Snap.

• The core action is your North Star metric. All product decisions, all new-user experiences, should funnel toward it.

• Identify it bottom-up (which action has the highest next-week return rate?) and top-down (if a user never does this, did they understand the product at all?).

• Avoid vanity actions: the metric must scale, and optimising for it must produce a product you actually want to build.

• MAUs and downloads mean nothing if they don't reflect users completing the core action.

• Level 2 — retention: The product must get better the more it is used, and users must have more to lose by leaving.

• Pinterest's algorithmic home feed: every pin improved your personalised recommendations and grew a board you couldn't replicate elsewhere.

• Evernote: thousands of personal notes create a near-irreplaceable knowledge base — but no growth loops, which caps its ceiling.

• Retention requires measuring cohorts. Track weekly cohorts on two axes: are users returning, and are they completing the core action? Look for the plateau — if cohorts keep declining, retention work isn't done.

• A common mistake: building too many features that diffuse new-user attention away from the core action, preventing activation entirely.

• Pure anonymity consistently breaks level 2. Without a persistent identity, users accumulate no benefits and have nothing to lose by churning. Pseudo-anonymity (Reddit, Twitter) preserves this through a persistent account.

• Level 3 — self-perpetuation: Every user action is kinetic energy; a great product converts it back into a better experience for all users.

• The strongest mechanism is a network effect: each pin on Pinterest enriched the entire recommendation graph.

• Secondary mechanisms are growth loops (pins triggering re-engagement notifications that pull dormant users back) and virality loops (collaborative features, SEO content, referrals).

• House Party and Clubhouse failed here: as the network grew, push-notification volume overwhelmed users and they stopped engaging — the flywheel broke its own mechanism.

• Without level 3, you must pay to acquire every user. TikTok is the rare exception — it spent over $1 billion on paid UA after proving levels 1 and 2, a path almost no consumer company can replicate.

Hierarchy of marketplaces — the three levels

• Level 1 — focus: Concentrate on the smallest market where you can create genuinely happy buyers and sellers — the thimble.

• GMV is a vanity metric in the same way MAUs are. A million dollars of GMV scattered across ten cities is far less valuable than a million dollars of GMV dominating one suburb.

• The goal is happy GMV: transactions where the experience is so much better than any substitute that both sides retain.

• Minimum viable happiness is the threshold — a high enough retention rate to know the product is ready to scale.

• DoorDash went after suburbs (low competition, captive customers); Postmates went after cities, categories, and retail at once. DoorDash won.

• Two scarce resources force the constraint: capital and founder attention. Spreading either kills the ability to reach tipping point.

• Level 2 — tip the market: Reach saturation in your constrained market, then trigger the flywheel that makes growth self-sustaining.

• Two types of loops work together:

  • Growth loops: supply attracts demand, or demand attracts supply, without requiring continuous founder effort. Examples — Recce (restaurant suppliers started sending their own customer CSVs to onboard), Etsy (sellers distributed branded business cards linking to their store), DoorDash (restaurants promoted their own DoorDash listings).
  • Happiness loops: mechanisms that continuously improve match quality as the marketplace scales. Search ranking that rewards sellers who create happy outcomes. Reputation systems that surface quality supply and churn low-quality suppliers.

• Tipping looks like: cohorts improving over time, organic growth starting to appear, suppliers or buyers leaning in without being recruited.

• Not all markets are tippable. Key conditions that enable tipping: fragmented supply (hungry non-incumbents who will lean in), a repeatable transaction (low-frequency categories like plumbing make it hard to own the demand relationship), and low competition in the target segment.

• Homogeneous supply (e.g. mechanical turk-style labour) resists tipping — adding the 5-millionth unit of supply doesn't improve buyer experience, so competitors can replicate the core value quickly.

• Level 3 — dominate the market: Once the playbook is proven in one market, expand aggressively across three vectors simultaneously.

  1. Deepen penetration of the existing market — expand use cases in geographies already tipped (Uber: black cars → UberX → UberPool).
  2. Enter adjacent markets — apply the proven playbook to new geographies or categories as fast as capital and attention allow.
  3. Never rest on dominance — consumer expectations continuously rise. Grubhub dominated food delivery then was overtaken because it couldn't move fast enough to add first-party delivery. HomeAway was overtaken by Airbnb. The history of marketplaces is a history of disruption.

• Market dominance only yields its full economic benefits once you are decisively number one. Being barely ahead of number two means fighting for every point of share with no contribution margin to reinvest.

• Raise and deploy capital in service of growth only after the playbook is proven — blitz-scaling before tipping a single market distributes resources too thinly for any one market to reach escape velocity.

Finding marketplace opportunities and what investors look for

• Three patterns for finding white space: (1) take a horizontal marketplace, identify its lowest-NPS category, and build a vertical around it (eBay → GOAT for sneakers); (2) find a niche no one is paying attention to and prove the market is larger than it looks (HipCamp, early Etsy); (3) change the atomic unit of supply to unlock inventory that was previously unavailable (DoorDash adding delivery to restaurants that previously had none).
• LLMs may open a fourth pattern: automating onboarding of long-tail supply that was previously too costly to reach.
• Underestimated markets are often the most attractive: less competition means more room to tip, and supply-side economics often expand the market as suppliers earn and reinvest (HipCamp hosts upgraded from tent pitches to yurts).
• What signals a strong marketplace opportunity: an earned secret (an insight about why now that others are missing), a founder with genuine focus, and intellectual honesty about what is and isn't working — vanity metrics signal the opposite.
• The "market current" test: a good market has momentum pulling the company forward; the wrong market requires massive effort just to make progress. Market size matters less than the direction and velocity of change.