Building a growth model and marketplace growth strategy

Executive overview

Most companies lack a structured way to compare product investments or understand which levers actually move the business. A growth model — an analytical representation of how a business grows, built in a spreadsheet — forces that understanding and creates a common currency for resource allocation.

For marketplaces, the same model applies but with added complexity: both supply and demand must be modelled, their interaction is hard to quantify, and second-order effects routinely surprise even experienced operators.

The core insight: retention is almost always the biggest lever in the model, and the best way to improve it is through early lifecycle experience — not re-engagement campaigns.

Building a growth model

• A growth model is not a forecasting tool — it is an opportunity assessment tool.
• 50% of the value comes from the act of building it; it forces genuine understanding of how the business works.
• Three core building blocks apply to almost every business: acquisition channels, retention, and monetization.
• Transactional businesses add unit economics (COGS, take rate); marketplaces add supply acquisition, retention, and supply-demand interaction.
• The most dangerous failure mode: stacking too many assumptions, especially in marketplaces where supply-demand interaction is hard to isolate.
• Preferred approach: one simple top-level conceptual model plus separate mini-models for each team or pod, each with its own north star and input levers.
• Use the model in annual planning as a common currency: teams run their proposed bets through it to generate comparable impact estimates.

Retention as the dominant lever

• Growth models consistently reveal that retention sensitivity is far higher than intuition suggests — small gains in retention compound through referrals, contribution margin, and LTV.
• Resurrection efforts (targeting churned users) almost always underperform; focus on new users instead.
• The biggest retention wins come from inflecting the early user experience — not from push notifications or re-engagement.
• Look for variability in the first week or month: identify which customers are having a below-average experience due to bad luck, and homogenise it upward.
• Uber and Lyft's first-week earnings guarantees are an example of this — smoothing early driver experience to prove what the platform is worth.
• Correlational analyses ("our best users do X, so let's make others do X") rarely work; focus on understanding real value drivers instead.

Marketplace health metrics

• GMV or transaction volume: the ultimate north star that brings both sides together.
• Unit economics: understand the per-transaction contribution margin; early marketplaces often have negative unit economics (Instacart, Uber).
• Liquidity: how reliably can each side do the thing they came to do? Express as a dimension customers care about — wait time for Uber, fill rate or conversion for commerce. Until the marketplace is liquid, nothing else matters.
• Share of wallet: what percentage of a buyer's total spend (or a seller's total business) flows through your marketplace? Depth beats breadth — a higher share of wallet predicts retention and reduces multi-tenanting.

Supply vs. demand prioritisation

• People systematically over-rotate on supply; demand is the only thing that ultimately matters.
• Demand is the currency: any supplier will accept customers delivered at a profitable rate.
• Supply acquisition is only justified to the extent it improves demand — more supply stops being valuable once it no longer increases liquidity for the buyer.
• Build dual-sided ROI equations that load supply CAC onto the demand CAC at the appropriate ratio; this lets you push acquisition to your payback threshold without relying on balance ratios.

Marketplace expansion

• TAM matters less than adjacency beyond a certain scale — focus on how close the expansion is to the current model, not just how big it is.
• Prioritise expansions where the same supply base or customer base serves multiple use cases (e.g. Uber Eats uses the same drivers and targets the same riders).
• Go-to-market must stay in lockstep with product — the race to liquidity is won by whoever delivers the best end-to-end experience first, not whoever spends the most on incentives.
• A small set of cohorts with great retention is more fundable and more defensible than high GMV with weak engagement.

Vertical vs. horizontal marketplaces

• Unbundling is consistently overhyped; most vertical spin-offs fail because they sacrifice the LTV benefits and network effects of scale.
• Vertical marketplaces can succeed when: (a) the sub-segment has a self-contained network, (b) the category has very high order value or high frequency, and (c) the horizontal incumbent has a trust or quality problem the vertical can solve.
• Competing on UX alone against a horizontal marketplace rarely works — the horizontal wins on SEM bidding power and cross-sell LTV.

Marketplace business model evolution

• Newer marketplaces charge higher commissions and do more work to justify them: lead-gen (5–10%) → managed marketplace (trust layer, Airbnb/Etsy) → heavily managed (logistics or financing, DoorDash/Faire).
• The logical endpoint of this trend is owning supply entirely — at which point it is no longer a marketplace.
• Marketplaces where supplier creativity matters (Etsy, Faire, Steam) will stay in marketplace mode longer; commoditised experiences (ride-sharing) will consolidate toward full ownership once autonomous supply arrives.
• A useful test: is there still a direct transaction between supply and demand? If not (e.g. Opendoor), it is no longer a marketplace.

B2B marketplaces

• B2B marketplaces are fewer partly because most founders are consumers and don't recognise B2B pain points.
• The structural constraint is fragmentation: concentrated markets (few large suppliers) have less need for a marketplace and more incentive to transact directly.
• High per-transaction dollar values accelerate disintermediation — when the commission on a single order runs to thousands of dollars, suppliers will just pick up the phone.
• Faire works because the independent brand and independent retailer sides are both highly fragmented and the transaction costs of the old model (line sheets, PDFs, net-60 terms) were genuinely painful.