86:08
product strategy
Casey Winters argues that almost no company today can reach IPO or sustain long-term growth on a single product, so leaders must forecast the S-curve of their core product's growth and deliberately time, staff, and evaluate a second product well before growth stalls — and that a second product only needs to strengthen ONE of acquisition, retention, or monetization (not all three, unlike a first product's PMF), but must do so at a scale big enough to move the whole company.
PMF (Casey's definition): customers are satisfied enough to stop leaving and start complaining about what else they want, rather than churning; this creates a positive growth slope you can scalably acquire against.
Four types of post-PMF product work (Reforge framework): Features, Growth, Scaling, Expansion — expansion is the talk's focus.
Five types of expansion ordered easiest to hardest: Geographic/Category, Format, Product Value, Platform, Strategic Diversification — classified by whether product, market, and core competency change.
Historically VCs sought single-product IPO winners (Google, Zoom, Duolingo) enabled by huge markets, weak competition, scarce talent, and strong economies of scale; those conditions have eroded (faster fundraising competition, tech-native incumbents, wider talent pools, breakable network effects), so multi-product strategy is now usually required.
You must model the S-curve of your core product's growth to predict when it will asymptote, and start the second product well ahead of that point, since new products take 1-3 years to find PMF.
At Eventbrite, before building new products, Casey first tried to make the existing creator→marketing→ticket-sales→new-creator growth loop turn faster by grafting on paid acquisition, SEO, and partner distribution — but the core loop only monetized ~10%, forcing a real second-product need.
Whether/when you need a second product depends on business-model-specific factors, not the business model category itself: competitive intensity, acquisition channel strength (e.g., network effects vs SEO vs virality), retention mechanics, monetization potential, market size/growth, and how natural the product adjacencies are.
Business model type is 'a total red herring' for predicting when a second product is needed — the individual business and market components matter, not the category (consumer, B2B, marketplace, subscription).
Paired case studies show identical-looking business models diverge sharply underneath: Pinterest vs Snapchat, Figma vs Canva, GrubHub vs Instacart, Duolingo vs Calm — in each pair the company that looked like it needed a second product less ended up needing one more, or vice versa, and valuation didn't track who 'won' at new-product development.
A second product doesn't need full PMF the way a first product does — it only needs to fix ONE weak link (acquisition, retention, or monetization) in the existing business, but must do so at a scale large enough to matter company-wide.
New products succeed most often when they're the only path left for the company to keep growing and the CEO is genuinely invested — not as diversified 'portfolio' bets; Casey rejects fixed-percentage portfolio allocation in favor of one focused bet on the core product's identified weakness.
Founders should not outsource new-product development entirely to PMs hired post-PMF, since those employees typically lack zero-to-one intuition; a founder needs to stay personally involved.
New-product teams should replace OKR-style success metrics with a 'learning feedback loop': a ranked list of assumptions to test, an expected path, and what's being learned, fed back to leadership quickly — since real progress may take a year or more before 'winning' is even possible.
Three failure modes for new products even with good process: (1) PMF exists but requires far more work than expected (Pinterest Q&A), (2) the destination is illusory and no real demand exists (Tinder Social), (3) the team drifts from the originally targeted business problem even while finding some traction (unnamed travel company).
Marketplaces must solve product-market fit ('liquidity') on both supply and demand sides simultaneously; specific custom thresholds (e.g., GrubHub's 'two orders a day' within 4 months) determine whether supply will stick around, and roughly 40% conversion/retention on both sides tends to let network effects take over.
Marketplace PMF takes roughly double the time of B2B SaaS PMF (Thumbtack: 4 years; Faire: ~6 months, among Casey's clients).
Whether to launch demand-first or supply-first when entering a new market is situational; today ~9 of 10 marketplaces launch supply-first, though GrubHub itself started demand-first before switching.
Blitzscaling-style capital-fueled subsidy strategies (à la Uber/Lyft's flat driver pay, or aggressive AdWords overspend) were enabled by the zero-interest-rate environment and have since fallen away; GrubHub instead grew profitably on tight payback-period discipline because it was capital-constrained.
For handling problems in the offline part of a marketplace's service (late delivery, bad food), the dominant pre-IPO strategy is to assume responsibility regardless of fault — GrubHub found customers who had a negative experience that was made right had HIGHER lifetime value than those who never had a negative experience.
SEO as a growth channel has 'yo-yoed' in effectiveness over time (worked for GrubHub in 2008, Apartments.com ~2005, Pinterest in 2014; became harder for startups mid-2010s due to Google's authority bias toward incumbents; briefly opened up via topical/niche authority; now clouded by AI content flooding search) — Casey's current favorite growth lever is distributing supply/user content into large, not-yet-optimized external networks (historically Google, now TikTok/Instagram Search).
Org design for incubating a second product: keep it organizationally separate (its own GM, engineering, product) until it either proves itself and gets folded back into the core, or stays independent long-term if its growth model is genuinely decoupled from the core business; incentive structures range from tying comp to the new unit's standalone performance, independent/phantom stock revalued periodically (Tinder inside Match Group — which led to a lawsuit over valuation), spinning the unit out entirely (TripAdvisor from Expedia, Qualtrics multiple times), or an LTIP tied to growth/profitability milestones (used at Apartments.com).
Four types of product work (Features, Growth, Scaling, Expansion) — Reforge's Product Strategy taxonomy of where post-PMF product effort goes: improving core value (Features), reducing friction to existing value (Growth), supporting more users/internal staff without breaking (Scaling), and moving into new segments/value props (Expansion). Apply: Use it to categorize where your team's current roadmap effort is actually going and to notice when expansion work is chronically zero even as core-product growth headroom shrinks.
Expansion taxonomy (5 types, easiest to hardest) — A classification of expansion moves — Geographic/Category, Format, Product Value, Platform, Strategic Diversification — ranked by how much the product, market, and core competency each change. Apply: Before choosing an expansion move, locate it on this ladder to gauge how much of your product/market/skills you'll need to rebuild versus reuse, and to estimate relative odds of finding PMF.
Casey's PMF definition — Product-market fit is when customers are satisfied enough to stop leaving and begin complaining about additional wants instead of churning, creating a positive growth slope. Apply: Use this instead of 'exceeding expectations' as your PMF bar, since expectations rise continuously and can't be permanently exceeded; watch for the shift from churn to feature complaints as your PMF signal.
S-curve growth forecasting — Modeling the trajectory of your core product's growth to predict when it will asymptote regardless of further optimization. Apply: Build the forecast early and start second-product investment 1-3 years ahead of the predicted asymptote, since waiting until growth actually stalls guarantees a growth gap.
Core growth loop mapping — Diagramming your product's existing acquisition→value→re-acquisition cycle (e.g., Eventbrite: creator lists event → markets it → sells tickets → attendees become creators) as a diagnostic tool. Apply: Map your own loop before building new products, to see whether the loop can simply be accelerated (via grafted acquisition channels) before committing to a whole new product.
Grafting acquisition loops onto a core loop — Adding new acquisition-focused mechanisms (paid acquisition funded by revenue, SEO, partner integrations, lifecycle emails/push) onto an existing core growth loop rather than building a new product. Apply: Try this first when a core loop still has room to turn faster, reserving a full second product for when this grafting hits its own ceiling (as Eventbrite's ~10% monetization did).
Data Network Effect — A network effect where user engagement generates data that improves recommendations/value for that user and similar users (Pinterest's model, contrasted with a friend/social graph). Apply: Recognize this as an alternative retention engine to a social graph when your product's value comes from interest/content matching rather than direct social connections.
Friend graph → topic/interest graph shift — Pinterest's pivot from friend-based recommendations (which worked for homogeneous early users) to an interest-based graph as its audience diversified. Apply: Consider shifting from social-graph to interest-graph personalization when expansion into new geographies/demographics increases interest heterogeneity among connected users.
Content loop (PLG acquisition mechanism) — A pattern where users create a piece of content and share it within their organization, driving further adoption (used by Figma and Canva). Apply: Design collaborative artifacts (files, designs, docs) to be naturally shareable inside a company as a bottom-up acquisition channel.
PLG → community adoption → top-down sales motion — A B2B growth sequence where individual users adopt a product bottom-up, it spreads through a professional community, and once enough users exist inside an org, sales sells top-down to the whole company. Apply: Expect (and design your sales process around) top-down deals only becoming viable after sufficient bottom-up penetration within target accounts.
Marketplace acquisition loop accumulation — The pattern where marketplaces start with SEO (demand side) and sales (supply side) and progressively add paid acquisition, TV, and incentivized referrals as accumulated LTV justifies it. Apply: Add acquisition channels incrementally as your marketplace's proven LTV grows, rather than trying all channels simultaneously from day one.
Cross-side network effects — A network effect where more supply attracts more demand and vice versa, characteristic of marketplaces. Apply: Use as the retention/defensibility engine to reason about in marketplace strategy, and recognize it can still be broken by a well-funded competitor (e.g., DoorDash vs. GrubHub).
Four-sided marketplace — Adding a fourth transacting party (Instacart added CPG advertisers to consumers, shoppers, and retailers) to fix unit economics that a simpler marketplace structure can't solve. Apply: Consider adding an additional market side only to solve a specific, otherwise-unsolvable monetization problem — Casey frames it as 'hard mode,' not a default move.
Slotting fees / digital slotting — The online analogue (Instacart Ads) of the real-world grocery practice where CPG brands pay for premium shelf placement, applied to paid placement in search results. Apply: Look for real-world monetization mechanisms in your industry (like physical shelf fees) that can be digitized as an ads product when your core marketplace transaction fee is capped.
One-big-bet model vs. portfolio theory — Casey's argument against allocating a fixed percentage of resources to 'innovation' vs. core work, in favor of forecasting the core product's growth bottleneck and making one large, targeted bet at that specific weakness. Apply: Instead of running several parallel low-conviction innovation bets, identify your single biggest projected growth constraint and commit concentrated resources to solving exactly that.
Retention cohort curve as PMF signal — A cohort retention curve that flattens over time, indicating a stable user base whose value can be reinvested into acquisition (via revenue) or distributed virally (via content). Apply: Plot retention by cohort and look for flattening as the signal that a user base is durable enough to fund scalable growth.
Four scalable growth channels — Paid acquisition, virality/referral, content/SEO sharing, and sales, as the four ways a retained user base converts into scalable growth. Apply: Check which of these four channels your retained/monetizing user base can actually fund or fuel before assuming you have a scalable growth model.
First-product PMF vs second-product 'single-lever' bar — A distinction that a first product needs full PMF (retention AND monetization enabling acquisition), while a second product only needs to strengthen ONE of acquisition, retention, or monetization — but at company-moving scale. Apply: When evaluating a second product's success, don't require it to independently satisfy full PMF; instead check whether it meaningfully improves the one specific lever it was built for, at sufficient scale.
Learning feedback loop process (vs. OKRs) — A new-product-team process model that replaces conventional OKR metrics (MAUs, revenue) with a ranked list of assumptions to test, an expected path, and explicit statements of what's being learned, reported back to leadership quickly. Apply: Run early-stage new-product teams on assumption-testing roadmaps rather than quarterly OKRs, since real learning/progress may precede any measurable 'win' by a year or more.
Three new-product failure modes — (1) PMF exists but requires far more work than expected; (2) the target destination is illusory and no real demand exists; (3) the team drifts from the originally agreed business problem even while achieving some success. Apply: Use these three patterns as a checklist when deciding whether to keep funding or cut a struggling new-product initiative.
Hamilton Helmer's Seven Powers — A competitive-moat framework (referenced by an audience question) covering multiple sources of defensibility such as brand, economies of scale, and network effects. Apply: Use it as a broader checklist of moat types beyond network effects, which is Casey's personal focus and not necessarily the right moat for every business.
Network effect taxonomy (Direct, Cross-side, Data) — Direct network effect (every added user increases value for existing users, e.g. WhatsApp), Cross-side network effect (two sides increase each other's value, e.g. marketplaces), and Data network effect (engagement improves recommendations for similar users, e.g. Pinterest). Apply: Identify which type of network effect your product relies on to reason correctly about how it strengthens or weakens as you scale or expand.
Assume responsibility for offline actions — A dominant pre-IPO marketplace strategy of taking responsibility for problems in the offline part of a transaction (late delivery, bad food) regardless of whose fault it was. Apply: Proactively compensate/fix offline failures attributable to your platform even when a third party (restaurant, driver) is at fault, since Casey found this raises rather than lowers customer LTV.
Cohort analysis framework (value-action x time) — A retention/analysis structure with a value-received action on the y-axis and a natural time cadence (daily/weekly/monthly) on the x-axis. Apply: Choose the time cadence to match your product's natural usage frequency, and choose a value-action metric that truly reflects value delivered, inventing a way to measure it (e.g. via survey) if it's not natively tracked.
Apartment.com vs Rent.com conversion-measurement strategies — Two contrasting approaches to an unmeasurable offline conversion (a signed lease): Apartment.com's trust-based subscription model without conversion proof, versus Rent.com's cash-incentivized self-reporting of confirmed leases. Apply: When you lack visibility into whether your product drove a real-world outcome, choose between trust-based pricing or paying users/customers to self-report that outcome, weighing each against your specific market.
Platform vs. Marketplace framework — Both are cross-side network effect businesses, but a platform must first build its own 'killer app' (like B2B SaaS) before it has enough scale to attract third-party developers to build on top. Apply: Don't attempt a platform/developer ecosystem strategy before your own core product is mature and well-served internally — Casey shut down a premature Eventbrite platform push for this reason.
Integration-first platform rollout — Launching integrations with popular tools your customers already use (e.g., Eventbrite + MailChimp) before enabling entirely new third-party businesses to be built on your platform. Apply: Prioritize proving PMF on integrations with existing popular tools before investing in a full third-party developer platform.
Marketplace liquidity framework — The marketplace-specific version of PMF requiring the product to work for both supply and demand sides simultaneously; demand-side value is largely driven by selection/conversion, supply-side by whether volume justifies staying. Apply: Evaluate a marketplace's readiness by checking both sides' thresholds separately rather than a single unified PMF metric.
Marketplace scaling metric framework — Identifying a custom supply-side retention threshold (e.g., GrubHub's 'two orders a day' within 4 months) and a supply-density threshold enabling scalable demand acquisition/retention (~40% on both sides as a rough scale-trigger). Apply: Define your own marketplace's specific supply and demand thresholds from data, then track progress toward roughly 40% conversion/retention on both sides as the signal that network effects will take over.
Blitzscaling — Reid Hoffman's strategy of aggressive capital-fueled growth and subsidy to grab a market fast, which Casey says was viable mainly in the zero-interest-rate era. Apply: Recognize this approach requires large capital reserves and favorable financing conditions; GrubHub's profitable, thoughtful growth is offered as the contrasting capital-constrained alternative.
Subsidizing one marketplace side to buy time — Paying or otherwise propping up one side of a two-sided marketplace (e.g., Uber/Lyft paying drivers flat hourly rates) to give the other side time to develop. Apply: Consider direct subsidy of the harder-to-build side only when you have the capital reserves to sustain it until natural liquidity develops.
Demand-first vs supply-first market launch — Two strategies for entering a new geographic market: building demand before signing supply, or signing supply before building demand; today ~9 of 10 marketplaces launch supply-first. Apply: Choose based on situational factors (as GrubHub did, switching from demand-first to supply-first) rather than assuming one approach is universally correct.
Payback-period discipline — Setting an explicit target payback period (e.g., GrubHub's 6-month AdWords payback) as a guardrail on acquisition spend. Apply: Use a payback-period target to decide how much you can spend per acquired customer/supplier, and treat any need to stretch that target as a deliberate, risk-aware decision rather than drift.
Liquidity game vs. scalable efficient growth game — A framing contrasting well-funded competitors who subsidize both marketplace sides to force PMF/liquidity versus capital-constrained companies pursuing efficient, LTV-disciplined incremental growth. Apply: Use this framing to understand why a competitor's spending looks 'irrational' to you — they may be racing to unlock liquidity in a market you've already unlocked.
Google 'Authority' (link-based) — Google's historical link-count-based trust signal that favors incumbent sites with more backlinks. Apply: Recognize this as a structural disadvantage for new SEO-dependent startups competing against established sites.
Topical/micro-topic authority SEO strategy — An SEO approach of owning an entire niche question-graph deeply rather than competing on broad site-wide authority. Apply: Build comprehensive content coverage of a narrow topic to outrank broadly-authoritative incumbents on that specific topic.
Search volume vs. competition 2x2 — A framework plotting a keyword/topic's search volume against its competition level to evaluate SEO opportunity, most winnable in high-volume/low-competition (typically fast-trending) quadrants. Apply: Look for rapidly trending topics with rising search volume but not-yet-established competition (Casey's pickleball example) as the sweet spot for new-entrant SEO.
Distributing content into large, under-optimized networks — A growth lever of pushing supply- or user-generated content into external networks larger than your own that haven't yet been fully gamed for search/discovery — historically Google, now TikTok/Instagram Search. Apply: Look for platforms with significant native search behavior in your category that competitors haven't yet 'SEO-ified,' and apply SEO-like content strategy there.
GM structure for incubating new products — Organizing a new or acquired product line with its own general manager, engineers, and product team, run independently until mature enough to fold back into the core. Apply: Use a separate GM structure when incubating something fragile so it isn't smothered by core-business priorities; fold it back in once it no longer needs independence, or keep it separate long-term if its growth model stays decoupled from the core.
Independent/phantom stock for internal incubation — A mechanism (used for Tinder inside Match Group) where an incubated product's equity is tracked separately and revalued periodically by independent auditors rather than paid in ordinary parent-company stock. Apply: Consider this when an incubated unit is expected to grow much faster than the parent, to properly incentivize its team — while being aware it can create valuation disputes (as happened with Tinder).
Spin-out for external valuation — Separating a business unit into its own company (e.g., TripAdvisor from Expedia, Qualtrics multiple times) so it can get a market-set valuation and issue independent stock. Apply: Use a spin-out when you need genuine external market validation of a unit's value that an internally-set valuation can't credibly provide.
LTIP (long-term incentive plan) — A compensation mechanism (used at Apartments.com) that pays bonus parent-company stock or cash upon hitting growth/profitability milestones for a new initiative. Apply: Use an LTIP tied to specific growth/profitability milestones when a new product's team should be incentivized without creating a separate stock class.
Despite Snapchat 'nailing' rapid second/third-product development (Stories, a hit) and Pinterest visibly failing at its own second/third products (Place Pins, Q&A), Pinterest ended up valued more — because its single core product could scale to a billion users end-to-end on acquisition, retention, and monetization, while Snapchat's core product (disappearing photos) was fundamentally hard to monetize with ads, so it needed new products just to have a shot at building a viable ad business at all.
Figma, despite launching into an apparently 'competitive' market against Sketch and Adobe, ended up needing almost no second product to keep growing, because the PMF gap it opened over incumbents was so large it never closed — while Canva, despite launching into an apparently 'non-competitive' market nobody else wanted, grew so fast it now looks like a genuine threat to Adobe and Microsoft, who could clone 70% of its value prop and intercept new customers — forcing Canva into aggressive suite-building that Figma never needed.
Instacart had to build a full four-sided marketplace plus an ad exchange (Instacart Ads, now the majority of its revenue and essentially all its profit) just to reach roughly the same valuation as GrubHub, which needed only a simple two-sided delivery marketplace — illustrating that 'marketplace' as a category label hides wildly different amounts of required work.
Duolingo and Calm both launched successful second products (Sleep Stories for Calm; though Duolingo's own second product, Tiny Cards, failed) and Calm even added a third, Calm for Business — yet Calm's valuation was still only about a quarter of Duolingo's, because the deciding factor in consumer subscription isn't product breadth but market size and acquisition cost, and Calm 'did nothing wrong' but never found as cheap a growth channel as Duolingo's virality.
GrubHub's 'pickup' second product actively cannibalized its own core delivery product (confusing customers, muddying the demand-side network effect) and the resulting growth slowdown is cited as a partial explanation for how Postmates gained a foothold in LA before GrubHub diagnosed the problem.
Casey frames the standard 'innovation portfolio' resource-allocation model (fixed % of resources to new bets vs core work) as something the best next-generation companies are moving away from, in favor of forecasting exactly where the core product's growth model will break and firing one well-resourced, CEO-backed bet at that specific weakness.
A second product is, in one sense, structurally easier to build than a first: it only has to win on one of acquisition/retention/monetization instead of all three needed for standalone PMF — but this cuts both ways, since even a successful, PMF-having second product can fail to 'matter' at company scale if it lands in a small-TAM niche with weak monetization (a failure mode Casey says is common at large companies like Meta and Google).
The comparison of Apartment.com vs Rent.com shows two opposite strategies for coping with unmeasurable offline conversion data (leases signed after a lead): Apartment.com bet on trust and a subscription model without proof of conversion, while Rent.com paid users $250 gift cards to self-report confirmed leases to get real data — and the company that skipped direct measurement (Apartment.com) ended up winning the market.
Match Group's internal incubation of Tinder used a separate, periodically-revalued 'Tinder stock' set by independent auditors instead of ordinary parent stock — and this very mechanism later became the subject of a lawsuit by Tinder's founders/early employees alleging undervaluation, illustrating the 'who watches the watchman' problem of internally-set valuations for incubated products.
SEO's viability as a scalable acquisition channel has repeatedly flipped between favoring incumbents (broad link 'Authority') and briefly favoring insurgents (topical/niche authority letting small sites out-rank major outlets like NYT) — and Casey believes it has now flipped again against everyone, since AI-generated content flooding search results is pushing Google to lean on Reddit as a trust signal, and Reddit itself is now getting spammed in turn.
«I don't have a book to sell you today uh I just have a bunch of random thoughts uh from a lot of pain uh trying to te startups growing for the long term which a lot of time means you have to figure out how to build a second product that will help you grow»
— 00:51
«when they feel satisfied enough to stop leaving and they start complaining about all the other things uh that you should have instead of just stopping to use the product»
— 02:42
«The best time to get started is, you know, today; the second best time was three years ago.»
— 16:47
«business model is a total red herring — it does not explain at all when you should think about your second product; you need to be analyzing more of these components of your individual business and your individual market.»
— 19:43
«new products work when they have to work when it's the only way you can continue to grow when the CEO is like really invested»
— 40:54
«I recommend Founders do not Outsource this one of the founders has to be involved»
— 47:30
«there's no way you would have a successful grocery Marketplace IPO if they didn't do this»
— 36:26
«if something goes wrong in the offline world, it'll be associated to your software even though it's not your fault.»
— 61:20
«we valued the feedback... users that had a negative experience with us when we made up for it had a higher lifetime value than people [who] had a negative experience... [or] never had a negative experience.»
— 62:04
«this line of thinking was possible in the zero interest rate environment and has sort of fallen away»
— 70:28
«they're playing a liquidity game... and you're playing a scalable efficient growth game»
— 75:09
«it feels like you could do everything right and still lose»
— 78:41
«who watches the Watchman on how valuable the company is»
— 83:30
«you're probably going to have something like an L tip a long-term incentive plan»
— 85:26
Reception
No comments are available to gauge audience reception.
The talk is a practitioner's synthesis rather than an academic study: its force comes from paired real-company comparisons (Pinterest/Snapchat, Figma/Canva, GrubHub/Instacart, Duolingo/Calm) that Casey uses to argue business-model category is a poor predictor of when a second product is needed, and from his own operating history (GrubHub, Eventbrite, Pinterest) as evidence for his specific frameworks around growth-loop modeling and second-product timing.
86:08