Contents
Every market has a graveyard of startups that died not because the idea was wrong, but because the world wasn't ready. Investigating the graveyard means systematically studying failed ventures to find compelling concepts that were defeated by timing, technology limitations, or execution failures — and relaunching them into conditions that have since changed in their favor.
Section 1
How It Works
The core insight is counterintuitive: failure is a signal, not a verdict. When a startup dies, most observers file the idea away as "didn't work" and move on. But the autopsy matters more than the obituary. A company can fail because the broadband wasn't fast enough, the payment rails didn't exist, the regulatory environment was hostile, the smartphones hadn't shipped, or the founders simply ran out of money before the market caught up. None of those causes invalidate the underlying demand.
The framework works by treating dead startups as free market research. Someone else spent millions of dollars and years of effort proving that a specific customer need exists — they just couldn't serve it profitably given the constraints of their era. Your job is to identify which constraints have since been removed. Webvan proved in 1999 that millions of consumers wanted groceries delivered to their door. The idea wasn't wrong. The infrastructure was: last-mile logistics were ruinously expensive, smartphone penetration was zero, and the gig economy didn't exist. Fourteen years later, Instacart launched into a world where all three constraints had evaporated.
The underlying principle is temporal arbitrage on validated demand. The graveyard contains ideas that were stress-tested against real consumers. The ones that attracted genuine user enthusiasm before dying of structural causes are the most valuable — they represent proven demand waiting for feasible supply. You're not guessing whether people want the product. You're betting that the world has changed enough to make delivery possible.
This is distinct from simply copying a failed company. The graveyard investigator doesn't replicate the dead startup's product — they extract the demand signal and rebuild from scratch using today's tools, infrastructure, and market conditions. The product may look entirely different. The insight is the same.
— Marc Andreessen, Andreessen Horowitz"There's an apparent paradox with failed startups: they often fail despite being right about the idea. The timing just wasn't right."
Section 2
When to Use This Framework
✓
Best Conditions for Investigating the Graveyard
| Dimension | Ideal conditions |
|---|---|
| Founder profile | Analytical operators who enjoy forensic research. You need the patience to study post-mortems, the pattern recognition to distinguish "bad idea" from "bad timing," and the technical fluency to assess whether today's infrastructure actually resolves the original constraint. Domain expertise in the failed company's sector is a significant advantage. |
| Stage | Ideation and pre-seed. The framework is most powerful when you're searching for what to build. It provides a structured alternative to brainstorming — you're mining history instead of imagining futures. Less useful once you already have a product in market. |
| Market conditions | Best deployed after a major technology shift (mobile, cloud, AI, blockchain), infrastructure buildout (payment rails, logistics networks, 5G), or regulatory change that removes a constraint that previously killed companies. The richer the shift, the deeper the graveyard you can mine. |
| Competitive environment | Ideal when incumbents have written off the category as "proven unworkable." The graveyard creates a psychological moat — investors and competitors who remember the original failure often dismiss the category entirely, giving you a window to build without competition. |
| Inputs needed | Startup post-mortems (CB Insights, Failory, TechCrunch archives), Wayback Machine captures of dead products, Crunchbase funding histories, founder interviews, patent filings from defunct companies, and a clear map of which technological or market constraints have changed since the original failure. |
The framework is unusually potent right now. The AI infrastructure wave is removing constraints that killed entire categories of startups in the 2010s — companies that needed human-level language processing, real-time image recognition, or personalized recommendation engines at scale. Simultaneously, the post-ZIRP correction has created a fresh graveyard of 2020–2022 startups that had genuine product-market fit but burned through capital at unsustainable rates. Many of those ideas are now available to be rebuilt with leaner cost structures and more disciplined unit economics.
Section 3
When It Misleads
⚠
Failure Modes & Blind Spots
| Blind spot | What goes wrong |
|---|---|
| Survivorship bias in reverse | You assume every dead startup had a good idea that was merely mistimed. In reality, many startups die because the idea was genuinely bad — the demand didn't exist, the problem wasn't painful enough, or the market was structurally too small. Not every grave contains treasure. |
| Constraint misdiagnosis | You identify the wrong reason the original failed. You think it was a technology problem, but it was actually a demand problem. You rebuild with better tech and discover the same indifference the original encountered. The autopsy must be precise. |
| The graveyard is already being excavated | You're not the only one reading post-mortems. If a failed concept is obviously viable now, multiple teams may be pursuing it simultaneously. The "psychological moat" of a dead category erodes quickly once one well-funded team enters. |
| Anchoring to the dead product | You study the failed company so closely that you unconsciously replicate their mistakes. You copy their UX, their pricing, their go-to-market — inheriting assumptions that were wrong even in the original era. The demand signal is valuable; the product decisions usually aren't. |
| Stigma contagion | Investors and partners who remember the original failure project that failure onto your venture. "Isn't this just Webvan again?" becomes a fundraising obstacle. You need a clear narrative for why this time is structurally different — not just optimistically different. |
| Incomplete constraint removal | The original failed because of five constraints. Three have been removed. You launch assuming the remaining two don't matter — but they do. Partial constraint removal can be worse than none, because it gives you false confidence. |
The single most common mistake is confusing "people signed up" with "people would pay." Many dead startups attracted impressive user numbers — Friendster had over 100 million registered users before its decline — but never found a sustainable business model. User enthusiasm is necessary but not sufficient. When you investigate the graveyard, you need to separate demand validation (people wanted this) from business model validation (people would pay enough to make this work). The former is common in graveyards. The latter is rare.
Section 4
Step-by-Step Process
Step 1 — ExcavateTools: CB Insights failure database, Failory, Crunchbase (filter by 'closed'), TechCrunch archives, Wayback Machine, AngelList historical data
Build a systematic inventory of dead startups in your domain
Don't rely on memory or anecdote. Build a structured database of failed companies in your area of interest. For each, capture: what they built, when they launched, how much they raised, how many users they reached, and — critically — why they died. Prioritize companies that raised significant capital (indicating investor-validated demand) but failed within 2–5 years (indicating execution or timing problems, not idea problems). Aim for 20–50 entries before moving to the next step.
Step 2 — AutopsyTools: Founder post-mortems, Glassdoor reviews, investor interviews, patent filings, SEC filings (for public companies), industry reports from the era
Diagnose the true cause of death for each candidate
For each promising dead startup, conduct a rigorous autopsy. Categorize the cause of death: Was it technology limitations? Market immaturity? Regulatory hostility? Capital inefficiency? Founder conflict? Business model failure? The most valuable candidates are those where the cause of death is external and time-dependent — meaning the constraint was outside the company's control and has since changed. Discard companies that died from internal dysfunction or fundamentally flawed value propositions.
Step 3 — Map the DeltaTools: Technology landscape analysis, infrastructure audits, regulatory tracking (GovTrack, regulatory agency websites), market sizing (Statista, IBISWorld), cost benchmarking
Identify what has changed since the original failure
For your top 3–5 candidates, build a detailed "constraint delta" — a side-by-side comparison of the conditions when the original failed versus conditions today. Be specific: "Smartphone penetration was 18% in 2007; it's 89% today." "Last-mile delivery cost was $12/order in 2001; it's $3.50 today via gig networks." "GPT-4 can now do what required a 50-person content team in 2015." The constraint delta is your investment thesis. If you can't quantify it, you don't have one.
Step 4 — Validate FreshTools: Landing pages (Carrd, Webflow), social media ads, Reddit/community surveys, concierge MVPs, pre-order campaigns
Test demand in today's market without referencing the dead company
Here's the critical discipline: do not assume the original's demand still exists. Markets shift. Consumer preferences evolve. Validate from scratch. Build a landing page describing the value proposition (without mentioning the dead company), run $500–$1,000 in targeted ads, and measure conversion intent. Conduct 30+ user interviews. The graveyard gives you a hypothesis; validation gives you confidence.
Step 5 — Rebuild from ZeroDeliverable: Product spec that explicitly documents every divergence from the original, with rationale for each decision
Design the new product using today's stack and economics
Resist the temptation to clone the dead product with better technology. Instead, start from the validated demand signal and design the product as if the original never existed. Use modern infrastructure (cloud, AI, APIs, gig labor) to achieve fundamentally different unit economics. Document every design decision and explicitly note where and why you're diverging from the original's approach. The output should be a product that serves the same need but looks nothing like its predecessor.
Section 5
Questions to Ask Yourself
Discovery
What startups in my domain raised $10M+ but shut down within five years — and what did their users love before the company died?
Which technology limitations, infrastructure gaps, or regulatory barriers existed when the original launched that no longer exist today?
Are there entire product categories that investors currently consider "dead" because of a high-profile failure — and is that reputation deserved?
What did early users of the dead product say in reviews, forums, or social media? Was there genuine enthusiasm or just curiosity?
Diagnosis
Can I clearly articulate the cause of death in one sentence — and is that cause external (timing, infrastructure, regulation) rather than internal (bad product, no demand)?
Has the specific constraint that killed the original been fully removed, or only partially? What percentage of the original's cost structure would be eliminated by today's tools?
Did the original fail because of one fatal constraint, or a combination of five? How many of those have actually been resolved?
Am I romanticizing the dead company's potential, or do I have hard evidence that the demand was real?
Validation
If I describe this product to 30 potential customers without mentioning the predecessor, do they express genuine willingness to pay — or just polite interest?
Can I achieve unit economics that are at least 3x better than the original, given today's infrastructure and tools?
Is anyone else currently excavating this same graveyard? How many teams are pursuing this resurrected concept right now?
[Narrative](/mental-models/narrative)
Can I explain to a skeptical investor — in 60 seconds — why this time is structurally different, with specific data points?
Do I have a compelling answer to "Isn't this just [dead company] again?" that goes beyond "we'll execute better"?
What's my plan for the moment when a journalist inevitably compares me to the failed predecessor?
Section 6
Company Examples
Chewy
Resurrected the Pets.com model with better logistics and customer obsession
Pets.com became the poster child of the dot-com bust, burning through $300 million and collapsing in 2000 just 268 days after its IPO. The idea — selling pet supplies online — was sound. The execution was catastrophic: the company spent lavishly on Super Bowl ads, shipped heavy bags of dog food at a loss, and operated in an era when e-commerce logistics were primitive and expensive. Ryan Cohen launched Chewy in 2011 into a fundamentally different world: FedEx and UPS had built out residential delivery networks, Amazon had normalized online purchasing of bulky goods, and the pet industry had grown to over $60 billion annually. Chewy's autoship subscription model — which Pets.com never developed — created predictable recurring revenue. PetSmart acquired Chewy for $3.35 billion in 2017, and the company went public in 2019, reaching a market cap above $40 billion at its peak. Same demand signal. Entirely different infrastructure and business model.
Airbnb
Succeeded where short-term rental predecessors failed by riding the smartphone and trust infrastructure wave
Before Airbnb, there were multiple attempts at peer-to-peer accommodation — VRBO launched in 1995, Couchsurfing in 2004, and various classified-ad-based rental services existed throughout the 2000s. None achieved mainstream scale because three constraints held them back: there was no ubiquitous mobile platform for real-time booking, no standardized digital payment system for peer-to-peer transactions, and no social identity layer (Facebook Connect launched in 2008) to create trust between strangers. Airbnb launched in 2008 and scaled from 2010 onward precisely as all three constraints dissolved simultaneously. The founders studied what had come before — Brian Chesky has spoken about analyzing why earlier home-sharing attempts stalled — and built specifically for the new infrastructure. Airbnb reached a $100 billion market cap at IPO in December 2020.
I
Instacart
Perfected the Webvan online grocery model 14 years later
Webvan raised $375 million, built massive automated warehouses, and attempted to create an end-to-end grocery delivery infrastructure from scratch. It filed for bankruptcy in 2001, becoming one of the most expensive failures of the dot-com era. The demand was real — Webvan had loyal customers who loved the service — but the cost structure was impossible: building warehouses cost hundreds of millions, last-mile delivery was ruinously expensive with full-time drivers, and online ordering was friction-heavy in a pre-smartphone world. Instacart, founded in 2012, resurrected the identical demand signal but inverted the cost structure. Instead of building warehouses, it used existing grocery stores as inventory. Instead of hiring drivers, it used gig-economy shoppers. Instead of desktop ordering, it built for smartphones. Instacart reached a reported $39 billion valuation in 2021 before going public in 2023. The constraint delta was enormous: gig labor, smartphones, and the "asset-light" platform model turned a $375 million failure into a multi-billion-dollar business.
S
Spotify
Resurrected the legal music streaming concept after Napster, Rhapsody, and others failed
Napster proved in 1999 that tens of millions of people wanted to stream music on demand — then was shut down by litigation in 2001. Rhapsody launched a legal subscription service in 2001 but never achieved mainstream adoption because broadband penetration was low, the music catalog was limited (labels were hostile), and the $9.99/month price point felt steep when piracy was free. Daniel Ek launched Spotify in 2008 in Sweden, entering a world where broadband was ubiquitous, labels had been sufficiently terrified by a decade of piracy to negotiate licensing deals, and the freemium model (which Rhapsody never tried) could convert free users into paying subscribers over time. Spotify reached 615 million monthly active users by late 2023. The graveyard of music streaming predecessors — Napster, Rhapsody, Rdio, MOG — provided both the demand validation and the cautionary lessons that shaped Spotify's approach.
Coupang
Resurrected the rapid e-commerce delivery model that failed in the West, adapted for South Korea's density
Multiple Western startups attempted same-day and next-day delivery in the 2000s and early 2010s — Kozmo.com burned through $280 million delivering convenience items in under an hour before collapsing in 2001. The model failed because U.S. suburban geography made last-mile delivery prohibitively expensive. Coupang, founded in 2010, recognized that South Korea's extreme urban density (Seoul has 16,000 people per square kilometer) and world-leading internet infrastructure fundamentally changed the economics. Coupang built its own logistics network — over 100 fulfillment centers across a country smaller than Pennsylvania — and achieved "Rocket Delivery" (next-day or same-day) for 99% of orders. The company went public on the NYSE in 2021 at a $60 billion valuation. The graveyard insight: the delivery model wasn't wrong, it was deployed in the wrong geography.
Section 7
Adjacent Frameworks
The graveyard doesn't exist in isolation. These frameworks sharpen the excavation or extend the insight:
Pairs well with
Industry timing arbitrage
The natural complement. Timing arbitrage provides the analytical lens for determining when a dead idea becomes viable again. Combine the graveyard's demand signals with timing arbitrage's infrastructure analysis for a complete investment thesis.
Pairs well with
The Chris Dixon [Idea Maze](/mental-models/idea-maze)
Dixon's Idea Maze framework maps all the paths a startup could take through a problem space — including the dead ends others have already explored. The graveyard provides empirical data for the maze: every failed startup is a mapped dead end that narrows your search toward viable paths.
In tension with
Category creation
Category creation asks you to build something the world has never seen. The graveyard asks you to rebuild something the world has already rejected. The tension is productive: the best graveyard resurrections often create a new category by combining a dead idea with a new delivery mechanism.
In tension with
Invent a new sport
Inventing a new sport is about radical originality — creating demand that doesn't yet exist. The graveyard is about recycling proven demand. These are opposing instincts, and founders should be honest about which one they're actually pursuing.
Apply next
Build a Copycat
Once you've identified a resurrectable idea, the Copycat framework provides the execution playbook for studying the dead company's product decisions, extracting what worked, and adapting for current conditions.
Apply next
Niche down
The original may have failed by going too broad too fast. After excavating the demand signal, consider launching in a single niche where the constraint delta is most favorable — then expand from a position of strength.
Section 8
Analyst's Take
Faster Than Normal — Editorial View
My honest read: this is one of the highest-signal, lowest-noise frameworks available to founders — and almost nobody uses it systematically.
The reason is psychological. Founders are drawn to novelty. The startup ecosystem celebrates invention, not resurrection. Telling an investor "I'm rebuilding a company that failed" requires a level of narrative confidence that most first-time founders don't have. It feels less glamorous than "I'm creating something the world has never seen." But the data is overwhelmingly in favor of the graveyard excavator. Chewy, Instacart, Airbnb, Spotify, DoorDash — some of the most valuable companies of the last decade are direct descendants of dead predecessors. They didn't invent new demand. They served existing demand that was waiting for the world to catch up.
The framework is especially powerful right now for two reasons. First, the 2020–2022 ZIRP era created a massive fresh graveyard. Hundreds of well-funded startups with genuine product-market fit — in areas like instant delivery, creator economy tools, remote work infrastructure, and climate tech — burned through capital and shut down not because the ideas were bad, but because the unit economics didn't work at 2021 burn rates. Many of those ideas are now viable at 2024 cost structures, especially with AI reducing headcount requirements by 30–60% for certain functions. Second, generative AI is the single largest constraint-removal event since the smartphone. Every startup that died because it needed human-level content generation, customer support, data analysis, or personalization is now a candidate for resurrection.
The founders I see making the most of this framework share three traits: they're obsessive researchers who enjoy reading post-mortems the way other people read novels; they have the intellectual honesty to distinguish "this failed because the world wasn't ready" from "this failed because nobody wanted it"; and they have the narrative skill to pitch a resurrected idea without triggering investor PTSD from the original failure.
The single biggest mistake is skipping the fresh validation step. The graveyard gives you a hypothesis, not a conclusion. Markets evolve in unpredictable ways. The demand that existed in 2001 may have been absorbed by a different product category entirely. Webvan's customers didn't sit around for 12 years waiting for Instacart — many of them adapted to Costco runs and meal kits. Instacart succeeded because it validated fresh demand in 2012, not because it assumed 2001 demand was still there. Treat the graveyard as a starting point for research, never as a substitute for it.
Section 9
Opportunity Checklist
Use this scorecard to evaluate whether a specific dead startup's concept is worth resurrecting. Score each item as yes (1 point) or no (0 points).
Graveyard Resurrection Scorecard
The dead startup attracted genuine user enthusiasm (not just funding) before it failed — evidenced by user growth, retention data, or public testimonials.
The primary cause of death was external and time-dependent (technology, infrastructure, regulation, cost structure) — not internal (bad product, no demand, founder conflict).
I can identify at least one specific, quantifiable constraint that has been removed since the original failure (e.g., "delivery cost dropped from $12 to $3.50 per order").
The constraint removal is structural and permanent — not cyclical or dependent on a temporary market condition.
No well-funded team is currently pursuing this same resurrected concept (or if they are, I have a differentiated approach).
Fresh validation (landing page tests, user interviews, or concierge MVP) confirms that the demand still exists in today's market.
I can achieve unit economics at least 3x better than the original, using today's infrastructure and tools.
I have a clear, data-backed narrative for why "this time is different" that would satisfy a skeptical investor in 60 seconds.
The product I would build looks meaningfully different from the original — I'm serving the same need, not cloning the same product.
The market size has grown since the original failure (more potential users, higher willingness to pay, or adjacent markets that didn't exist before).
I have domain expertise or local knowledge that gives me an advantage in understanding why the original failed and what needs to change.
Section 10
Top Resources
01
Book
The foundational text on why good companies fail — and by extension, why their failures often contain the seeds of the next generation's success. Christensen's framework for understanding when incumbents are structurally unable to pursue an opportunity is essential for diagnosing whether a dead startup failed because of timing or because the market genuinely couldn't support it.
02
Book
The practical sequel to The Innovator's Dilemma, focused on predicting which disruptions will succeed and which will fail. The "signals of change" framework is directly applicable to graveyard investigation — it gives you a structured method for assessing whether the conditions that killed a previous attempt have actually shifted.
03
Essay
Dixon's essay argues that the best founders have a deep understanding of all the paths through a problem space — including the dead ends. The graveyard is the empirical record of those dead ends. This short piece reframes failed startups as navigational data rather than cautionary tales, and it's the closest thing to a philosophical manifesto for the graveyard investigation framework.
04
Book
Horowitz ran Loudcloud/Opsware through the dot-com crash — a company that nearly died and was resurrected by pivoting into a different market with the same core technology. The book is a masterclass in understanding the difference between "this company is dying because the idea is bad" and "this company is dying because the market isn't ready." Essential reading for anyone trying to diagnose cause of death.
05
Podcast
The best single source for detailed case studies of companies that succeeded where predecessors failed. Episodes on Instacart, Spotify, Airbnb, and DoorDash all explicitly trace the lineage from dead predecessors to billion-dollar successors. The hosts' forensic approach to company history models exactly the kind of analysis the graveyard framework demands.
Why this matters next
mental modelsSurvivorship Bias
Chewy applied the Survivorship Bias mental model
mental modelsIdea Maze
Chewy applied the Idea Maze mental model
mental modelsNarrative
Chewy applied the Narrative mental model
mental modelsScale
Chewy applied the Scale mental model
mental modelsEnvironment
Chewy applied the Environment mental model
mental modelsCost
Chewy applied the Cost mental model
Continue exploring
RRUNIAACTRND
Framework
Recent funding rounds
Analyze companies that have recently secured significant investment, identifying
Framework
Unbundling
Breaking down a bundled product or service into separate, standalone offerings,
Framework
Industry timing arbitrage
Apply newly developed technology from one industry to another that hasn't yet ad
Framework
Acqui-Deaths
Identify opportunities created when large companies acquire startups, potentiall
Framework
Three-Star reviews
Find business opportunities by analyzing moderately satisfied customers' feedbac
Framework
Niche down
Focus on a highly specific market segment or customer base, becoming a specialis
More like this, in your inbox
I send a newsletter every week — free, no spam, unsubscribe anytime.