The Golden Triangle of Pre-Seed / Seed Investing
How to Identify the 25% of Startups That Will Survive!
75% of startups die before their third anniversary — a figure that rises to 82% for tech startups in Europe, according to the latest CB Insights data.
As investors, our mission is twofold: dramatically minimize this percentage in our portfolio while maximizing our exposure to calculated risk to optimize our DPI.
How can we achieve this? For this purpose, I use the "Golden Triangle" of pre-seed evaluation — a structured framework to filter the excellent from the ordinary.
I. The Team: The Essential X-Factor
The brutal reality, confirmed by FirstRound Capital's post-mortem analysis of 428 startups: the team is the #1 failure factor for pre-seed/seed startups, responsible for 67% of failures. Two essential corollaries to remember:
Hasty partnerships are lethal (42% of team-related failures)
Complementarity is not optional (38% of successful teams have well-defined complementary profiles)
Too many teams form out of mere convenience: "we worked at the same company" or "we went to the same school." Spoiler alert: these fragile foundations collapse under pressure. The data is unforgiving: 71% of co-founders who have known each other for less than 12 months before launching their startup separate before Series A.
Here's my three-step evaluation methodology: → Analysis of intersecting trajectories
I begin by dissecting individual and collective paths. How did they converge on this venture? Have they weathered storms together? The answers reveal the solidity of their foundation.
Key data I look for:
Duration of the pre-startup professional relationship (ideally >18 months)
Shared crisis management experience (minimum 1 significant event)
Measurable history of shared accomplishment (with verifiable KPIs)
→ Triangulation of skills and temperaments
The perfect balance I systematically seek:
A mercenary, visionary CEO overflowing with energy
(MBTI profile ENTJ/ENTP in 73% of cases)A Cartesian CTO, capable of structuring chaos and saying "no"
(INTJ/INTP profile overrepresented at 68%)A human-centered COO/CPO, driving execution and cultural safeguarding
(ENFJ/ESTJ at 61%)
This specific constellation creates productive internal dynamics and an adaptive external interface with the market.
→ Alignment of intentionality
The deep "why" that drives each founder is revealing. I need macro alignment on ambition, but micro motivations sufficiently distinct to avoid friction. An infallible test: individual interviews with identical questions to detect dissonances that predict who will abandon ship after the tenth door slams.
II. The Tech/Product: The Architecture of Scalability
Beyond the flashy pitches, the technical reality of the product often constitutes the first barrier to scaling. My investigation is structured around three axes:
→ Operational maturity
Brutally revealing metrics that I systematically evaluate:
Average critical bug resolution time (<24h is ideal in early stage)
Ratio of implemented vs. abandoned features (>70% is a strong signal)
Customer feedback and integration cycle (<14 days between feedback and implementation)
Technical team structure and workflow (looking for 1-2 week sprints, not quarterly cycles)
Quantified technical debt (requires an honest audit that reveals team maturity)
→ Depth of IP and data
Particularly crucial in AI ventures, I look for tangible and defensible entry barriers. My fear: financing an AI agent that will be duplicated in two weeks by OpenAI or another giant.
My evaluation criteria for IP/data:
Volume of proprietary data (minimum 100K+ unique entries in niche markets)
Exclusivity of data sources (ideally 40%+ data not publicly accessible)
Defensive patents filed or pending (>2 for deep tech startups)
Unique technical architecture (at least 1 major differentiating component)
Robust feedback loops (system self-improvement mechanisms)
→ Resistance to strategic obsolescence
I am convinced that LLM/SLM architectures will need to diversify to achieve profitability. Giants will inevitably adopt a "copy-paste" strategy of the best AI agents, which they will distribute via their platforms. I therefore meticulously scrutinize the product's ability to create or maintain a defensible advantage against this threat.
To quantify this resistance, I evaluate:
Product iteration speed (<3 weeks per major release)
Number of unique features difficult to replicate (minimum 3-5)
Dependency on external APIs (<30% of critical functionalities)
Unique acquisition cost of proprietary data (must be >2-3x higher than CAC)
Existence of "data moats" with quantifiable network effects (each user must enrich the system cumulatively)
III. The Real TAM vs. The Fantasy TAM
All pitch decks display astronomical TAMs extracted from McKinsey reports. The reality is quite different and constitutes the third pillar of my evaluation. The Startup Genome study reveals that 42% of startups fail due to the absence of a real market for their product – a figure that rises to 67% in B2B SaaS.
→ The delta between theoretical market and addressable market
I'm interested in the real size of targeted segments and the differences in personas across the market. This analysis often reveals a considerable gap between the displayed ambition and the truly conquerable territory.
My empirical 1/100/10 rule:
Real SAM = Theoretical TAM ÷ 100 (on average)
Initially attackable market = SAM ÷ 10 (in seed phase)
In 76% of the pitches I evaluate, market size is overestimated by a factor of 50 to 500. Verification tests I apply:
Decomposition of TAM by addressable sub-segments (minimum 3-5 segments)
Bottom-up market size calculation (preferred to top-down approaches)
Quantified adoption barriers by segment (transition costs, organizational inertia)
Competitive dynamics by sub-market (% already captured by existing solutions)
→ Commercial conquest strategy
How do they plan to address their market? With what sequence? At what cost? I expect surgical knowledge of CAC, sales cycles, friction points, and adoption levers.
Commercial realism metrics I require:
Initial CAC by market segment
LTV/CAC ratio by segment (minimum 3:1 for B2C, 5:1 for B2B)
Average sales cycle duration (with breakdown by stage)
Conversion rate by funnel stage (benchmarked against industry standards)
Acquisition cost of the first 10, 100, and 1000 customers (progression must be realistic)
→ Price/value coherence
Pricing often reflects the real understanding of the market better than any discourse. Disconnected pricing is the relentless indicator of a deeper disconnection with customer needs.
I push founders to quantify:
Customer ROI by segment (ideally >5-10x the solution price)
Tested price elasticity (not theoretical) on a minimum of 3 customer segments
Detailed competitive benchmark (price positioning vs. functionalities)
Willingness-to-pay by customer persona
I ensure that this market analysis is anchored in the founders' personal story and demonstrates a genuine shared conviction. The best founders have often experienced the problem they're addressing.
Successful early-stage investment is not a simple capital-equity exchange. It's a strategic alignment between the founders' gaps and the investor's ability to fill them to accelerate value creation.
The calculation is simple but ruthless:
The Triple-A equation: Ambition × Aptitude × Adaptability = Probability of success
The paradox of weak signals: the most promising startups often present characteristics that trigger alarms in conventional investors
My advice to founders: prepare to be dissected on these three dimensions. To investors: develop your own "Golden Triangle" and apply it with rigor and discernment.
What about you? How do you evaluate early-stage startups? What are your early indicators of success? What warning signals make you flee from a seemingly promising deal?
Previous articles you'll love :
Catch you soon!
Great insights, Arthur!
Great nuanced product questions - love it! I'm wondering if it is worth focusing on TAM that early in the journey though. I've studied multiple VC winners from Seed - Series A/B deck and the majority doesn't get the TAM right in Seed, but it doesn't really seem to matter. Everything you're asking here I'd apply for Series A+ but not before. Thoughts?