What Our 6 Wins and 5 Losses Teach About Investment Risk
A post-mortem across 38+ angel investments: the 3 patterns behind every win, 3 patterns behind every loss, and the risk framework that changed everything.
Whether you're looking for an angel investor, a growth advisor, or just want to connect — I'm always open to great ideas.
Get in TouchAI, startups & growth insights. No spam.
TL;DR: After 38+ angel investments, 6 wins, 5 confirmed losses, and the rest still unrealized, the patterns are undeniable. Every win shared three traits. Every loss shared three different traits. This is the honest post-mortem I wish existed when I wrote my first check.**
I made my first angel investment in 2019. I had built and sold a SaaS product, had some capital sitting in a bank account earning nothing, and had a founder friend who needed a small check with strategic value. I wrote a $25,000 check with almost no formal analysis. That company eventually returned capital but not much more. It was not a win or a loss — it was tuition.
Today the portfolio looks like this:
| Status | Count | Notes |
|---|---|---|
| Confirmed wins (returned >1x capital) | 6 | Mix of exits, secondary sales, significant markups |
| Confirmed losses (written off) | 5 | Shutdowns or zombie companies |
| Unrealized (still active) | 27+ | Mix of early stage, growing, and stalled |
| Total investments | 38+ | Across SaaS, AI, developer tools, fintech, consumer |
The wins and losses feel outsized in importance not because they represent a majority of the portfolio — they don't — but because they are the clearest signal about which decisions correlated with value creation versus value destruction. The unrealized investments are still writing their stories.
The first thing I want to be honest about: 38 investments over six years is a small sample size for statistical conclusions. This is not a Sequoia-style analysis with hundreds of data points. What I have are real patterns from lived experience, and real money either made or lost. I am sharing what I actually observed, not a theory built from second-hand case studies.
The second thing: most angels do not talk publicly about losses. They share wins on Twitter and stay silent on failures. I think that is a mistake for the ecosystem. If we cannot be honest about what went wrong, we cannot improve our process. So here is the unfiltered version.
I spent three months going back through every winning investment, reading my original notes, re-interviewing the thesis I had at the time, and comparing it against what actually happened. Three patterns showed up in all six wins without exception.
Every founder behind a win had two things that are hard to fake: genuine obsession with the problem and a non-obvious path to early distribution.
Obsession is not passion. Passion is enthusiasm. Obsession is when a founder has been living with this problem for years before starting the company, has a perspective that no one else has, and cannot stop thinking about it even when they should be sleeping. You can feel it in diligence conversations. When you ask them about the problem, they go three levels deeper than you expected. They tell you about customer conversations you did not ask about. They correct your assumptions with data they had been collecting for months.
Distribution moat is harder to identify upfront but was visible in retrospect in every win. It usually took one of these forms:
The combination of these two traits — obsession and distribution moat — was the single strongest predictor of a winning outcome in my portfolio.
Every win in my portfolio entered the market when an underlying technology had just crossed a critical threshold. Not before. Not after. Exactly at the transition point.
This is easier to see in retrospect than in real time. But there were signals I learned to look for:
The companies that entered six months too early burned through cash trying to build infrastructure that did not exist yet. The companies that entered three years too late found themselves in a crowded market with no differentiation. The winners entered at the moment when the technology was just capable enough and the market was just aware enough.
One of my wins in the AI tooling space entered the market in early 2023. At that point, large language model APIs had just become cheap enough for the unit economics of an AI-native SaaS product to work. Six months earlier, the token costs would have made the business model structurally unviable. They saw the inflection point and moved before the crowded wave followed.
This pattern surprised me when I first identified it. None of my wins tried to go broad from day one. Every single one found a highly specific, initially underserved customer segment and owned it before expanding.
The non-obvious part matters. If the customer segment is obvious, it is already crowded. The winners found customer segments that incumbents were ignoring, either because the segment was too small to justify the attention of larger players, or because serving them required domain expertise that generalist teams did not have.
One fintech company I backed went after a segment that every major bank had internally labeled as "not worth the cost to serve." The founder had worked at one of those banks and knew exactly why that segment was ignored — and knew the segment's economics were actually quite good if you designed the product correctly from scratch rather than retrofitting an existing bank product. That kind of non-obvious customer segment focus is one of the core patterns behind why AI startups fail when they skip it in favor of going broad too early. They owned that segment for two years before anyone noticed.
The pattern across all six wins: hyper-specific customer focus at launch, then expansion once the initial segment was saturated and the product had enough breadth to appeal to adjacent customers.
| Win Pattern | Visible Pre-Investment | Visible Post-Investment | How to Identify It |
|---|---|---|---|
| Founder obsession + distribution moat | Partially | Fully | Multi-level problem depth; early user network; referral mechanism |
| Technology inflection point entry | Partially | Fully | Recent cost curve shift; new platform API; 12-18 month window |
| Non-obvious customer segment | Partially | Fully | Ignored segment with good LTV; domain expertise; early design partners |
The losses were more emotionally instructive than the wins. When something fails completely, the reasons are usually loud and obvious in retrospect. The question is always why you did not see them clearly enough at the time.
This is different from "bad founder." Every company I lost money on was founded by someone genuinely impressive. The problem was not founder quality — it was founder-market fit. The founder was brilliant but working on the wrong problem for them specifically.
The signals I missed at the time:
In two of my five losses, the founders eventually pivoted the company to a problem they were more personally connected to. In both cases, the pivot came too late and too expensive. The capital had been spent building in the wrong direction.
The honest version of this pattern: I let halo effect bias me. When a founder had an impressive pedigree — a prior exit, top-tier school, previous startup experience — I gave less weight to the question of whether this specific problem was in their bones. It was not.
Three of my five losses were not wrong on the market — they were wrong on the timing of the market. Two were too early. One was too late.
Too early is dangerous because the team looks brilliant, the vision is clearly correct, and the market will eventually be there. The problem is that "eventually" can be three years away. At a startup's burn rate, three years is an eternity. The company runs out of money before the market materializes, or pivots so many times trying to find revenue that the original thesis is abandoned.
Too late is a different failure mode. The market is clearly there, the problem is validated, but competitors have already established distribution and customer relationships that make it nearly impossible for a new entrant to win on product alone. One company I backed entered a market that was genuinely ready — the timing signal was correct — but did not account for how entrenched the market leader had become. The product was better. The distribution could not compete.
The specific signal I missed in the too-early cases: the founding team was relying on analyst reports and conference buzz as evidence of market readiness rather than actual paying customers. "Everyone is talking about this" is not market validation. "I have 10 enterprise customers paying $50k ARR" is market validation.
This is the loss pattern I am most embarrassed about because it was the most avoidable. In three of my five losses, the unit economics were broken at launch and the team was deliberately not looking at them.
The argument for ignoring early unit economics is seductive: move fast, find product-market fit, optimize the business model once you have proven demand. There is a version of that argument that is correct. But I misapplied it to companies where the unit economics were not just unoptimized — they were structurally broken. The product had a cost structure that could never achieve healthy margins at any scale because of fundamental decisions made about the business model.
The specific signals I should have caught:
| Loss Pattern | Signal Missed | When Visible | Avoidable? |
|---|---|---|---|
| Founder-market fit mismatch | Intellectual vs. experiential problem ownership | Pre-investment | Yes, with better diligence |
| Market timing miscalculation | Analyst buzz vs. paying customers | Pre-investment | Partially |
| Unit economics ignored | CAC:ACV mismatch; hidden costs | Pre-investment | Yes, with model review |
I am not going to name the company. The founders are still building and their competitors do not need a roadmap. But I am going to describe the investment in enough detail that the lessons are concrete.
I met this founder through a mutual contact in 2021. They had spent six years as a practitioner in a B2B vertical that most investors found unsexy. Think operations-heavy, mid-market, dominated by legacy software from the 2000s. The kind of industry that does not have a high-profile tech conference or a glossy media narrative.
The founder's pitch was not about AI. It was not about disruption. It was about one specific workflow that every company in their industry did manually, and the observation that this workflow was the number-one source of employee frustration and errors in that workflow cascaded into expensive downstream problems.
The check size was small — under $50,000. The valuation was reasonable. I had no thesis that this was going to be a unicorn. I had a thesis that this founder knew their problem space better than anyone I had ever met in that category, and that the market was underpriced by venture because the industry lacked a sexy narrative.
The company found product-market fit quickly. Not because they had a revolutionary product — the first version was genuinely unremarkable — but because the founder already had 40 warm introductions to potential customers on day one. When you spend six years building credibility in an industry, you have a distribution network that a first-time founder pitching the same product would spend 18 months building from scratch.
The first 20 customers came in six months, all from the founder's existing network. The product got better because the feedback loop was tight and the founder understood what customers were actually asking for versus what they said they wanted. By month 18, the company had enough case studies that inbound from the same industry segment started arriving without paid acquisition.
The inflection point came when the company added a feature that automated a second, adjacent workflow. That feature increased ACV by 40% without meaningfully increasing CAC, because existing customers bought the expansion immediately and the sales pitch to new customers now covered two pain points instead of one. NRR crossed 120%.
A strategic acquirer from adjacent software bought the company. The acquirer wanted both the technology and the customer relationships — particularly the founder's industry credibility and the fact that 60% of the customer base was on multi-year contracts.
My return was meaningful but not fund-making on its own. What made it instructive was that every component of the outcome was visible — not certain, but visible — at the time I wrote the check. The founder's domain depth, the warm distribution network, the underpriced market, the tight product feedback loop. The win was not luck. It was the accumulation of identifiable advantages compounding over three years.
Unsexy industries are often underpriced precisely because they are unsexy. The lack of investor attention means the founder can close rounds at reasonable valuations, the press is not following the company and creating noise, and the customer base is stable and not being constantly pitched by 15 competing startups.
Distribution advantages compound faster than product advantages. A better product can be copied. A founder's six years of industry relationships cannot be replicated quickly by a well-funded competitor. The founder-led growth playbook shows how the best operators systematically build that distribution moat before it becomes obvious to the market.
Small checks into high-conviction thesis investments outperform large checks into fashionable thesis investments. The fashionable investments attract crowded cap tables, high valuations, and competition from better-capitalized firms. The unsexy investments let you price in and own a meaningful stake.
This one I will describe more carefully because the lessons are more uncomfortable.
The company was in a category I found intellectually exciting in 2022. The founders had impressive backgrounds — one had a prior exit, the other had deep technical credentials. The market narrative was strong. Every analyst report I read confirmed the category was growing fast.
I wrote a larger-than-usual check. The valuation was high but defensible given the competitive round and the investor pedigree on the cap table.
Looking back at my original diligence notes, the warning signs were present. I documented them and then rationalized them away.
Signal 1: The founders could not show me a customer who had renewed. Their product had been live for eight months. The initial customer cohort was signing on, but no one had come up for renewal yet. I noted this in my diligence document and wrote: "Early stage, expected." What I did not adequately model was that in their pricing tier, customer lifetime was critical to unit economics. Eight months of customer history was not enough to know anything real.
Signal 2: The go-to-market was entirely inbound from the category hype. Customers were showing up because the industry press was hot on the category. The company had essentially zero outbound motion. When category hype fades — and it always fades — there was no engine to replace it. I noted this as "early-stage distribution dependency" and did not weight it appropriately.
Signal 3: The CAC was described to me verbally, not in a model. When I asked for the unit economics deck, the founder sent a spreadsheet that had placeholder assumptions in some cells. The CAC figure was an estimate based on three months of pipeline data, not closed customers. I asked follow-up questions but did not push for actual closed-customer CAC before wiring funds.
The category hype faded. Several well-funded competitors entered the space with substantially lower prices. Churn picked up in months 10-14 as initial customers completed their first year and either did not renew or negotiated down significantly on renewal pricing.
The company raised a bridge round at a flat valuation to extend runway. I did not participate. The bridge bought them 18 months. They went through two pivots. The second pivot produced some traction in an adjacent category but not enough to raise a Series A. The company shut down after the founders accepted jobs at larger companies.
The total loss was 100% of my invested capital.
The round composition is a signal. When a founding team has access to top-tier institutional money but their first check is coming primarily from angels in a competitive round, ask why the tier-one firm did not lead. Sometimes the answer is fine. Sometimes the answer reveals that sophisticated investors saw something I missed.
Inbound demand driven by category hype is not the same as product-market fit. Product-market fit means customers are paying for your specific product because of your specific product's value. Category hype means customers are trying things because the analyst reports told them to. These produce the same revenue line in early months and completely different retention curves in month 12.
I had too much money in this check relative to my conviction level. Larger check size did not correspond to higher conviction — it corresponded to FOMO about the competitive round. When check size is driven by FOMO rather than conviction, you have already made the error.
After loss number three, I built a formal diligence framework. After loss number five, I updated it significantly. Here is what changed:
I now require three reference checks — two professional, one from a former employee if the founder has led a team before — before committing to any investment above $25,000. The full pre-investment framework is in the 5 angel checks I run before writing a check. This added friction that killed two deals I wanted to do on a quick timeline. Both companies I missed doing well, and I have no regrets about the discipline because the framework has also caught things I would have missed.
The reference check questions that have been most revealing:
I now ask for a unit economics model built from actual customer data, not projections. For pre-revenue companies I ask for a bottoms-up model with their assumptions made explicit. The question is not whether the model is right — early-stage models are always wrong. The question is whether the founders understand the key drivers well enough to articulate them, and whether there are any structural problems baked in.
Structural unit economics problems I now check explicitly:
I restructured position sizing around three conviction tiers:
| Conviction Tier | Criteria | Check Size |
|---|---|---|
| Tier 1: Highest conviction | Know founder personally, deep domain, clear distribution | $50-100k |
| Tier 2: High conviction | Strong diligence, meets all framework criteria | $25-50k |
| Tier 3: Exploratory | Interesting thesis, incomplete diligence or early signal | $10-25k |
No check is written in Tier 1 based on round pressure or FOMO alone. The Tier 1 classification requires me to be able to articulate clearly why this is Tier 1 in a written document before transferring funds.
After processing wins and losses, I collapsed everything into a risk matrix that I now use as my primary diligence filter. Every investment gets scored on five dimensions before I commit.
| Risk Dimension | What I'm Evaluating | Red Flag | Yellow Flag | Green Flag |
|---|---|---|---|---|
| Founder-market fit | Experiential vs. intellectual connection to problem | First-time in domain, no adjacent experience | Adjacent experience, strong learning velocity | Deep domain, prior operator in exact space |
| Market timing | Technology readiness and market awareness | Analyst reports only, no paying customers | Early customers, category emerging | Paying customers, category inflecting |
| Distribution | Path to first 100 customers | No network, no channel, relying on ads | One channel, untested at scale | Multiple distribution vectors, early conversion data |
| Unit economics | Structural business model health | CAC > ACV, margin < 50% | CAC:ACV 1-1.5x, margin 50-65% | CAC < 0.5 ACV, margin > 70% |
| Capital efficiency | How long capital buys and what it buys | 12 months runway at current burn with no clear path | 18 months, defined milestones | 24+ months, clear fundraising narrative |
A company needs four greens and one yellow to get a Tier 1 check from me. Three greens and two yellows gets a Tier 2 check. Any red flag requires explicit documentation of why the red flag is acceptable before I write a check.
One insight that took too long to learn: risks in the same company are often correlated in ways that are not obvious at investment time. A market timing risk often correlates with unit economics risk, because a company that is too early will spend down capital building market awareness before customers are ready to pay, which distorts the real cost of customer acquisition. A founder-market fit risk often correlates with distribution risk, because a founder without deep domain experience often lacks the relationships to build an efficient distribution channel.
When I see two yellow flags in correlated dimensions, I now treat that as equivalent to a red flag.
Most angels I know built their portfolios the wrong way: opportunistically, check by check, following deal flow without a portfolio-level thesis. I did this for the first three years. It produces a portfolio that is fine but not optimized for the asset class. The diversification strategy post covers the power law math and why portfolio breadth matters more than most angels admit.
The asset class math for angel investing is unforgiving if you do not understand it going in:
That last point is the one most angels intellectually accept and practically ignore. If all the returns come from outliers, your portfolio construction should be designed to maximize the probability of holding a meaningful position in an outlier, not to minimize losses or pick the most likely winners.
There is genuine disagreement among experienced angels about whether to concentrate (fewer, higher-conviction investments) or diversify (many smaller investments). My current view:
The case for concentration: If you have genuine edge in identifying quality investments, concentrating capital in your highest-conviction bets maximizes returns. A $250k check in a company that returns 100x generates $25M. A $25k check in the same company generates $2.5M. The information you have at investment time matters enormously.
The case for diversification: If you do not have systematic edge — and be honest with yourself about whether you do — then diversification gives you more chances to accidentally hold an outlier. Across a 25-investment portfolio with $25k checks, you need approximately two investments to return 10x and one to return 50x to generate a 3x net portfolio return. That is achievable.
My current approach is a hybrid: diversify broadly in Tier 3 investments to maintain deal flow and outlier exposure, concentrate in Tier 1 investments where I have genuine conviction and relationship.
| Portfolio Size | # of Investments | Average Check | Expected Composition |
|---|---|---|---|
| $250k-$500k | 10-15 | $25-35k | Mainly Tier 3 exploratory; 1-2 Tier 1 conviction |
| $500k-$1M | 15-25 | $30-50k | Mix of Tier 2 and Tier 3; 3-5 Tier 1 conviction |
| $1M-$3M | 25-40 | $40-75k | Diversified with meaningful Tier 1 positions |
| $3M+ | 40+ | $50-100k | Institutional-style portfolio construction |
One of the most value-destroying mistakes an angel can make is deploying 100% of capital in initial investments with no reserve for follow-ons. The companies in your portfolio that are performing well will raise again, often at higher valuations, and your ability to participate in those rounds — to maintain pro-rata rights — is one of the primary ways angels generate returns beyond the initial check.
I now allocate approximately 30-40% of my angel capital for follow-on investments in winners. That means my initial check deployment rate is lower than it would otherwise be, but the portfolio return profile is substantially better.
Let me give you the honest numbers, because most introductory angel investing content glosses over the math.
Multiple studies of angel portfolio returns — including Kauffman Foundation research, AngelList data, and academic studies of angel networks — produce roughly consistent findings:
| Outcome Category | % of Angel Investments | What Happens |
|---|---|---|
| Total loss (0x) | 50-60% | Company shuts down or goes zombie |
| Return of capital or small gain (0-2x) | 20-25% | Modest exit, acqui-hire, or fire sale |
| Meaningful return (2-10x) | 10-15% | Good outcome, contributes positively |
| Strong return (10-30x) | 5-8% | Great outcome, anchor of portfolio return |
| Outlier return (30x+) | 1-3% | The investment that drives portfolio returns |
In my portfolio, my distribution is roughly consistent with these base rates, though the sample size is too small to draw strong conclusions.
If 1-3% of investments drive portfolio returns, and you make 20 investments, you may get zero outliers by chance. This is not a failure of judgment — it is the statistical reality of the asset class.
The implication: to have a statistically reasonable chance of holding at least one outlier investment, you need to make more investments than most angels make. The minimum portfolio for credible outlier exposure is approximately 25-30 investments. Below that number, you are playing a lottery with a small ticket count.
| Portfolio Size | Investments | Expected Return Multiple | What Generates the Return |
|---|---|---|---|
| $250k | 10 | 1-2x | No meaningful outlier probability |
| $500k | 20 | 1.5-3x | Statistical chance of 1 meaningful outlier |
| $1M | 30-35 | 2-5x | Probable 1-2 outliers if deal quality is reasonable |
| $2M+ | 50+ | 3-10x | Portfolio starts to reflect asset class base rates |
Looking at my six wins:
This is the power law in a real portfolio. The top two investments are not 2x better than the others — they are 10-20x better. The lesson: the most important decision is not avoiding losses. The most important decision is ensuring you have meaningful position size in potential outliers.
The corollary to that lesson, which took me too long to internalize: the five losses cost less in total than what I would have made from investing that same capital in a higher-conviction Tier 1 position. The opportunity cost of bad investments is not just the capital lost — it is the capital that could have been deployed into known winners at earlier stages.
Deal flow is entirely relationship-dependent at the angel stage. My deal flow comes from four sources: founders I have previously backed introducing me to other founders, operator networks I have built over years in the startup ecosystem, co-investors who share deals across their networks, and a small number of angel networks and scout programs. I do not source well from cold inbound — my hit rate on unsolicited pitches is extremely low.
At minimum: pitch deck, financial model with unit economics assumptions, cap table, any signed customer contracts or LOIs, and three reference contacts. For Tier 1 investments I also request access to a data room with operating metrics, any legal diligence (IP ownership, employment agreements), and founder background information.
For Tier 3 exploratory investments: two to three weeks. For Tier 2: three to six weeks. For Tier 1: up to three months. The process length has increased substantially since my early investing years. My hit rate on investments that meet my framework criteria has improved accordingly.
My policy is that domain knowledge gaps can be compensated for by strong co-investors with domain knowledge, or by exceptional founder credibility in the domain. I will not lead a Tier 1 investment in a domain I do not understand without a co-investor I trust who does. For Tier 3 investments I accept more domain uncertainty because the check size is small enough that it functions as an option rather than a concentrated bet.
The question is whether there is credible evidence that the core thesis is still intact. If the struggling company has found a different path to the same destination — same customer problem, different distribution, different product approach — I may participate in a bridge. If the company is pivoting away from the original thesis entirely, I treat the decision as a new investment with new diligence rather than doubling down on sunk cost.
Writing too few checks at too high a valuation in too fashionable categories. The three variables compound: fashionable categories attract high valuations, and high valuations require exceptional outcomes to return capital. First-time angels are also more susceptible to halo effect bias — investing in impressive founders regardless of founder-market fit.
Proactively and honestly. I reach out quarterly to every portfolio company regardless of performance. For struggling companies I offer specific help rather than generic encouragement — introductions, recruiting, customer connections, whatever is actually useful. I try to be the investor who founders can tell bad news to, because the earlier I know about problems, the more likely I can help address them.
With the benefit of perfect hindsight, yes. With the benefit of what I knew at the time of investment, I regret two of the five. Those two had red flags I chose to rationalize. The other three had weaker signals that I could not have caught with better diligence alone — the companies ran into market conditions or execution challenges that were genuinely difficult to predict. I try to distinguish between process errors (which I should have caught) and outcome errors (which even good process cannot guarantee).
It depends on your definition of worth it. From a pure financial return perspective, at $500k deployed across 20 investments, your expected return is modestly positive if your deal flow is reasonable — but not dramatically better than index fund returns when adjusted for risk and liquidity. The value of angel investing at that scale is not primarily financial. It is the access to deal flow that compounds into better deal flow, the network of operators and founders, the pattern recognition that helps your own companies, and the asymmetric upside if you happen to hit a real outlier.
If you are deploying $500k purely for financial return, index funds are probably better. If you have genuine edge in a domain and are willing to do the work, angel investing can be exceptional. Most people should not be angel investors. The ones who should know who they are.
The competitive dynamics for good deals are much more intense. In 2020, a thoughtful angel with domain expertise could lead seed rounds in genuinely good companies at reasonable valuations. In 2026, every good company has a cap table with AI-focused micro-funds, operator syndicates, and deep-pocketed angels competing for allocation. The advantage of being a check writer has diminished; the advantage of being the angel who brings genuine value — customers, hires, strategic guidance — has increased. The best deals I see today are going to angels who are also operators, board members, and advisors before they are check writers.
The exact pre-investment checklist I run before every angel check: founder-market fit, timing, unit economics, references, and momentum — with scoring tables and decision matrix.
A rare look inside a 38-company angel portfolio: the tracking system, vintage analysis, wins, losses, and the decision framework that evolved over seven years.
A brutally honest case study of a portfolio company failure — what we missed, what went wrong, and what I changed about my investing process after.