Venture Partner @ Voyager 🚀, founder/ex-student #VC @ Campus Capital Oxford 🏫, founder @ stealth #startup, idiot @ large 🌏
9177 words

praxeum - the failure log #1

I've been working on my startup for a year - and it hasn't gone anywhere close to the way I would have wanted it to.

It's been a truly humbling, soul crushing exercise. Customer acquisition has been slow, as has been fund raising. Having previously started funded projects and gotten job offers and scholarships easily, I thought the success would be easy. I really didn't appreciate how lucky I was.

I've had to learn to tame my jealousy as the corporate alternatives I've forfeited or the peers who are "ahead" creep into my mind.

I've had to learn to check my ego and lick my wounds as I get rejected and found consolation in the philosophy of the Stoics and the Daoists.

I've had to count my blessings at my past successes - and to humble myself in the realisation that success is a lot of luck. I'm really not as great as I think I am.

But, I don't have any regrets, I think. I took a risk - and I'm happy to take the consequences. I prefer hard experiences and harder failures, then the perfectly paved road imagined for me.

I've learnt more and grown to be a better human being. My life before - untested with low stakes and no failures has been a cocoon. This risk I took prompted me to take a deeper look at myself, and my beliefs about risk, luck, rewards, individuals and society. For once, I've truly had to examine and understand myself, and experiences outside my Hong Kong/international school/Oxford bubble.

I've learnt to be a better entrepreneur and VC. I know what's at stake now - and know that every late cheque I write or decision I delay impacts.

I've also (re-)learnt, that I don't give up. I apparently keep swinging. I'm so grateful for the friends who believe in my mission - Heidi and Gordon, who are pushing us to new heights with this latest rollout. This week is the beginning of the prax+ rollout - another round of engagement with schools, teachers and parents.

Here's to hoping that after this year of learning, this product I've poured so much love in, will find some love with users. I hope that in its quest to democratise and disrupt education, prax+ will stand for more than the better IGCSE preparation - and will one day enable someone's failures and growth. I hope it increases the potential of our next generation - we're kinda counting on them.

Startup/VC Metrics - SaaS

More personal notes - lifted from the best blogs

Business and Financial Metrics

  • Bookings (value of contract) vs. Revenue (recognised when revenue provided) NB Letters of intent and verbal agreements are neither revenue nor bookings.
  • Recurring Revenue vs. Total Revenue ARR (annual recurring revenue); ARR per customer: Is this flat or growing?; MRR
  • ARR should exclude one-time fees, professional service fees, and any variable usage fees.
  • Gross Profit
  • Total Contract Value (TCV) vs. Annual Contract Value (ACV) Re ACV - is it growing, i.e. are you adding more value?
  • LTV
    • Revenue per customer (per month) = average order value x by the number of orders
    • Contribution margin per customer (per month) = revenue from customer - variable costs associated (selling, administrative and any operational costs associated with serving the customer)
    • Avg. life span of customer (in months) = 1 / by your monthly churn.
    • Why? When you graph average lifespan of customer against monthly churn, because churn is a % of customers leaning, you have an exponential decay graph.
    • LTV = Contribution margin from customer x by the average lifespan of customer.


  • Gross Merchandise Value (GMV) vs. Revenue
    GMV (how much the consumer side of the marketplace is spending) as measure of the size of the marketplace.

  • Unearned or Deferred Revenue and Billings

  • SaaS companies only recognize revenue as the service is delivered, so that “booking” is a liability line item called deferred revenue.

  • Billings is a much better forward-looking indicator of the health of a SaaS company than simply looking at revenue because revenue understates the true value of the customer, which gets recognized ratably.

  • A SaaS company could show stable revenue for a long time — just by working off its billings backlog — which would make the business seem healthier than it truly is.

  • CAC (Customer Acquisition Cost) … Blended vs. Paid, Organic vs. Inorganic

Product and Engagement Metrics

  • Registered Users (vanity)
  • Active Users definitions
    • In social and mobile platforms: MAUs, WAUs, DAUs, and HAUs -On content sites: “uniques” and visits
      -On e-commerce sites: no weight on MAU/DAU
  • Month-on-month (MoM) growth
    • CMGR [CMGR = (Latest Month/ First Month)1/# of Months -1] also helps you benchmark growth rates with other companies.
  • Churn: Monthly unit churn = lost customers/prior month total

  • Retention by cohort

    • Month 1 = 100% of installed base
    • Latest Month = % of original installed base that are still transacting
  • Gross churn: MRR lost in a given month/MRR at the beginning of the month.

  • Net churn: (MRR lost minus MRR from upsells) in a given month/MRR at the beginning of the month.

    • Gross churn estimates the actual loss to the business, while net revenue churn understates the losses (as it blends upsells with absolute churn).
  • Burn Rate: Monthly cash burn = cash balance at the beginning of the year minus cash balance end of the year / 12

    • Net burn [revenues (including all incoming cash you have a high probability of receiving) – gross burn] is the true measure of amount of cash your company is burning every month.
    • Gross burn on the other hand only looks at your monthly expenses + any other cash outlays.
  • Downloads (but a vanity metric)

Business and Financial Metrics

  • Total Addressable Market (TAM)

    • But using the size of an existing market might actually understate the opportunity of new business models:
    • Bottom-up analysis, preferred, which takes into account your target customer profile, their willingness to pay for your product or service, and how you will market and sell your product.
  • Average Revenue Per User (ARPU)
    For pre-revenue companies, investors will often compare the prospects of a company against the known ARPU for established companies. For example, we know that Facebook generated $9.30 ARPU in FY2015Q2 from its U.S. and Canada users.

  • Gross Margins
    A few examples to illustrate the point: E-commerce businesses typically have relatively low gross margins, as best exemplified by Amazon and its 27% figure.

    • cf most marketplaces (note here the distinction between e-commerce) and software companies should be high gross-margin businesses.
  • Sell-Through Rate & Inventory Turns

    • Sell-through rate :number of units sold in a period divided by the number of items at the beginning of the period
    • In marketplace businesses: like high + improving rate so that suppliers are seeing good returns on the effort they put into posting listings on the marketplace.
    • In businesses that buy any kind of inventory — retailers, wholesalers, manufacturers — the sell-through rate is a key operating metric for managing inventory on a weekly or daily basis. It can reveal how well you matched supply of your product to demand for it, on a product-by-product basis.
    • Inventory turns is a more useful metric than sell-through rate in inventory-based businesses, because it:

      • Talks to the capital efficiency of the business
        • Provides clues as to the quality of the inventory, i.e. demand

Economic and Other Defining Qualities

  • Network Effects shown by, e.g.

    • The sales productivity of OpenTable grows substantially over time, as large increases in the number of inbound leads.
    • The number of diners seated at existing OpenTable restaurants grows substantially over time.
    • The share of diners who come directly to OpenTable to make their reservation (versus going to the restaurants’ websites) grows substantially over time.
  • Virality is the speed at which a product spreads from one user to another.

    • NB viral growth does not necessarily indicate a network effect.
    • Virality is often measured by the viral coefficient or k-value:
    • Here’s the basic math behind the k-value [there are some other more nuanced and sophisticated calculations here]:
      1. Count your current users.
      2. Multiply that count by the average number of invitations that your user base sends out.
      3. Figure out how many of those invited users took the desired action within a defined period of time. As with all measurements, pick a meaningful metric for this action. =. This means you started with 1,000 people and ended up with 1,750 people through this viral loop during your defined time period. The viral coefficient is the number of new people divided by the number of users you started with; in this case, 750/1000 = 0.75.
    • Anything under 1 is not considered viral; anything above 1 is considered viral.
  • Economies of Scale, e.g. decreasing unit cost over time.

Other Product and Engagement Metrics

  • Net Promoter Score (NPS):

    1. Ask your customers the above question and let them answer on a 0-to-10 Likert-type scale, with 10 being definitely likely
    2. % of promoters = number of respondents who ranked 9 or 10, divided by total number of respondents
    3. % of detractors = number of respondents who ranked ≤ 6, divided by total number of respondents
    4. NPS = % of promoters minus % of detractors

    • Obvious issues: only surveying a subset of customers, comparing companies, which leads to misunderstanding and gaming scores. Rather, focus on same company NPS trends — and pay close attention to optional comments from users.
    • When looking at NPS, we look for a couple of things:
  • Cohort Analysis

    • Cohort analysis breaks down activities/ behavior of groups of users (“cohorts”) over a specific period of time that makes sense for your business — for example, everyone who signed up for your service in the first week of January — and then follows this group of users longer term: Who’s still using your product after 1 month, 3 months, 6 months, and so on?
    • Here are the steps for a cohort analysis:
      1. Pick the right set of metrics rather than a vanity metric (like app downloads)
      2. Pick the right period for a cohort — this will be typically be a day, a week, or a month depending on the business
      3. Period 1 (day, week, or month) — 100% of install base takes some action that is a leading indicator for revenue, such as buying a product, listing a product, sharing a photo, etc.
      4. Period 2 — calculate the % of install base that is still engaging in that action a week or month later
      5. Repeat the analysis for every subsequent cohort to see how behavior has evolved over the lifetime of each cohort

    -The two trends we like to see in cohort analyses are:
    1. Stabilization of retention in each cohort after a period such as 6 or 12 months.
    2. Newer cohorts performing progressively better than older cohorts.

  • Sources of Traffic

  • Customer Concentration Risk: revenue of your largest customer or handful of customers relative to total revenue. So if your largest customers pay you $2M/year and your total revenue is $20M/year, the concentration of your largest customer is 10%.

Startup/VC Investment Math

My own personal primer.


Pre-money valuation: $5M
Founder owns: 5,500,000 shares
Total shares: 10,000,000 shares

  1. Simple equity investment, e.g.
    Investment of: $2M
    = “Post-money” valuation: $7,000,000
    = Investor purchasing 28.57% of the company ($2M/$7M).

  2. Price per share = pre-money valuation / number of fully diluted shares, e.g. Price per share = $5M / 10,000,000 shares = $0.50

    Sanity check: investor’s $2,000,000 will buy at $0.50 per share for a total of 4,000,000 shares. 4,000,000 shares/14,000,000 shares = 28.57%.

  3. Founder's resultant ownership:
    Number of shares he owns / by the total number of shares (5,500,000 / 14,000,000) OR the percent before * the decrease of ownership (55% x (100%-28.57%))


Pre-money valuation: $15
Investment: $5M
Pre-pool and pre-money shares: 10M

Without employee pool:
Price per share = pre-money valuation / number of fully diluted shares
Price per share = $15,000,000 / 10,000,000
Price per share = $1.50

  1. With employee pool, e.g. 10%
    The pre-money valuation takes this into account.
    Price per share = pre-money valuation (which doesn’t change) / number of fully diluted shares + option pool shares (which does change)

    Note: when an investor demands that the option pool is a certain size, it almost always means that the option pool size is that certain size after the financing is complete.

    Post-money = founder's equity + 10% of rest (employee pool) + investor's equity = founder's equity + 10% of rest (employee pool) + ($5M/$20M) of rest
    = 65% (original) + 10% employee pool + 25% investor
    = 65% [$13M] [Founders] + 10% [$2M] [Option Pool] + 25% [$5M] [Investors] = 100% [$20M] [Total]

    = Price per share, pre money is $1.30.


Convertible debt converts into equity at a qualified financing.
Founders own: 1,000,000 shares = 100% of the company

  1. 1st Investment: Investor CD comes along and invests in the company with a convertible debt deal done: $100,000, 4% interest, qualified financing of $1M, valuation cap of $3M. 20% discount.

    Convertible debt does not affect the cap table/company ownership if and until the debt converts. Nothing changes.

  2. 2nd Investment: Investor Y comes along and invests $200,000 at a valuation of $800,000 (not a convertible debt deal.) What does the cap table look like?

    The convertible debt did not get triggered because the qualified financing of $1M was not reached.

    Price per share = 800,000 / 1,000,000
    Price per share = $0.80
    Shares purchased by investor: $200,000 / $0.80 = 250,000

    Founders: 1,000,000 shares = 80% of the company
    Investor Y: 250,000 shares = 20% of the company

  3. 3rd Investment: Investor Z comes along and invests: $1M at a pre-money valuation of $2M. What happens in that case?

    Price per share = pre-money valuation / number of fully diluted capitalized shares.
    Price per share = 2,000,000 / 1,250,000 = $1.60
    Shares purchased: $1,000,000 / $1.60 = 625,000

    Investor CD's terms: $100,000, 4% interest, qualified financing of $1M, valuation cap of $3M. 20% discount.

    That convertible debt will convert into shares at the lesser of (a) 80% of the price per share paid by the purchasers in the qualified financing (remember the 20% discount); or (b) the valuation cap divided by the fully-diluted capitalization immediately prior to the closing.

    80% of the $1.60 is $1.28
    OR the valuation cap is $3M, so the price per share if the valuation is $3M is: Price per share = 3,000,000 / 1,250,000 = $2.40
    So the price per share will be $1.28 as it is lower.

    Number of shares purchased by Investor CD via conversion: $100,000 / $1.28 = 78,125

    ####Resulting cap table after Investor Z:
    Founders = 1,000,000 shares = 51.2%
    Investor Y = 250,000 shares = 12.8%
    Investor Z = 625,000 shares = 32%
    Investor CD: (conversion) = 78,125 shares = 4%

GETTING OUT - how about when investors exit the deal with a liquidation preference?

Liquidation preference gives preferred shares the right to be paid out first following a liquidation event. If a startup is liquidated for less than the investors invested, then their liquidation preference would allow them to get money while the common shareholders get nothing.

  1. The key to understanding liquidation preference is the liquidation preference multiple (bolded).
    Startup acquired for $15M Investor W invested $5M with 2x liquidation preference = investor gets $10M before any common shares paid out

What about the rest of the proceeds (the remaining $5M)?

  1. Full participation (fully participating Preferred Stock)
    E.g. the Series A Preferred (Investor W) participates with the Common share pro rata on an as-converted basis

    VC invests $10M of 1x fully participating Series A Preferred at a $10M pre-money valuation = VC would own 50% of the common stock.

    If the startup is acquired for $60M, the VC will take $10M (from liquidation preference) + $25M (from participation), for a total of $35M.

  2. Capped participation (partially participating preferred)
    This means that holders of Series A Preferred stock are paid their liquidation preference preferentially (meaning before common stockholders get a dime) and then they also get to share any of the proceeds that are left over, with the common stockholders until an aggregate of X times the original investment is reached.

    Even in capped participation, the VC still has the option to convert their shares to common and participate fully in the proceeds, but the entire class of shareholders must agree to forego their participation rules.

    VC invests $10M with 1x liquidation preference with a $30M (3x) cap, and that $10M represented a 50% ownership stake in the company.

    If the company sold for $60M:
    *the investor would get $10M (liquidation preference) plus $20M of the remaining proceeds to hit their cap of $30M (50% of the remaining proceeds, $22.5M, would put them above their cap).
    *If they converted all their shares to common and took 50% of the proceeds, they would also receive $30M.

    If the exit value were $65M:
    *the investor would get $10M (liquidation preference) plus $20M of the remaining proceeds to hit their cap of $30M (50% of the remaining proceeds, $22.5M, would put them above their cap).
    *the investors will get a better outcome by converting to common $32.5M, as no ceiling at $30M.

  3. Non-participation (straight preferred)
    With non-participation, “the balance of any proceeds shall be distributed pro rata to holders of Common share.”

    For example, let’s say a VC invests $10M on 1x non-participating Series A Preferred for a 50% ownership stake in the company.

    If the startup is acquired for $60M:
    the VC can either take $10M (from liquidation preference) + $0M (due to non-participation)
    OR take pro rata share of 50% for a payout of $30M

  4. The liquidation stack - seniority structures to determine which class of shares get paid out first:
    *Standard. Later stage investors receive their liquidation preference first (e.g. D, C, B, then A).
    *Pari passu. All investors receive their liquidation preference at the same time.
    *Tiered. Classes are grouped together (e.g. F and E, D and C, B and A).

Digital 📱 Prescriptions 💊

What is it?

Prescription Digital Therapeutics (PDTs) are basically prescription medication - but in a digital form, i.e. replacing a pill with the phone.

PDTs are not Digiceuticals - the Luminosity 🧠💪 of the world. PDTs are subject to randomized clinical trials 👩‍⚕️ and held to the same standards of other prescription medicine. Digiceuticals aren't - they're wellness products similar to the conventional Vitamin D supplement.

How effective are they? 📈

Very. By definition, PDTs are required to meet the same standards as other conventional therapies. They have received state recognition, e.g. FDA approval. As a result, they are just as if not more effective effective than conventional therapeutics.

Getting people to buy into this, however, will be the problem. So, for the uninitiated, let's break down why PDTs would be effective.

Let's start by refuting prevalent and subconscious concerns. The reason most people feel uneasy about PDTs are because the prototypical approach to disease is as follows:

person 🤒 is sick ➡️ person gets drug 💊 or treatment 🩺 ➡️ drug does science magic 🧬 ➡️ person gets better 😄

The thinking goes, PDTs don't really fit into this framework - because what change can an app that doesn't interact with the body do?

  1. Through behaviour modification 😢➡️😋 - a lot of conventional treatments work through behaviour modification and not the interaction of biological and chemical matter - and indeed in areas like cognitive behavioural therapy we have seen great strides in PDT. PDTs can either work as an additional tool to bridge the gap between efficacy (potential impact) and efficiency (real life effectiveness), or, as replacement themselves. It's easy to grasp that smartphones can change our behaviour to hook us onto Facebook, so it should be easy for them to change our behaviour in ways beneficial for our health.
  2. AI + Data 🖥 - PDTs are often distilled over a smartphone, allowing for the use of its technology to better track behaviour. This provides valuable, quantitative feedback as to how well the patient is doing. In situations where treatment isn't just a drug-and-done, and where there is experimentation, this will provide much quicker routes to personalisation and effective treatment.
  3. Constancy 🏭 - there is a certain level of replicability and guaranteed levels of treatment.

  4. Low cost 🤑 and ease to replicate - in addition to being great alternative tools in the fight against disease, they also make possible the cheap and quick dissemination of personalised treatment to many people .

In short, the inability to trust PDTs stems from fundamental assumptions of our health and care system: that illnesses is a state that can be fixed with chemical reactions. 🧬

This ignores that very real mental 🧠 and trial-and-error 🧪component of many illnesses, and assumes an unshakeable acceptance in the shortcomings of current methods of therapy.
This perhaps speaks to the incredible amount of privilege, owing to the efficacy of 🏥🚑 modern medicine being good enough to make us reluctant of experimentation. This would be a mistake - as healthcare and progress much march forward. As I've mentioned before - one of the fundamental barriers of healthcare innovation is us - and the paradigms we choose: from [false binaries between "well" and "ill"]pe( to ignoring the mind in healthcare.

What to do? The G2M

Many others more qualified than me, have already pondered on how to physician buy-in is essential, and how to integrate it into their workflows, lower the learning curve, support early adopters, and create confidence. Many others have discussed best ways of monetisation - from D2C to working with insurers. The jury is out on all of that - and I'm excited to see what will happen.

I want to talk more about the power of habit. The PDT struggle to get buy-in reminds me of the story of Febreeze 🧼 - and how it struggled to get people to buy its scentless, odour-busting product at first - until they added a useless scent to piggyback off an existing habit of cleaning. I wonder if PDT might be better of Trojan-horsing it as well. Instead of selling it as a replacement to therapeutics, we might find better use by bundling it with placebos - and selling them as a "medication enhancer" and "tracker". We're adding into the habit loop, rather than changing it entirely - which can only be much easier.

Open Source 🔓 Venture Capital 💰

Quick thoughts today - I've been talking to a lot of VCs recently, and was surprised to find out about a major pain-point shared: purpose built tools. 🛠 (Some research shows some exist 😅, but are apparently not widely used yet.)

On second thought 🤔, it makes a some sense. There are only 1000 VC firms in the US, and maybe a handful more angel investors - so you can imagine the total number of potential users would be quite small, even globally. This also isn't a Bloomberg Terminal situation - VCs don't really need these tools enough to justify paying large enough subscriptions to make it profitable. Though, Crunchbase does exist 🤷🏻‍♀️ ...

So I don't expect any sort of real startup to exist in this area, but I could see a set of open source tools built by the community, or by a smaller startup. Some government/startup communities are building tools that soft of touch on the periphery of this.

Ok, on to the tools/features

  1. Data driven decision making 📊

    • Increasing deal flow data, e.g. scraping connections made on Linkedin by other VCs 🗂
    • Surface data 📈: industry data, scraping Twitter and FB for word of mouth deals
    • Dynamic filters - based on NLP etc.
    • Dealing with data: scaling and hacking where sample sizes are small, or ambiguity is high❓
  2. Deal flow and process management 📨➡️

    • Plug into Twitter and other promotional places 📣
    • (Bonus: ability to share deal flow with other corporate arms/student VC teams) 🌏
    • Track deal on investment process
    • Decision making conciliation and tracking - Bridgewater-like tools to help make and track decisions
    • Post-cheque tracking and management - included rejected startups, to see how to improve and reflect on issues ✍️
    • Relationship 🤝 management - NLP/data to analyse relationships with portfolio
    • Team/collaboration tools 👩🏻‍💻👨🏿‍💻
  3. Sharing deal-flow + LP control 🔐 and access + MFA/2FA

    • With the rise of new VC funds and VC-as-a-service, it would be prescient to realise LPs sometimes want to see deals as well
    • Sharing with other VC firms, dealflow better suited to them

The upside to this

The upside to the conversation, is that there seems to be a growing consensus that VC game is becoming less dependent on connections, and more about making good decisions and helping the portfolio. The data is out there - not tidy, not centralised or necessarily apparent, but it is out there. That's a huge thing!

It's a hopeful flag towards a march towards increased diversity 🏳️‍🌈👨‍🦽 in VC and startup founding.

The march towards democratisation: open source resources

Other than a set of useful tools, there are a couple other things we could consider moving ➡️ towards. Only when we reach transparency, do we allow investors to make better decisions. And only when investors can make the best decisions, do they best allocate capital to drivers of growth and innovation in our society 🚀. By sharing approaches to encourage and guardrail change, can we make progress occur.

  • sharing ↔️📩 data between VC firms and investors
  • standardised term sheets
  • open source incubation and acceleration programmes 👍
  • pooling mentorships and resources 👩‍💼👨🏽‍💼
  • open source regulation re: new technologies and new business models 👨🏾‍⚖️

2020 - Books 📚 in Review

I've just finished my New Year's Resolution to read 40 books this year. These are some of the best - the must reads.


1. Meditations by Marcus Aurelius 🏛
A must read and tonic for the soul. Also really interesting how Stoicism converges with Taoism (I also enjoyed 道德经) and converges with new behavioural economics - many of the meditations really center around taming System 1 thinking with System 2 thinking, to let go of ego and find inner peace.

2. How Asia Works 🌏: Success and Failure in the World's Most Dynamic Region by Studwell
This is a must read for anyone vaguely interested in Asia, who wants an alternative to the Euro-American centric economic history we are taught. It also helps conceptualise the current stages of development of Asian economies (and institutions) in the context of their development - and a convincingly nuanced perspective of why comparing "Western" society to "Eastern" is like comparing apples to oranges.

3. World After Capital 🤑 -->👩🏻‍🎓 by Albert Wegner
This was the last book to make the 40 I wanted to read this year. It's just fantastic. It's a incredible analysis and prediction of how our fundamental ways of thinking and organizations (as a species) must evolve, and the consequences if it does not. Loved the Universal Basic Income (UBI) drive-by too, as I'd been meaning to read up on it.

4. The Psychology 🧠 of Human Misjudgement by Munger
A must read primer on human psychology and irrationality - all things in life boil down to human behaviour - either our own or others, so it is important to know why we are irrational, and control that, and also to see that behaviour in others, to warn them of it too.

24 Standard Causes of Human Misjudgment
1) Under-recognition of the power of what psychologists call “reinforcement” and economists call “incentives”
2) Simple psychological denial
3) Incentive-cause bias
4) Bias from consistency and commitment tendency
5) Bias from Pavlovian association
6) Bias from reciprocation tendency
7) Bias from over-influence by social proof
8) Better to be roughly right than precisely wrong
9) Bias from contrast-caused distortions of sensation, perception and cognition 10) Bias from over-influence by authority
11) Bias from deprival super-reaction syndrome
12) Bias from envy/jealousy
13) Bias from chemical dependency
14) Bias from mis-gambling compulsion
15) Bias from liking distortion
16) Bias from the non-mathematical nature of the human brain
17) Bias from over-influence by extra-vivid evidence
18) Mental confusion caused by information not arrayed in the mind
19) Other normal limitations of sensation, memory, cognition and knowledge
20) Stress-induced mental changes
21) Other common mental illnesses and declines
22) Mental and organizational confusion from say-something syndrome

5. The Snowball 🧊: Warren Buffett and the Business of Life by Schroeder
Warren Buffet's authorised biography. An incredible look in one of the best investors.

Memorable Quotes:

On humility

“When I was a kid,” Warren would later say, “I got all kinds of good things. I had the advantage of a home where people talked about interesting things, and I had intelligent parents and I went to decent schools. I don’t think I could have been raised with a better pair of parents. That was enormously important. I didn’t get money from my parents, and I really didn’t want it. But I was born at the right time and place. I won the ‘Ovarian Lottery.’”

.... “Whenever my version is different from somebody else’s, Alice, use the less flattering version.”

He dreaded falling prey to what a Harvard Law School classmate of his had called “the Shoe Button Complex.” [...] Cornering the market on shoe buttons made him an expert on everything. Warren and I have always sensed it would be a big mistake to behave that way.”

On taking risks and straying from the pack

It might seem easier to go through life as the echo—but only until the other guy plays a wrong note.

.. his Inner Scorecard—a toughness about financial decisions that had infused him for as long as anyone could remember—kept him from wavering.

On startups and tech (aka the famous Sun Valley Conference)

That was so even though Herbert Allen himself thought the “new paradigm” for valuing technology and media stocks—based on clicks and eyeballs and projections of far-off growth rather than a company’s ability to earn cold hard cash—was bunk. “New paradigm,” he sniffed. “It’s like new sex. There just isn’t any such thing.”

On Graham and investing

Warren felt there was a conflict of interest inherent in the business. [...] “You can’t make a living that way. The system pits your interests against your clients.”


6. The Seven Husbands of Evelyn Hugo 💃🏼 by Reid
Perhaps one of the best fiction books I've read in a while. I recommended it to my best friend who devoured it. It's a tour-de-force. A profoundly human book - about struggle and hustle, love and lust, shame and power. It's unintentionally a great study on human nature, with a romance for the ages.

Memorable Quotes:
On human nature

“It’s always been fascinating to me how things can be simultaneously true and false, how people can be good and bad all in one, how someone can love you in a way that is beautifully selfless while serving themselves ruthlessly.”

“No one is just a victim or a victor. Everyone is somewhere in between. People who go around casting themselves as one or the other are not only kidding themselves, but they’re also painfully unoriginal.”

On struggle

When you're given an opportunity to change your life, be ready to do whatever it takes to make it happen. The world doesn't give things, you take things.”

“ yourself a favor and learn to grab life by the balls, dear. Don’t be so tied up in trying to do the right thing when the smart thing is so painfully clear.”

On sexism

“Isn't it awfully convenient,” Harry added, “that when men make the rules, the one thing that's looked down on the most is the one thing that would bear them the greatest threat? Imagine if every single woman on the planet wanted something in exchange when she gave up her body. You'd all be ruling the place. An armed populace. Only men like me would stand a chance against you. And that's the last thing those assholes want, a world run by people like you and me.”

“You wonder what it must be like to be a man, to be so confident that the final say is yours.”

On love

“I love you so much, sweetheart. So, so much. And it's in part because of things like that. You're an idealist and a romantic, and you have a beautiful soul. And I wish the world was ready to be the way you see it. I wish that the rest of the people on earth with us were capable of living up to your expectations. But they aren't. The world is ugly, and no one wants to give anyone the benefit of the doubt about anything. When we lose our work and our reputations, when we lose our friends and, eventually, what money we have, we will be destitute. I've lived that life before. And I cannot let it happen to you. I will do whatever I can to prevent you from living that way. Do you hear me? I love you too much to let you live only for me.” For a moment, I thought she might flood the backyard. “I love you,” she said. “I love you, too,” I whispered into her ear. “I love you more than anything else in the entire world.” “It’s not wrong,” Celia said. “It shouldn’t be wrong, to love you. How can it be wrong?” “It’s not wrong, sweetheart. It’s not,” I said. “They’re wrong.”

“It is two A.M., and you are tired. You miss the love of your life. You want to go home. You would rather be with her, in bed, hearing the light buzz of her snoring, watching her sleep, than be here. [...] You imagine a world where the two of you can go out to dinner together on a Saturday night and no one thinks twice about it. It makes you want to cry, the simplicity of it, the smallness of it. You have worked so hard for a life so grand. And now all you want are the smallest freedoms. The daily peace of loving plainly.”

Special mention to other great books read just before 2020
1. 死神永生 (三體 ,#3)
3. The Intelligent Investor
3. The Emperor of All Maladies: A Biography of Cancer

Problem with Education 🏫 Today - Part 2

Verdict ⚖️ and disruption possibilities 🚀
Disruption is inevitable and welcome 🎉. The question is, what form it will take. Let's start with what existing disruption we see, and where we think it'll end up. I've structured the disruptions below chronologically ⏳, where possible. Many disruptions also are combinations of the below.

Old-school education archetype: groups of students taught by teachers, sitting standardised tests 👨‍🎓👨‍🎓👨‍🎓 👨🏻‍🏫

The below three trends really all come down to improving student's performance in the education system. If traditional schooling is a road between birth and employment, then these interventions are wheels, so students can move faster and easier 🛹.

Existing trend #1: School-performance enhancing solutions

  • The most obvious disruption has been old-school supplements. This has obviously come in the form of tutors 🎓, revision books and notes 📚, fuelled by parents who are willing and able to pay people for more.

Existing trend #2: Moving the above online 💻

  • Obviously, though incredibly slowly, we have make marketplaces or online versions of all of the above.
  • Standouts include VIPKID and Byju in the online tutorial space, as well as Quizlet in the educational study tools space

Existing trend #3: Personalisation and data/AI driven education 👩🏻‍💻👨‍🎓👨‍🎓👨‍🎓

  • Increasingly, we are seeing new online tools to personalise, and thus make more effective, learning. For example, we see spaced repetition flashcards, or learning visualisation tools.
  • Standouts include Anki.

The first paradigm shift is, moving outside the demands of of traditional school. Increasingly people see that extra-curricular activities as useful. If traditional schooling is a road between birth and employment, then these interventions may be seen as shortcuts or side roads branching off from the main road. 🛣

Existing trend #4: Extra-curricular learning (+ online) 🎹

  • In more developed economies, especially in Asian cultures that prize education, we are increasingly seeing extra-curricular classes, e.g. music lessons. These usually teach a "hard skill" like being able to play music, often done due to a) traditional values, many English style boarding schools that still rank highly stress this b) transferrable skills, or c) university admissions boosting effects.
  • Within this trend, we're now seeing a move and focus on "soft skills", e.g. critical thinking skills. People are now increasingly comfortable with skills that may not be immediately measured or shown, due to a) increased need in new economy, b) demands of international facing schooling, e.g. IB or AP style that lead to better university admission outcomes.
  • Standouts include tuition centers, e.g. Capstone in HK.

*The second paradigm shift is, replacing the demands of of traditional school. If traditional schooling is the road most travelled between birth and employment, then these solutions offer an alternative path entirely. 🛤 *

Existing trend #5: Technology Bootcamp 🏕

  • Already near and around Silicon Valley, coding bootcamps arise for people who want to pursue jobs in technology. These essentially acts as a replacement for traditional education.
  • Standouts include Flatiron School.

Existing trend #6: Alternative schools 🚀

  • Additionally, interestingly, new schools aim to replace traditional pre-college education, with their own curriculums/teaching formats.
  • Standouts include Sora, which allows for student directed learning. Standardised curriculum content is broken down, and reassembled into projects and areas of reasoning the student cares about.

The third paradigm shift is towards a lifelong schooling. If traditional schooling is a road most travelled between birth and employment, then these solutions are similar to a discovery the road never ends. 🚦

Existing trend #7: Lifelong up-skilling 🧓🏻

  • Increasingly, we are moving away from a one-company, rank-climbing career arc, to multiple life changes during this new stage of technological innovation and change.
  • People in my/the new generation are expected to see automation/technology, creation of new industries, fluid working, globalisation and dramatic career switches. Traditional education is not preparing people for this.
  • Standouts include Blinkist, or MOOCs.

The fourth paradigm shift is outside of a strict binary of "learning" vs "not". If traditional schooling is a road most travelled between birth and employment, then these solutions represent the realisation that learning is not just a road, but also the trees that line the path 🌴🛤🌴.

Existing trend #8: Unbundling the campus 🥳

  • Increasingly interventions acknowledge how education is more than book/school/tuition learning. We see education comes from others - mentorship and alumni, our peers or on projects.
  • Special mention to Handshake and Facebook Campus.

Being at the leading edge of a field doesn’t mean you have to be one of the people pushing it forward. You can also be at the leading edge as a user. It was not so much because he was a programmer that Facebook seemed a good idea to Mark Zuckerberg as because he used computers so much. - Paul Graham

The next big disruptions, are likely to be a combination of the above that is packaged in a way pleasing to the consumer. Otherwise, they may be ones that facilitate the move award from credentialism to the real world. Moreover, as we start to blend learning and other areas of life, new and successful disruptions will navigate this expanding boundary successfully to best serve users.
That really is the key - whichever startup best serves their users, i.e. parents, will really win. And they have a lot of subconscious, unmet needs.


A VC recently described to me how he sees EdTech progressing. Interestingly, it sort of coincides with the "shifts" I described above, only he suggests three waves instead of four shifts. The three waves are 1) dumping content online, e.g. Khan Academy and YouTube, 2) online marketplaces, e.g. VIPKID and 3) now the shift towards community based EdTech. Sweet. 🤘🏻

The Problem with Education 🏫 Today

Problem 1: Education 👨🏻‍🏫 ≠ learning 📚

  • Our Miseducation - A daring assumption to make at 22, but nonetheless one I’m happy to commit to currently. The problem I will introduce is a Gordian knot, but let’s tease out some issues.
  • Let’s start from the beginning, what is the point of education. I vaguely think the point is to prepare children for (their, our, your) future. The problem is, our education doesn’t do any of this. 🤷🏻‍♀️
  • We teach things that are useless in real life. Of course, an argument can be made that these teach us transferrable skills, but then why don’t we teach the real skill itself?

We don’t teach our kids the important things.
*We don’t teach psychology 🧠, i.e. we don’t teach the real world. Our curriculum is to real life, what classical economics is to behavioural economics. We don’t teach children how generally things work in the real world, but teach them how thinks precisely should work in the world.
*Heuristics of the mind - perhaps the most important thing we should teach people is how our brain sabotages us, and how we should avoid that to think clearly.

Even worse, we teach our kids the wrong things. 🙅🏻‍♀️

  • In schools, we spoon-feed people the “right” and “wrong” answer, without teaching them how to find and solve problems.🥄
  • We use only one yard stick 📏 to measure students, instead of letting kids develop what they’re good at.
  • Indeed, (in Hong Kong) we punish creativity, and initiative. Indeed, in local schools, we punish people for being attacked by other students. We inoculate people against innovation, which cannot be the right thing.
  • We mistake elite education and memorisation with real learning and understanding.

Problem 2: One size fits all

  • Our current teaching systems are one size fits all. There is little tailoring of content, attention and testing.

Problem 3: Lack of innovation in education
*All this, of course, boils down to a lack of innovation in schools. We just adopt past models of schooling, without thinking about - psychology, technology or what the future will look like.

  • Some must consider research:
    • Bloom's Two-Sigma problem 📈
    • Memory hacking, e.g. spaced repetition
    • The Pygmalion Effect

For Verdict ⚖️ and disruption possibilities 🚀, see part two

5G 📶: misunderstood and overrated

5G’s promise 🎉

  • Faster connection - near real time
  • Reduces latency, allowing more data to be delivered and processed faster 💨
  • Increased usability: connecting Internet of Things or IoT
  • This would allow us to harness connectivity in ways yet unseen, e.g. remote surgery 🤯

But wait, whoops...🧐

  • Infrastructure 🏗- still requires fibre optic connection - sorry folks, you may not get this if you don’t live in New York🌇 (+ inequality, bleh)
  • Shorter wavelengths + Blocked by glass - scalability block again 🛑

Verdict ⚖️

Make no mistake, 5G is paradigm shifting. 4G made possible all the fun stuff like Uber we take for granted today with better and reliable connections. 🚕 5G represents a similar revolution.


...overhyped re consumer

  • TL;DR, 5G is a) too expensive 💰 and b) results are not worth it for the ordinary consumer.
  • Infrastructure demands and technical shortcomings limits its meaningful use for the ordinary consumer
  • The uses possible also are overrated. 5G can only bring limited fire now.🔥
  • E.g. real time connectivity is already pretty great. One of the problems with 4G is that it makes a lot of things possible, like Uber and on demand services. Sure, we can have 4K videos on the go. (Definitely not - if Quibi is any indication consumers don’t like high quality entertainment on the go. ) 📺
  • Furthermore, not everything needs to be connected to the internet. (And yes, I’m talking about some of the past startups I’ve seen). 😬

So what now?

  • Whilst the consumer angle is less promising than we’ve made it out to be, the real money might lie in the play for B2C. ⚽️
  • TL;DR they can afford it, and it will be worth it for large corporates.
  • Companies/large institutions are ideal use cases for 5G. They have the ability to pay for, design and implement the infrastructure for 5G happen. And it can pay off. 💰
  • For example, 5G can be configured for controlled environments like in factories, for construction/manufacturing. It can also yield significant effects - communication between devices/machinery may bring increase efficiency. 🏭

There are also a lot of interesting points that lie in periphery. 🧐

  • Let’s talk about edge computing. By deploying edge computing in networks, the distance between the endpoint and the data center is reduced. In the case of 5G specifically, which already reduces the latency between the data center and the endpoint, an edge compute reduces the actual physical distance between these two destinations.
  • Edge computing is interesting because though they’re often thought of in connection with 5G, thought it doesn’t have to be the case. Thus, these under-appreciated cousin of 5G deserve more love for their wide applicability. ❤️
  • Moreover, this leads me to the most unloved cousin of 5G: the 5G/data-center/periphery technology REITS.
  • Land is necessary for 5G to work well.
  • New infrastructure for 5G + blocks to scalability requiring more physical deployments + edge computing = REITs cash in 💰🏘

Credit and Debit Card Processing in the US 💳

Process: Issuing bank 🏦 —> customer 💵 hands information to payment gateway 🚪—> payment processor 🏭 handle the rest of the transaction details —> card network 🌍 authorises credit —> acquiring bank 🏦 pays merchant 💼

How it works?

  • Many middlemen between customer and merchant charges fee = merchants lose ~ $3 in a typical $100 transaction 💸

Disruption possibilities 🚀

  • Bypass system entirely, e.g. Kenya’s M-Pesa and China’s Alipay
  • Reduce middlemen/edit existing system, e.g. Apple Card
  • Re-imagine the economic distribution, e.g. pay back consumer based on fees received
  • New credit cards – e.g. for children
  • Corporatisation of credit cards/Payments or Fintech as a service - e.g. Railsbank
  • Efficiency generating/adjacent technologies, e.g. measuring customer sentiment/optimising conversion etc.
  • Cross-border disruption 🌍

Verdict ⚖️

Disruption is inevitable. 🎉

  • Obvious and large pain points with tech solutions

Key is: what form the disruption will take? 🔑
Element #1: What are we replacing?

  • Historically, the credit/debit card system in the US replaced cash.
  • Vs In China, they replaced cash with e-wallets.
  • What disrupts the credit/debit system will replace credit cards in a settled and comfortable economy
  • This replacement will need to serve the same functions and provide the same perks with less cost.
  • Whilst some of this might come from reducing middlemen, we will still need to eventually make cuts to perks, as they (and the entire system) are funded by middlemen fees.
  • = New revenue streams will need to be found.
  • 💡 However, can be noted that credit/debit is more obsolete, and so doesn’t need to be replaced?
  • On-demand age moves us further away from need of credit for unnecessary costs. FinTech solutions turn large expenses into smaller, more regular transactions.
  • Additionally, credit/debit cards provide one function, whereas we’re moving towards conglomeration and the “everything” app.

Element #2: How are we replacing? 🏛

  • Technical and governmental infrastructure required for current debit/credit card system means any replacement has to deal with that entrenched interest against changing.
  • It also means that any replacement that doesn’t re-use the existing system (in novel ways), short of bypassing it, will be limited in ability to disrupt it.
  • That said, to bypass it entirely, infrastructure is required that may not exist in the West. China has ID cards to track all citizens, India’s government has a banking tech stack that provides infrastructure. 🏗

Element #3: Trust enabling? 🤝

  • Consumer behaviours is not to be understated. Consider HK 🇭🇰 vs Shenzhen 🇨🇳 – one is a near cashless society, whilst the other, despite being the same country and having similar resources to expansion is still very much a cash society. Reasons include a comfortable populace with little reason to ‘risk’ moving onto a new system.

Element #4: Market dynamics

  • Incumbents will play a role. Increasing fragmentation, e.g. penetration into China is unlikely.
  • 🤯Interestingly, this suggests different middlemen, but that they will not be as reduced as we make them out to be.
  • Politics will also continue to play a role – payments are still international, and tensions are still high. 🥵 But note that China has a strong hold on Asia and Africa.

🎯 Key Question: what is the credit/debit card to the target consumer of the start-up? What is the solution?

Denise 🤘

Future of Healthcare 👩‍⚕️: thoughts from an unqualified humanities major

There are too many problems with the healthcare system to note. 😢 We’re skipping to the verdict.

Verdict ⚖️

It’s inevitable that disruption is coming. Several headwinds inform this: increasing dissatisfaction, more ‘on-demand’ technology, greater data processing. The key is what disruptions we are likely to see.

Non-educated guesstimate trend #1: from binaries ⚖️ to sliding scales ↔️
Historically, health has existed in a binary. One is either “well” or sick, receiving care or not, a patient or not. The future we’re likely to see is a realisation that this is a false dichotomy, and thus a move to a more nuanced scale of wellness.

Driving factors:

  • on-demand economy and increasing touch points with the end user 📱
  • IoT and new medical devices

The above proposition seems simple, but it has a profound implication on the type of care we will see in the future.
* We will see more, earlier, and smaller interventions 👩‍⚕️📱
* Instead of waiting for a ‘hospital’ or ‘doctor visit’ event for a medical intervention, we can achieve smaller interventions with any number of new touch points.
* We have everything from the run of the mill telemedicine/online therapy/mindfulness sites, i.e. taking existing services and breaking them down to a granular level that can be more easily accessed.

* More interesting is the idea of a greater scope of primary care, i.e. in addition to digitising what medical care we have now, we add to that other services, e.g. mental health, addiction management.
* Taking the above to the extreme, instead of doctor’s offices, specialist clinics, and hospitals we should fully expect to see an increase of offerings in-between, e.g. micro-hospitals and clinic-hybrids. 🏥
* As primary care moves out of hospitals and online, we should see hospitals and new institutions evolve, some becoming smaller in size and more specialised.
* The uber-isation of patient importance (yes, ugh, I know but it’s an accurate description) 🙋‍♀️
* We will see an increased emphasis on patient experience as a metric. Direct to consumer innovations, that don’t go into the belly of the beast will need to appeal to our psychology to drive adoption. We’ve gotten used to being able to complain about quality of service.
* Moreover, other ancillary services will make use of data, consolidation and network effects to take on more services than before. In addition to bringing the service to the ‘patient’, services will appear in front of the ‘patient’. For example, a point of contact like the pharmacy will evolve past dispensary: they may also help validate identity, eligibility, prescribe, and even provide some of the smaller health interventions discussed above.
* 🔑 In short, we push smaller and more interventions closer to the “patient”!
* + On a more macro level, we may catch pandemics earlier.

Non-educated guesstimate trend #2: from sliding scales into all parts of life 🌏
* Taking the above to the logical conclusion, if there is no states of healthy/unhealthy, wellness and care bleeds into other areas of life.

* That may take the form of something as simple as continuous monitoring of health with the Apple watch.
* More interestingly, it may see health as not just being a minimum acceptable form of being, but as a component of productivity. This is an especially interesting spin on employer-driven healthcare in the US, where it is not seen as a liability or a necessary perk, but merely a metric to boost productivity.
* One great upshot of this is the destigmatisation of care to mental health.

Non-educated guesstimate trend #3: away from hypothesis driven care 🕵️‍♀️ to personalisation of care 🎯

  • Big data and AI deserves will be huge.
  • “When you hear hoofbeats think horses not zebras”🦓 : medicine is partially an exercise in hypothesis testing, i.e. we see symptoms, we guess it is most likely this and we go down in order of probability
  • With big data and AI, e.g. genomic testing, we can move from that to personalised care and treatment. This has a lot of exciting implications, including scalability, accuracy, and quality of care. I am especially excited about it’s effect on mis- or missed diagnoses surrounding rare chronic illnesses.
  • More interestingly, we may see big data and AI driving predictions of issues, feeding trend #1, above.

Non-educated guesstimate trend #4: blockchain and data management 📊

  • With increased scaling of data (and hopefully data privacy structures that work), disruptions to help with patient data and administration is inevitable, e.g. blockchain

Non-educated guesstimate trend #5: human-centric design 👩‍🎨

  • UX/UI design of medical devices. Gone are the day of, let’s make it beep. Now, companies like MIC are designing feedback systems to interface with hospitals effectively, moving closer to the usual domain of Silicon Valley startups.

Non-educated guesstimate trend #6: cyber-security in healthcare - an under-loved sector 🥋

  • I was recently surprised to discover how often cyber-security concerns play into healthcare, e.g. ransomware attacks with a woman in Germany being the first death to be linked directly to a ransomware attack at Duesseldorf University Hospital.
  • This problem has an obvious cause: legacy existing IT infrastructure old and vulnerable. This can only grow worse with IoT in healthcare.
  • Apart from the obvious need for cyber-security, ransomware and data hostage taking will interact with new insurance/FinTech in healthcare, which is it’s own other beast entirely.

PS another key implication of more data heavy health care, is re: governmental structure underpinning it: the US will likely have different problems and evolutions in technology to China. 🏛