Note to self: What I read about startups

DISCLAIMER: This text is plagiarized.

Definitions:

Startups that scale prematurely are classified as inconsistent and startups that scale properly are classified as consistent

Types of startups:

  • The Automator / Type 1: These startups are product centric with a self-service customer acquisition strategy, that focus on quick execution and often automate a manual process. The majority of them target consumers in existing markets.
    • Examples: Google, Dropbox, Eventbrite, Slideshare, Mint, Pandora, Kickstarter, Zynga, Playdom, Box.net, Basecamp, Kayak
  • The Social Transformer / Type 1N: These startups have a self-service customer acquisition strategy and often create new ways for people to interact. They are almost always confronted with the challenge of reaching critical mass. If they surpass this threshold they can often have runaway user growth in a winner-take-all market.
    • Examples: eBay, OkCupid, Skype, Airbnb, Craigslist, Etsy, IMVU, Flickr,
      LinkedIn, Yelp, Facebook, Twitter, Foursquare, YouTube, Mechanical Turk,
      PayPal, Quora
  • The Integrator / Type 2: These companies thrive on acquiring customers by generating leads with marketing and closing them with inside sales reps. They are product-centric and rely on early monetization typically through subscriptions in smaller markets. They often take innovations from Automator startups and rebuild it for smaller enterprises.
    • Examples: Intuit, Square, Adobe, PBworks, Uservoice, Mixpanel, Dimdim,
      HubSpot, Marketo, Xignite, Zendesk, GetSatisfaction, Flowtown
  • The Challenger / Type 3: These startups are focused on closing high paying customers in large but fragmented markets. They are highly dependent on a small number of deals being successful and usually operate in complex and rigid markets. To be successful they need to find a repeatable and scalable sales process.
    • Examples: Oracle, Salesforce, MySQL, Red Hat, Jive, Atlassian, Palantir,
      NetSuite, WorkDay, Zuora, Cloudera, SuccessFactor, Yammer

Lessons:

  • Startups that try to scale before they have reached product/market fit and streamlined their customer acquisition process don’t do very well. In fact, no inconsistent startup was able to get more than 100,000 users.
  • Inconsistent startups grow faster in the early stages, probably due to forcing or over-engineering growth and then relatively flatline by the scale stage. Meanwhile, consistent startups have slow growth in the beginning and take off in the scale stage in “hockey stick” fashion (a linear line on a log scale is a smooth exponential curve). By the scale stage, consistent startups are growing more than 20 times faster than inconsistent startups per month.
  • Inconsistent startups almost never reach a strong monthly run rate.
  • Startups that are inconsistent depict themselves as better than they are. The illusion holds up until the scale stage when consistent startups have a valuation that skyrockets and inconsistent startups have a down round.
  • Startups that scale prematurely has teams that are significantly larger than the consistent startups that haven’t scaled yet, but their team size also rarely gets as large as startups that do scale properly since they aren’t able to sustain their growth. Scaling the team prematurely is a problem because it’s very hard to align a large team if there are still frequent changes in direction of the company, as is usually the case in the early stages.
  • Raising too much money too early can be harmful to startups. It puts a company under pressure to scale even though they are not ready.
  • If startups raise too much money before the scale stage they have a very high chance of being inconsistent. Founders may think they can be disciplined if they raise too much money, but the numbers show that by and large they aren’t.
  • Before a startup can cost-effectively acquire customers they should not be spending a lot of money on customer acquisition. If you consider spending more than $15,000 on customer acquisition before you are ready to scale, then the data shows that Inconsistent startups are 2.3x times more likely to spend too much on customer acquisition.
  • All startups overestimate their valuation during discovery and then it drops once they actually start validating their product. Inconsistent companies have rose-colored glasses that are significantly darker.
  • Inconsistent startups over-engineer their products and spend too much time on building out features that are not absolutely necessary. We often see engineers that are entirely convinced that their product can only work if it has the same product complexity of a mature product such as Facebook or twitter. Most of the time this complexity leads to lower market adoption and eventually failure for the startup.
  • In the beginning, startups can get easily lost in building a product without validating the actual demand for it. Based on interviews most inconsistent startups are under the impression that they are an exception to the rule. They believe they have found a special insight for a disruptive startup that no one else has. Unfortunately, most of these startups fail.
  • Consistent startups spend more time discovering who their customers are, whereas inconsistent startups are focused on validating that customers want their product. Consistent startups are searching. Inconsistent startups are executing. It’s widely believed amongst startup thought leaders, those successful startups succeed because they are good searchers and failed startups achieve failure by efficiently executing the irrelevant.
  • Especially in an early stage, it is dangerous to outsource the product development. All dimensions: product, customers, business model, financials, and the team are typically changing at a high rate. If product development is not done in-house startups will have a hard time keeping up with the daily or even hourly feedback loop startups have at this stage.
  • Startups that are tackling new markets are more likely to be inconsistent because they have more uncertainty than existing markets. Tackling an existing a market where the product is differentiated by being cheaper has the highest certainty because people almost always prefer the same value at a lower price, whereas differentiation by better or niche rely on more subjective qualities.
  • Trying too hard to monetize leads to inconsistency. While money can be an important validation indicator, stressing it too heavily will lead startups to ignore opportunities and drift towards non-scalable opportunities that are likely to turn into small business or custom consultant shops.
  • The level of ‘difficulty’ or ‘uncertainty’ for the different types of startups is in the following order: Type 2, Type 1, Type 1N, Type 3 (see above for the description of types).
  • That consistency does not vary by estimated market size. There is an equal distribution of consistent and inconsistent startups across all the different market sizes.
  • The frequency of your product release cycle has no effect on consistency.
  • The education of the founder has no impact on whether a startup is consistent or inconsistent.
  • The gender of the founder has no impact on whether a startup is consistent or inconsistent.
  • The time the founders have known each other has no impact or little impact on whether a startup is consistent or inconsistent.
  • Age has no impact on whether you are consistent.
  • The amount of products a company is handling, therefore, does not influence the performance
  • Google Analytics, homegrown solutions, and spreadsheets are the top 3 three tools that are used by more than 90% of all startups to track their metrics and make decisions.
  • Geography has no impact on whether you are consistent

Other findings:

  1. Founders that learn are more successful: Startups that have helpful mentors, track metrics effectively, and learn from startup thought leaders raise 7x more money and have 3.5x better user growth.
  2. Startups that pivot once or twice times raise 2.5x more money, have 3.6x better user growth, and are 52% less likely to scale prematurely than startups that pivot more than 2 times or not at all.
  3. Many investors invest 2-3x more capital than necessary in startups that haven’t reached problem solution fit yet. They also over-invest in solo founders and founding teams without technical cofounders despite indicators that show that these teams have a much lower probability of success.
  4. Investors who provide hands-on help have little or no effect on the company’s operational performance. But the right mentors significantly influence a company’s performance and ability to raise money. (However, this does not mean that investors don’t have a significant effect on valuations and M&A)
  5. Solo founders take 3.6x longer to reach scale stage compared to a founding team of 2 and they are 2.3x less likely to pivot.
  6. Business-heavy founding teams are 6.2x more likely to successfully scale with sales were driven startups than with product centric startups.
  7. Technical-heavy founding teams are 3.3x more likely to successfully scale with product-centric startups with no network effects than with product centric startups that have network effects.
  8. Balanced teams with one technical founder and one business founder raise 30% more money, have 2.9x more user growth and are 19% less likely to scale prematurely than technical or business-heavy founding teams.
  9. Most successful founders are driven by impact rather than experience or money.
  10. Founders overestimate the value of IP before product market fit by 255%.
  11. Startups need 2-3 times longer to validate their market than most founders expect.This underestimation creates the pressure to scale prematurely.
  12. Startups that haven’t raised money over-estimate their market size by 100x and often misinterpret their market as new.
  13. Premature scaling is the most common reason for startups to perform worse. They tend to lose the battle early on by getting ahead of themselves.
  14. B2C vs. B2B is not a meaningful segmentation of Internet startups anymore because the Internet has changed the rules of business. We found 4 different major groups of startups that all have very different behavior regarding customer acquisition, time, product, market, and team.

Yoneda Lemma in terms of burritos

Suppose you find a machine on your doorstep called the Burritron that makes any kind of single filling burritos you want. It needs nothing as an input except another machine called the FoodinatorFoodinator can transform a stone into a type of filling you need for your burrito. Each type of foodinator can only convert stone into one type of filling. For example, the Bean Foodinator can transform a stone into beans. Meat Foodinator can transform a stone into cooked meat. A Do Nothing Foodinator does nothing to the stone and just gives it back as it is.

You want to figure out how the Burritron works, so you get a Bean Foodinator that can convert a stone into beans, and plug it into the Burritron. It pops out a burrito with just red bean filling in them.

You still want to figure out how the Burritron works, so you get a Do Nothing Foodinator that does nothing to stone and plugs it into the Burritron. It pops out a burrito with a stone in it! A stone burrito.

So here is your reasoning about how the Burritron works:

  1. Within the Burritron there must be a stone burrito otherwise plugging in the Do Nothing Foodinator cannot make the Burritron produce a stone burrito.
  2. The Burritron must use the Foodinator to make the fillings because that is the only way it can convert a stone into the filling we need.

Thus even though you did not know how Burritron is made you now know enough to reverse engineer the Burritron.

More importantly, you have now “understood” a stone burrito in terms of normal burritos. You now just need to figure out how each of foodinators do their job and you are all set.

Got Purpose

Since I do not plan to get married and/or have children, I’ve always lacked something to consider more important than myself that actually exists and is not just an idea.

I have decided that the happiness and well-being of people like me are something worth considering as something more important than myself.

They will exist as evolutionary cul-de-sacs but they deserve well-being because they did not mean any harm.

People like me are happiest when they are not bullied, ridiculed or confined by any group and are left alone to voluntarily interact with people they wish to trade with.

To make them the happiest people on this planet I will have to make their tyrants weak.

Signalling

[Most of this stuff is plagiarized]

Example 1: Catholic devotion

Suppose you’re a Catholic who wishes to signal your commitment to Catholicism for whatever reason (maybe stronger commitment gets you more respect). One thing you could do is refrain from murder since this is an important Catholic doctrine. However, this is an extremely weak signal. Why? Because the costs of refraining from murder are very low, whether you are a Catholic or not. If, however, you refrain from using a condom, that is a strong signal of Catholicism. Why? Because those with weak commitment or no commitment to Catholicism, and so do not believe so strongly in the immorality of condoms, regard the net costs from refraining from using a condom as significantly higher.

Theory

This is called signalling.

Signaling is defined as “a method of conveying information among not-necessarily-trustworthy parties by performing an action which is more likely or less costly if the information is true than if it is not true”. Some signaling is performed exclusively to impress others (to improve your status), and in some cases isn’t even worth that. In other cases, signaling is a side-effect of an otherwise useful activity.

To construct a theory based on signalling you need two key ingredients:

  1. an unobservable trait that, if observable, would be rewarded
  2. an observable action (or set of actions) which is expected to be less costly the more of the unobservable desirable trait you possess

That is, the trait to be signalled (1) and the signal itself (2).

Example 2: Social Justice

Another example is social justice advocacy. The unobservable trait in question is fairly obvious: the strength of devotion to social justice causes. Anyone who observes mainstream culture today understands that there is a social premium on being perceived to be supportive of social justice. To be more in favour of social justice is to be more moral and more respectable in today’s Western society, to obtain a higher position in an informal moral hierarchy. Even if many people dislike social justice advocates, if the advocates themselves regard their strength of belief as a virtue, they will try to signal it. Simply saying that you’re committed to leftist causes counts for nothing because almost anyone can do that—so signalling is required. That’s ingredient (1).

Ingredient (2) is less obvious. I believe the signal is simply the positions social justice advocates take. It seems plausible to me that many things that social justice advocates believe many people would find unpleasant to believe. Examples of this are the idea that the riots in Baltimore and Ferguson were a good thing and that whites should pay reparations to nonwhites.

Take the case of police shootings of black Americans, which has received a lot of media attention recently. If “social justice” were the goal of social justice activists, you’d expect them to focus their energies on cases where the evidence against the police is strongest. However, the signalling theory predicts the opposite. The stronger the evidence against the police, the more likely someone weakly committed to social justice (or not committed to social justice at all) will side against the police. Therefore the signal being sent is weakest in those cases. The signal is in fact strongest when the evidence against the police is weakest.

A case study in this is to compare the Michael Brown shooting with the death of Eric Garner. In the case of Michael Brown, all the evidence available supported officer Darren Wilson’s story. The Michael Brown case caused riots and received much longer and more intense coverage than the Garner case.

More generally, if “black lives matter” was really about protecting black lives, you’d think they’d be concerned that more black Americans are killed due to sneakers alone than are killed by police. You’d also think they’d be concerned about the so-called Ferguson effect. However, from a signalling perspective, both of these blind spots make perfect sense. Everyone can see that deaths over sneakers and another black-on-black crime is bad, and it’s precisely for this reason that it’s ignored—it’s a weak signal.

Example 3: Cults

My mother is in a cult which was recently banned by the Catholic Church, and so I have experienced first hand the effects of signalling. Most cults are based on an unobservable virtue like faith. There are however observable actions which cannot be undertaken with ease if you are a non-believer or a moderate (e.g. mind-numbing chants). Cults through means like ostracization, shame and fear punish the moderates who do not exhibit signalling behaviors. The cults also bestow status and respect on those who do signal more. The combination of punishment and reward for doing difficult deeds signalling unobservable virtues can be used to mobilize people in any direction.

Example 4: Quackery

Fraudulent medical practitioners often make treatment efficacy a consequence of commitment in the treatment. They also offer observable actions one can engage in to signal your commitment. These observables are often very difficult or cause unease to someone who is not committed. So in order to signal commitment, the patient engages in very difficult activities. Some have even died seeking such a status of a committed patient to a quack.

My state of mind

I have been part of these groups that engage in arduous observable signalling as a proof of an unobservable virtue. My experience was that I was never bestowed with respect in such groups even when I signalled correctly. So I stay in groups with observable merit e.g. programming.

In fact, I am so averse to unobservable virtues than I even stay away from traditionally stable systems like marriage which are no longer based on observable economic necessity, but based on one’s unobservable commitment to virtues like tradition, romance and love which require observable signalling through pretentious gestures.

What can not propaganda do?

I think we need to find that out if we have to win against the currently establish order. The recent Brexit plebiscite is a good case study regarding this. We need to study, how despite all the media, the thought leaders, and the elites unanimously preaching that Brexit will be the end of the world, somehow the British people found the guts not to blindly obey them out of fear.

Seems like fear gets too boring after a while. Also when everything is rape, racism, and homophobia; when everybody is Hitler or a white supremacist, all the name calling simply doesn’t make any sense.