Posted on

Insights on managing Big Data from meet up with Dean and Ved

From Dean (Reputation.com)

  • Enterprise sales as an acquisition strategy is feasible because revenue per account ranges in the USD millions – e.g. 70 million USD
  • Once an auto company like Ford or GM signs up, they will start bringing their dealerships in
  • The infrastructure needs to be able to support the size of the data which can be up to billions of rows
  • Scaling of infrastructure to handle load ever increasing data becomes critical for the continued growth of the data company
  • Data Product will appear broken when user attempts generate report while the data is still being written into the database
  • The key challenge is that different solution is suitable for different operation
  • Types of data operation include
    • writing into the database
    • reading from the database
    • map reduce to generate custom view for data in the database to support different types of reporting for different departments in the client companies.
  • Successful data companies will create different layers of data management solutions to cater to the different data needs
    • MongoDB
      • good for storing relatively unstructured data
      • querying is slow
      • writing is slow
      • good for performing map reduce
    • Elastic Search
      • good for custom querying for data
  • Dev ops become a very important role
    • migration of data between different systems can extend up to weeks before completion
    • bad map-reduce query in codes while start causing bottlenecks in reading and writing causing the data product to fail
    • dev ops familiar with infrastructure might on occasion have to flush out all queries to reset
    • The key challenge is the inability to find bandwidth for flushing out bad queries within the codebase
  • Mistakes in hindsight
    • In hindsight lumping all the data from different companies into the same index on MongoDB does not scale very well
    • Might make better sense to create separate database clusters for different clients
  • Day to day operations
    • Hired a very large 100 strong Web Scraping company in India to make sure web-scrapers for customer reviews are constantly up
    • Clients occasionally will provide data which internal engineer (Austin) will need to look through before importing into relevant database
  • Need to increase revenue volume to gear up for IPO
  • The Catholic church has 10 times more money than Apple and owns a lot of health care companies.

From Dan (Dharma.AI), the classmate of Ved

  • Currently has 15 customers for their company
  • Customers prefer using their solution versus open source software because they can scale the volume of data to be digested and solution comes with SLA
  • Company provides web, mobile and table solutions which client companies’ staff can use in the field to collect demographic and research data in developing countries
  • The key challenge is balancing between building features for the platform and building features specific verticals:
    • Fields differ between industry: fields in the survey document for healthcare company will be very different for fields in the survey document for an auto company
    • Fields differ between across company size: survey format for one company might be different as compared to another in the same industry but of different size
    • Interface required is differs between companies
  • Original CEO has been forced to leave the company, new CEO was hired by PE firm to increase revenue volume to gear up for IPO

From Ved

  • As number of layers increase in the hierarchy, it becomes increasingly challenging for management to keep up to date on the actual situation in the market
  • New entrant of large establish competitor might sometime serve as an opportunity to ride the wave
  • when Google decided to repackage Google Docs for Education, it was a perfect opportunity for Edmodo to more tightly integrate into Google and ride that trend rather than being left behind
  • Failure to ride the wave will result in significant loss of market shares
  • It takes a lot of discipline to decide on just focusing on the core use case and constantly double down on it.
  • Knowing that a critical problem, which could potentially kill the company, exists versus successfully convincing everyone in the company that it is important to address it are two different things.
Posted on

Book summary: Your First 100

Your first 100

New Customer versus existing ones

  • New customer is 6 – 7 times harder to acquire than to keep one
  • probability of selling to new prospect 5-20% versus 60-70% existing customer
  • loyal customer worth 10X their first purchase

Customer and feelings

  • Loyal customer are buying into an experience they know they will get from your brand.
  • Profiles
    • readers
    • subscribers
    • buyers

Brand

  • What people say when you are not in the room, Jeff Bezos
  • It is tied to their sensory cues
  • Brand strategy is where you tell people how you want your brand to be perceived by your presentation of the different components
  • Your brand should be very targeted and not sell to everybody, only to those who will benefit the most from your offering. That is the basis of differentiation

Experience – components in a brand

  • product
  • website
  • delivery
  • customer service
  • payment

Touch points

  • Pre-touch point: someone mentioned you to them
  • premier first touch point: landing page
  • pivotal touch point: reading up more about your product on your website
    • category 1: no idea about the problem you are solving
    • category 2: aware of the problem
    • category 3: aware of problem, trust and love your content
    • category 4: ready to buy but have questions
    • category 5: made a purchase
  • prime touch point: onboarding & magic moment
  • post touch point: getting referrals

Content tilt

  • To focus on going deep and narrow in the content you create

Prospect behavior

  • 95% not ready but
  • 70% will eventually buy from you or your competitor
  • Only 3% are actively buying at anytime.

Related readings

  • What customers crave, Nicholas J Webb
  • Top of mind, John Hall
  • Lead generation for complex sale, Brain Carroll
  • Content Inc, Joe Pulizzi
  • Sticky Branding, Jeremy Miller
Posted on

Coaching for founders

  • As a CEO, being in the trenches feels a lot like being punched in the face over and over again
  • A lot of dysfunctions in teams boils down to not being clear about roles and responsibilities
  • always talk about the task and not about the person
  • The (blame of the screw up) Buck stops with the CEO whose role and responsibility includes being the ultimate decision maker
  • It is the CEO’s job to bring up difficult conversations that are necessary
  • performance issues boil down to two types motivational, perceived or ability
    • motivation because feeling burned out
    • perceived because role n responsibility not clear
    • ability then not supposed to clear  skillset mismatch
Posted on

Book summary: Thinking in Bets

Thinking in Bets
Decision Theory Model

Overview

  • Good quality decisions do not always yield good outcomes
  • All decision makings in real life are made under uncertainty. All decisions are essentially bets about the future
  • Decisions made in Chess are not made under uncertainty because every single permutation can be pre-computed unlike Poker.
  • Most real life decisions are not zero-sum games

On outcomes

  • Real life outcomes are probabilistic
  • Outcomes are influenced primarily by the quality of our decision (skill) and luck
  • While the outcomes might not always be positive, having a process in place to constantly improve the quality of decision making will tilt the odds in our favor

Implications

  • Do not change strategy drastically just because a few hands did not turn out well in the short run
  • For each premise understand what the base rate is
  • Learn to be at peace with not knowing
  • Recognize the limits of our own knowledge
  • A great decision is a result of a good process. A good process attempts to accurately represent our own state of knowledge
  • Watching: It is free to learn from other people’s experience

Cognitive biases that impede good decision

  • Decisions are the outcomes of our beliefs
  • Hindsight bias impedes against quality decision making
  • Guard against black or white decision making
  • Availability bias means lagging any prior conflicting data, our default setting is to believe what we hear is true
  • Selective bias and consistency bias, means we are unwilling to change our mind despite contrary signals from the environment
  • Avoid attribution bias

Related Readings

  • Theory of Games and Economic Behavior, Jon Von Neumann
  • Ignorance: How it drives Science, Stuart Firestein
  • Stumbling on Happiness, Daniel Gilbert
Posted on

Insight from dinner with Sujit

  • hard to find experienced PM with data background in the Bay Area
  • Mountain View just issued permit to build more houses. Price of existing houses got depressed by 10%
  • Dressing well allows ability to charge USD75 per hour more
  • Most data scientist hang out at twitter, vidhya analytics and Kaggle
  • Growth frameworks is the easy part, its getting everybody to row in the same direction that is hard.
Posted on

Insights from house party at Qyunh’s

On Crypto-currency

  • Major institutions like hedge funds and banks are trying to figure out how to increase exposure to this currency type while mitigating against volatility
  • Governments are exploring into how this new  form of currency can be regulated
  • Companies in this space are increasingly gearing towards providing missing infrastructure like secure wallet as opposed to creating new Token
  • Transaction throughput
    • Visa gateway allows for 50,000 transactions per second
    • Lightning networks as well as off chain transaction provide ability to temporarily by-pass bottleneck
  • Promising candidates
    • EOS (Chinese currency) supports up to 1,200 transactions per second
    • TRON (Chinese currency)
    • Ethereum

On Data API

  • New travel portals need to be able to prove traction before existing providers like Viator is willing to become partners
  • Normal affiliates get 7% commission while partners get 18% commission
  • This is a missing step in this eco-system that startups will need to gap before they can reach partner status.

On AR/VR

  • Much of efforts is done on AR right now as opposed to VR. Where there is actual industrial application

On Reality

  • Accurate interpretation of signals provided by reality requires the ability to set aside predisposed lens
  • Morality is one of the hardest lens to set aside when interpreting signals
  • Morality derived construct like the Good and Evil dichotomy is inherently subjective, while useful for marketing, branding and messaging purposes, is not useful for analytical purposes.
Posted on

Book Summary: Lost and Founder

Radical Candor/Transparency

It is hard but it works – needs to be tampered with empathy

On being product focused

  • Consulting is limited by time and people – not scalable
  • Effective Product-focused business
    • reach
    • scalability
  • Start with a product informed by your consulting – real life problems others face

Impediment to shifting focus

  • Too comfortable
  • not enough time
  • difficulty finding the right customers for the product

On being a founder

  • Great founders enable a vision
  • forget about being hands on most of the time
  • job scope changes every six months – for any road block encountered focus on sufficing the requirements instead of perfecting it
  • you rarely get to do what you love to do
  • be cognizant on when to lead and when not to – have the specialist do the job
  • Cultivate self awareness in strength and weakness – structure company to work around them
  • Attribute of founder is instilled with near-permanence in the organization while those of supporting team fluctuates
  • the hardest parts of the business is less a reflection about the business than about the person experiencing them
  • Build expertise before building network, build network before building company
  • Focus on and reward the behavior, let the outcome take care of itself

On Values

  • Authentic values force hard decisions – held to be more important than money
  • have real costs: Impede certain behavior and strategy
  • Values are discovered instead of set
  • Used as a yard stick for recruiting new members to the cause. Helps get pass the competence versus cultural fit dilemma

On recruiting

  • CTO should be those that should be oriented towards education instead of shielding you from the nitty gritty details (black box)
  • Use your value system as a yard stick
  • Hiring for diversity will make the mental model of the organization more holistic
  • Great managers / coaches might not be great individual contributors

On markets and pivots

  • Pivots Are expensive don’t make it a habit – only resort to this tactic when the original hypothesis is not longer valid
  • focus on the market and then find a field ignored by others because it appears unsexy. From there craft a solution
  • Err on the side of execution

On investors

  • Need to take money for the right reason
  • Investors interest will tend to get out of alignment overtime (return multiples and investment horizons)
  • 80 percent of returns are by 20 percent of investments
  • They need at least a 10X to break even in a position for all the other losing positions they took
  • They don’t bring much value to the table
  • follow up with CEOs they invested in to understand how they react in a shit storm
  • Can help provide information on salary ranges

Choosing a market

  • If you can keep your ego in check you can chase after smaller markets and don’t need VC money
  • Great ideas are born of mediocre ideas that become better by
    • time spent iterating
    • humility learning
    • surviving
  • look for searches that indicate problems
    • Google Adwords
    • Moz’s keyword Explorer

Knowing your customers

Defining your user base

  • Call 3 different types of users
  • Find out why they subscribed and stayed
  • Craft messaging toward this group of people

Discounts are a doubled edged sword – while they might attract signups, these folks tend to have a higher churn rate

Schedule regular interactions with your user so that you can understand their habits. It helps you get to an empathetic position with them.

On Products

  • Feature set needs to be coherent enough to be able to deliver value
  • Early adopters
    • have very different expectation as compared to early majority –
    • hence more forgiving
    • ok accepting MVP
  • Retention triumphs acquisition any day

Marketing

  • Optimize for acquisition loops that reinforces the UVP instead of linear acquisition channels

Focused Execution

  • Practice the discipline of focus.
  • Important to Focus and not waver around unnecessarily. Its a waste of resources
  • A very focused and simplified product offering will help users to more easily understand and adopt it
  • Helps keep teams lean as a by-product
    • ROA improves dramatically
    • helps avoid future layoffs
  • Focus on what will not change in the next 10 years

Related references

  • Lean Startup, Eric Ries
  • Sprint, Jake Knapp
  • Venture Deals, Brad Feld
Posted on

Reflections on communicating with your users via email

While sending a personal email to each individual user who directly uses our API just now, it occurred to me the main difference between talking with someone you have relationship with (like your mum/girlfriend/wife…?) versus simply doing a mass email blast to a group of “strangers”, is the potential lack of warmth and empathy in the latter on the sender’s side.
 
No one really likes being treated like a number on a Excel sheet. It sucks.
The key challenge becomes how do you scale your communication as the amount of people you need to communicate with increases without alienating them. Or does it even matter?

Related Reference

  • Permission Marketing, Seth Godin
Posted on

Reflections on organizational observations

On the benefits of being under resourced

An organization that is under-resourced is similar to an individual who regularly practices intermittent fasting.

In the case of the individual, weaker cells are cannibalized by the body to produce new cells. Studies has shown decreased probability of developing cancer and Alzheimer, coupled in some cases with increased longevity.

In the case of the organization, operations are forced to be focused and only opportunities backed by stronger market signals prioritized and pursued. Wastage and distractions are structurally curtailed. ROI improves as a by product.

On organic growth

An organization that is organically grown tends to be happen-chance and exhibit somewhat illegible and opaque structure.

In contrast, an organization that is deliberately designed tends to exhibit a legible and transparent structure.

Counter intuitively, it is the former that exhibits more resilience in times of environmental stress. The observed illegibility and opaqueness is attributed to the process of organic adaption to its environment. Chief historical examples are:

  • Military of Carthage
  • Genghis Khan’s Mongolian horde

Related readings

Posted on

Key take aways from the Blockchain revolution

The Blockchain revolution
  • ensuring the integrity of data exchanged among these billions of devices without the need for a trusted third party
  • allow people that do not have access to the service of these third part into the digital economy
  • easier ability to get compensated for your work or ownership of digital property
  • Ronald Coase on types of costs
    • Searching cost
    • Coordination cost
    • Contracting cost
  • Dimensions of search
    • horizontal search – wide search across the web
    • vertical search – within a specific website
    • sequence – blockchain?
  • Innovation typically comes from the edge
    • monopolies have a lot of resources but lack the culture and will to explore, Yochai Benkler
    • This can be attributed to high levels of bureaucracy within the core

Related references