Class 1

Digital Disruption – The Incumbent’s Dilemma

 ‘Value chain’ meaning the link between producer and end-consumer.

Companies exist somewhere along this and get a margin out of the
value created there.
 When you say everything is digital, there is a sense that some
digital tech is a part of that value creation – tech as a “strategic
partner” rather than an add-on – central to operations on some level
– at least the back end.
 To be a digital good is to have:
i. Reproducibility
ii. Transmissibility
iii. Manipulability
iv. Storability

Each of these cut costs along the value chain – freeing cash for
other things – efficient.

 New business models – particularly digitisation – changes user

behaviour. This creates new ops, widens your value network – shifts
the ecosystem. This is disruption.
 “overserved” – (I want low quality, but only higher quality is
available) creates ripe market for a new player to come in and offer
something at a lower quality and still take away those customers.
 Disruptors normally have a better innovation model because when
they try to hit the bottom tier (budget cars, steels etc.) – they value
efficiency more across their processes.
 Disruptors are also focusing on a specific aspect while incumbents
are busy focusing on their own value prop or doing wider, more
spread out operations.
 They disrupt the traditional value network – the processes, areas
where profits are coming up etc.
 The disruptor must be able to innovate a lot faster than the
incumbents – otherwise there is no point – the incumbent will be
able to catch up soon.
1. Is it serving an underserved part of the market?
2. Making process better to serve the top line too.
3. [Write the 6 questions as per the Christensen framework]

Class 2

Introduction to AI: History, Advances, and Importance for Managers

  Is AI a cost leader or a differentiator? Both, and in a significant way.
Leverage on creating competitive advantage, but also for creating
value faster and more efficiently (therefore cost leader).
 Network effects an important aspect of AI – the sum is greater than
the parts. The insight from the data a lot more important than the
data itself. The fundamentals of AI is always data.
 AI need not take complex decisions, but the fact that it can ‘enact’
a decision makes it AI. E.g. the thing that unlocks your phone, Chat
makes decisions each time it chooses what the next word will be
(which then sets the trajectory of the whole chain of thought).
 If it has the ability to enact a decision – no human intervention –
then this is AI, not ML.
 Counterfactual analysis is something AI is quite good for
 IBM Deep Blue – played chess on a series of if-else statements,
defeated world champ. Showed that our intelligence can be written
into AI, which can then make decisions etc. – and can be made into
an actual usable application.
 Causality is a lot more intensive to establish than correlation – for
firms, the latter is alright – whereas policy firms need causality.
 “Fair” AI – The context of the business says whether something is
fair.
 To see where you are deriving value from AI – ‘IDEA Framework’:
o Identify problem- what is the business problem?
o Define objective – what is the business objective?
o Enact decision – what is the business decision?
o Analyse data – what is the data mining problem?
 A problem statement should always be a statement.
 Deep Learning – neural networks – large amounts of unorganised
data – allows you to infer quality analysis from non-linear data.
 Reinforcement learning is like a blind person walking – give it an
environment and tell it to learn – there is no right and wrong.
 Enact a decision and do something with what you’ve learnt 
AI.

Class 3 – GenAI, Prompting & Ethical Aspects of AI

 Prompting: Right set of questions with some structure that gets us

the answers we want.
 Learning from the past and applying it in new context: creative?
 Weights (regression betas) are parameters – ChatGPT being a 5
billion parameter means that there are 5 billion betas – all to just
produce that first word. Tokens transforming and giving a particular
output.
 Sequence of predictions – structure for AI – the architecture of
weights in between (middle; modulating) makes for better odds of
this neural network yielding the right answer.
 Creating a “model” means learning these weights (betas) – that the
model then uses – think like a mathematical model. The weights
look like some number between 0 and 1.
 Model: Set of collection of weights – going through it somehow gives
the mathematical prediction at the end of the day.
 Bigger models (more parameters) means that they know about
more contexts – not necessarily that the answer itself is better.
 Rag – AI to answer only in a specific context with certain constraints
in a particular environment. Fine tune it with particular information
but it is deployed in a context. E.g. ISB admissions bot.
 Parameters and number of pathways is the difference between
models.
 Open Weight Model:
 CHAT GPT: Makes money only on subscriptions. (Chat GPT as
different from GPT-4).
 Claude is an application, Sonnet is a model.
 Applications make money by selling APIs where the application is at
the back end.
 API: Connectors that let the application talk to the model.
 Llama, Deepseek etc. – they have a different business model where
they don’t charge for API and users making other models – they also
don’t host the app – make you host it on your own – These are Open
Weight or Open Source models – allow you to keep more of your
data to yourself (not giving leverage in the form of information to a
single app). They all have very different business models – for e.g.
Meta has access to whatsapp chats – can use info from its AI in ads
etc. too
 Prompt engineering – getting the results you want.
 Task Definition (clearly state what you want it to do – “explain”,
“analyse”, “generate”)  Contextual background (tell it context +
give it a persona)  Output format (specify exactly how you want
the response structured – bullet points, paras etc.) 
Constraints/Guidelines
 Techniques:
i. Zero Shot Prompting: Technique where you give it a one-shot
task – this allows you to leverage LLM’s advanced
understanding of these concepts. E.g. explain quantum
computing to a 5 year old.
ii. Two-Shot: Give some examples
 iii. Chain of Thought: Instruct LLM to think step by step or reason
through the problem – helps with complex reasoning.
iv. Self Consistency: Ask for multiple independent solutions to the
same problem and identify the most consistent answer.
Improves accuracy for mathematical and logical reasoning
tasks.
v. Uncertainty Routed Reasoning: Instruct LLM to indicate its
confidence level and provide multiple approaches when
uncertain. This creates more reliable responses and highlights
areas.
 Especially when using agents, important to know what structure is
to be followed when doing prompts.

Khan Academy

- KA disruptive, Khanmigo sustaining

- Build for competitive advantage, rent/buy for parity – here,
the AI itself is not the CA – it is the underlying pedagogy and content
that is the CA. So makes sense to rent the AI – GPT.
- Holy grail: 1:1 tutoring level + quality
- Fundamental shift:
o Adaptive tutoring
o Active mode of learning
o Changes org, talent, costs for KA
 Makes the company AI-centric (not content centric)
 Working with external partner (Open AI)
- Disruptive / fundamental shifts have the ability to change cost
structures, org, or even mission statements. Cannot get money from
the same sources you were getting before – Google and Microsoft
(huge investors) suddenly became competitors.
- Hallucination huge risk for business functioning, brand, loyalty. Still
went for it – why? User base is leverage – if I don’t do it someone
else will – if someone for-profit does it, then the moat for them will
become so high that you will not be able to get back into it and
achieve your mission.
- Ethical dilemmas - give AI? Issue being that AI (1) random (2)
training (3) who accountable? (4) without such considerations.
Accountability unclear – means areas of improvement limited.
- Should be able to clearly define ethical bounds. Know where it works
well.
- Disclaimers cant do enough
- KA wanting to be a main player in the market so they can control
what happens in this space – go back to Chat and refine the model
itself with feedback.
Class 4

- People who ask more questions and are uncertain get better use out
of AI.
- GenAI shapes the illusion of competence – leads you to have a
disproportionate amount of self confidence – overestimate
competency and understanding.
- The skill you want to be developing is being able to detect when it’s
wrong
- The issue of bias is generally when what is ‘not in the data set’ is
not what the model is also trained on – this means that AI detecting
criminal area will be disproportionately biased towards black areas
because of the data that it is trained on – and the fact that it can’t
be trained on the ‘absence of data’.
o This means the model learns only from the outcomes it can
observe, and because it cannot see the counterfactuals, such
as people who were never released but may not have
reoffended, it ends up reproducing and amplifying the
historical biases embedded in the data.
- With models, it is also tricky to pinpoint which part of the model
points someone to be something (e.g. visual model classifying prof
as ‘woman’ on the basis of hair) – for a neural network with billions
of nodes this is tricky af. Accountability issues – this is
Transparency and Explainability. The opposite of this is a Black Box
model.
- Ethical AI, key considerations:
o Bias and Fairness – Chicago case
o Transparency & Explainability – making AI decisions
understandable and interpretable by humans – bank
loan
o Accountability & Responsibility – defining who is
responsible for AI decisions and ensuring ethical use –
Khanmigo
- Some models (e.g. language models) neural networks work better –
for some other types (like loan) decision trees work better.
- Develop metacognition
- AI is here anyway; build and learn to use it
- Agentic AI – AI with even more agency to make decisions – continue
the workflow based on its own thinking
- Agent is the part that comes before the workflow – agent is
triggered, and then the workflow follows.

Class 5: Metaverse
Web 1.0. – static web pages

 The passion for the internet seems to have been premised on a

constant need for information – whether that be Yahoo making the
news, Google organising the news, ads then capitalising on that
information.
o Serving needs  build trust  make ecosystems

Web 2.0. – transactions

Next paradigm shift: Metaverse (Web 3.0.) – there will be fresh needs,
fresh behaviour, and new people catering to them

 Decentralised – the word of Web 3.0. – this should ideally

ultimately mean that there is final resting power with the user itself
and not with the corporations.
o Security, privacy, decentralisation
o Blockchain comes in here – to decentralise and distribute
power – the more information an agency/corp has, the more
power they have, should they choose to do something with it
o To attribute ownership to the person who actually built the
virtual assets – blockchain - virtual assets: NFTs.
 Meta, for instance, has data across teams / across siloes by
integrating with Whatsapp, Insta etc. – to become a key habit.
 The news feed and the ability to let the user completely scroll were
the only things that Orkut didn’t have that FB had – Google owned
Orkut, they already had a user base too, yet FB was able to disrupt.
 Metaverse – Augmented reality + virtual reality –> extended reality
 Can have ways to take out biases, not sure how it’ll sustain
 Augmented reality is reality++; more contextual information in
space.
 Use cases:
o Enterprise simulation:
o Social & Comms:
 Things to think about: Maybe appearances could
change, people could misrepresent
 But alt – think interview where you neutralise things like
accents, sounds, gender, appearance etc.
 Change in user behaviour, control of information shared
 People’s behaviour might change when such technology
is adopted.
o Gaming: Immersive gaming
 Change in behaviour always leads to new opportunities.
 Bystander effect – behaviour changes when we think we are being
perceived.
 o Deploying a store in a place like the metaverse needs to be
studied – for instance, Nike found that people didn’t like the
idea of going into an empty store and wanted people to be
there.
 Gaming & Simulations are the first use cases for the metaverse.
 App store world view –
 Interoperability is very difficult to build and hold – to have multiple
protocols and to integrate within them – this is hard when there are
multiple metaverses.
 The most important thing with social media-connected tech
companies is capturing the market early – first movers advantage.
 How do you evaluate who will win in a market when a new paradigm
shift happens? Who do you place your bets on?

Highest reward – Time the market x New

Sure to get something when there is a paradigm shift – Drive the market x
Existing Needs

Framework to evaluate who wins when a new paradigm shift happens:

Mobile Paradigm Time the Market Drive the Market

New Needs Uber (I) App/Play Store (II)
Existing Needs Photoshop (III) Chips – Intel (IV)
Winamp Cloud Services
New needs take away from companies that already cater to existing
needs.

(III) is the incumbent’s dilemma.

(I) Least risk, high reward

(II) Highest risk – but also high reward
(III)
(IV) Low risk, but higher rewards because you’re helping someone
else

CLASS 6 – BLOCKCHAIN

 4 Pillars of Blockchain: Security (Privacy), Immutability,

Transparent, Decentralised
 To track ownership and attribute accountability to the players
 Decentralised system where we all collectively track and
authenticate something – blockchain – a bad actor cannot come and
change it later – there is a chain of authentication for it.
 The idea is – why should the bank make money out of doing
nothing? Why is the bank getting transaction fees and middleman
 money when the rest of us can just do it between ourselves? They
get information too – no need to monitor. Trust the complete
information and consensus of the users.
 Torrents also a decentralised system – you would be getting it off
any of the other users.
 Why is crypto valuable?
o Trust
o Scarcity – designed with the idea that there will only be a
limited number of combinations possible for the bits (therefore
finite number of bitcoins possible)
o Bitcoin mining: (in short) Consensus mechanism – it is adding
on to the next block (#, private key)
 How to operationalise?
o Ledger information carrying over – a series of ledger so you
know who owes who what
o If ledger of financial transactions, cryptocurrency – this is
encrypted – so the security you need so that people cant
irregularly use this information or access it
 Core components of a blockchain:
1. Blocks & Transactions – prev block ref, time etc. –
transaction data
2. Hashing, crypto: ensures data security & integrity
3. Consensus mechs: proof of work (PoW) & proof of stage
(PoS)
4. Smart contracts – self-executing contracts w coded terms
 Smart contracts: The contract is entirely codified – a series of if-else
statements and it is immutable. People validate that there is indeed
ownership.
 ONDC and NammaYatri – blockchain
 Tech is not unidirectional – AI etc. will feed into this and vice versa
 Tech helps you build value – and value should be a focus.
 Centered around consensus – it is about agreement

Class 7

Quantum Computing

Proof of stake – higher price – the idea that the higher “stake” has the
higher right – can lower the reward also proportionally because it doesn’t
take as much effort to produce now etc.

Music industry was de-bundled – value generation & value captured got
closer together
Underlying framework:

Frontier Tech  New Needs/Behaviours  New Business Ops  New

Business Models

Qubit – Quantum bit – 0 and 1 at the same time – existing on the value
that it is really at – like a flipped coin spinning on the side. This is giving
you more info than a standard bit that needs to collapse to either 0 or 1.
This allows it to explore multiple paths at once.

When qubits interact, there is some quantum entanglement

Qubits allow you to find things, or find a solution from a bunch of

things, way faster. TO find where something breaks. Especially great for
counterfactuals.

[[Watch MKBHD IBM quantum computing]]

- Not faster in everything – what they’re good at is finding structure in

tons of data.
- “Getting a boat”
- Observe the pond and make connections from what it sees in how
these waves interact
- Can break encryptions
- RSA: multiplying large prime numbers – difficult to factorise and get
to the underlying info – now maybe can be factorised easier.
- Will need quantum-safe algorithms
- Quantum is inherently a probabilistic system, not deterministic. So
when the qubits interact, your solutions may not always be the
same – even for something as simple as addition. You don’t want to
use this for all use cases, avoid simple use cases.
- We’re nowhere close to where we need to be – gaps: 2-3 magnitude
from target
- ~400 qubits is where we are right now – need millions to do the kind
of operations we think we need to do
- Qubits are very sensitive to environment, go bad more easily – this
process is highly delicate
- Harvest now, decrypt later: what hackers do at the moment.

Mids

MCQ (no partial marking, & T/F section – choose + reasoning (1-2 lines); 1
side handwritten

Do the different frameworks – Winners & Losers types too

Value Chain
- With time, the value chain always changes – as do the places where
value is captured & created.
Session 8 (Vishal K.)

Disruptors use different business models in conjunction with the tech in

order to defeat incumbents.

Deconstruct the business model of a product:

Dropbox

- Deeply technical founder

- Enterprise v. Consumer: enterprise – margins – CAC – match other
companies – separate product- scale – consumers easier to go to
integrated models – enterprise more sticky too (long-term contracts,
more stickiness) – more data, more lock-in.
- Build for competitive advantage, Buy for parity – if this is just
one part of what I’m doing, then there is no point for me to build this
up.
- Structure on how to make this decision:
o First: Customer Value Proposition: PROBLEM 
DIFFERENTIATED SOLUTION  TOM (TECHNOLOGY &
OPERATIONS MANAGEMENT PLAN)
 Is this product even valuable for customers?  What is
the problem we are solving? (1) My inability to access
my data whenever and wherever I want (2) Loss of data
(3) Difficulty in collaboration
 Integrating these three capabilities into a single product
– these are problems that occur in tandem – so a single
product that solves for all this.
 Targeting non-tech-savvy consumers (“individuals”) –
pick the customer segment where you can add max
value to
 Trojan strategy – Land and expand – picks customer
basis the hypothesis that they will help them go to
another segment
o Differentiated offering (each of these addressing the needs of
the user)
 Reliable
 Simple
 Incremental upload – this increased speed – which was
important at the time because internet speed low + this
was worse than offline work. Understanding the core
issue and seeing how you can differentiate a solution
using tech – unarticulated user need (comes from
understanding user goals).
 Ability to breach firewalls
o TOM:
 Using Amazon infra
o GTM:
 Launches video and goes to market first in the Hackers
website – this captures innovators because of the desire
of innovators to give feedback and their skin in the
game.
 They are also risk-taking
 SEM (Search Engine Marketing) didn’t work bc it works
when there are targeted searches – people don’t know
what they are searching for
 Freemium model – look at the conversion ratio -
 Traditional GTM didn’t work:
 Referrals to drive up virality (>1)
 Shared and public folders to drive up direct
network effects – this also transforms users to
makeshift customer service execs
 B2C: Person who buys the product is the one who uses
the product
B2B: Managers buy (focus on security, privacy,
integration, performance), but the entry-level users who
use the products value performance and are different
from them
 Here, there is nothing being spent on marketing – they
think of the cost of free customers as marketing spend
(for CAC)
 (1) Timing (costs, enterprise) (2) customisation: value
add (3) Higher data usage – usability (transfer is
underway)
 “Hope is not a strategy”
 LTV – calculate enterprise stickiness for 7 years
 Fundamentally you make this decision on quant and
qual:
 Diversification is a good idea when you have FCF
(free cash flow) internally without needing to
borrow or go with VC cash
 With Dropbox at this point they will have higher
FCF only when there is higher conversion – at this
point that will only happen when they build their
brand
 You analyse internal and external resources quant
and qual
  Quant is always the important part of this – qual
augments the story.

The story beyond

They are at $2B valuation, 0 debt, $500M FCF – massively

They went to enterprise in 2013 – after their conversion rate went from
2.5% to 5% (where they then got FCF) – started building 2012. This is
despite Google, Microsoft etc. entering the industry and going at it
aggressively – technology as a core offering: continued to be the case.

1. CVP
2. TOM (differentiated)
3. GTM – financially viable?
4. Meaningful profit formula

Class 9

Value – what you get out of a product?

- Value creating entity: song/music (information good)

- Value delivering entity: physical CD

CD-ripping software was able to separatee the two and then resell it. So
then the economy lets you trade just the value (information) without the
delivery. Information asymmetry allows you to arbitrage on this value.
Information arbitrage – all big companies are either doing this or helping
other companies do this.

Supply-side economy of scale: industrial economies (e.g. cycles)

Demand-side economy of scale: digital platforms (Uber in US – not

here)

Business models for information goods

Network Effects

 Both of these types could be + or – (not just dependent on what is

happening but also what side we are looking at it from)

1. Cross-side Network Effects: Between the different platforms

2. Same-side Network Effects: Inter-se one of the user groups /
demand sides

Platforms with varying levels of complexity:

1. [Link]
 Platform hosts infra (website)
Brings the two sides together
Defines the rules of engagement
Value of the platform changes depending on the number of people
interacting with the platform
Two or more groups with distinct needs and different benefits
There are network effects
2. Amazon
Hybrid platform
Does both merchant and platform role
3. Google
A dedicated ads company
There are at least 3 sides to Google in the ads space: Ads,
Users, Websites
Google makes an insane amount of money with ads – the numbers
show that people click on the ads (hence the revenue) – so contrary
to expectations, there may be lesser cross-side negative network
effects than you would guess. Look at the data always.

 Core intention is to build a platform where you maximise

mono-homing and minimise multi-homing – also try to lock-in
(make switching cost harder) – e.g. AWS makes it difficult to pull out
your data (terminate) by charging for that – storing is free.
 The idea with subsidising both sides (when platforms like
Zomato start off) the idea is to promote monohoming so that
the users don’t switch later when you start monetising one or both
sides.
 Google ads take into account bid $ and quality number (undisclosed
metrics).
 First price auction (conventional bidding) works for open auctions,
second price works better for closed auctions where there is high
information asymmetry.
 Buyer’s remorse usually happens for closed auctions on first price
auction – the score is a coefficient of the quality – a process where
the second person’s price works for you (to avoid winner’s remorse)
along with a denominator of your quality score (so if it’s higher then
the denominator becomes higher, making the price per ad lower).

The idea of this is: optimise for the ecosystem to make switching
difficult and try to ensure that people do monohoming. The big question is
also whether platforms are worth it with extremely aggressive
strategies. At the ecosystem level – with all the subsidies and without –
you want to make sure that you’re using the right amount of money while
still sustaining your network effects such that switching is difficult.

Principles of subsidising:

1. On Quantity
2. On Quality
3. Assymetric data network effects side

Things may still not work with this – because need to optimise for the
users.

Class 10

General strategy being that you want to maximise unit economics – but
here, you are OK (and it is almost desirable to be) bleeding money – to get
users, so that the other side comes, and then you monetise it all. Works
well for WTA markets.

Winner Takes All does not mean 100% - just means a significant share.

Winner Takes All market? (questions viewed holistically)

1. Are the multihoming costs (setup + ongoing) significantly high for at

least one side?
2. Does the side with high multihoming costs experience significant
positive network effects with the other side? (I will take users from
the other side with me)
3. Are the opportunities for differentiation limited?

Costs – not just financial, also set up effort etc.

- For things like shops listing on Amazon, same-side network effects

are also negative. The idea of ‘I need enough of them to be there
with me’ for users to be attracted to me is a critical mass needed
for the cross-side network effects to kick in.

Case: Google (Search) – WTA?

- Miles ahead in terms of share

- 2010 – Bing took significant share and was the fastest growing
o Bing became ‘vertical search engine’ – did exclusive data
deals with flight and hotel aggregators
- Some differentiation seems to be possible: privacy, language
(niches), data network effects, AI summaries, agentic etc.
- Google responded very strongly – bought the companies that had
these data acquisition agreements – this is highly irrational
behaviour but if you’re in a non-WTA market (here, first two
questions were no, so Google is not a WTA market)
 o BUT it can engineer network effects into certain markets so
that it behaves like a WTA market
- Google’s biggest threat is hyper-specialised vertical search engines
– users now go directly to the thing they want and search there –
Google is not the first platform people use to check now

If WTA, I can bleed to death.

If not, then how aggressive can I sustainably be?

Class 11

Merchant/E-commerce Model

 E-commerce and quick commerce work on similar principles.

 Why do we order online so much? (Value in e-commerce)
o Convenience: lower mileage costs
o Variety
o Discounts
o Available 24/7 – no temporal/spatial limits
o Price comparison instantly / transparency
o Richer data – with filters, customer reviews, personalised
recommendations, vendor behaviour, platform policies etc.
o Easy return experiences
o Newer models (subscriptions?)
o Privacy (in a strict sense of not literally having to go to the
store and have people see what you’re buying and carrying –
in other ways there is no privacy at all)
 These can all be categorised as (1) Customer Experience (2)
Selection (3) Price. Applicable in case of retail also – but usually in
retail they’ll have to make trade-offs (on at least one parameter).
 Competition of price on e-commerce:
1. With other online retailers
2. With offline retailers

What – in these – provides online e-commerce markets the most salient

competitive advantage?
In value-based sale (non-MRP) retail seems to be lower prices than online.
But in other products, generally prices are lower online.

 Real estate costs

 Optimisation costs
 Price discovery (competition in prices lowers prices)

 High value sales don’t happen online – luxury watches, Birkin bags –
the sales cycle is a lot longer and fundamentally very different.
 Menu costs are a lot higher offline – online, prices are changed all
the time.
 Frequency (and therefore inverse of quantum) of price change is
more online then offline.
 Online the mean is lower, but the standard deviations are wider (the
dispersions are wider) online – this is for the same product
(controlled for geography and retailer too) – so there is price
differentiation between users online too.
 Same product across different retailers across different geographies
– unit price is not a differentiating factor between the different
online ? (I think so too – price may not be the key differentiator
online v. offline)
 If selection is important, this implies that we are looking at the tail.
Multiple SKUs. Lesser predictability in selection. High variety means
high demand variability, meaning that there is a higher inventory
issue (meaning higher inventory costs).
o Amazon handled this by building up their own warehouses and
supply chains that used technology as the core strategy
o Information asymmetry is the core challenge to the
supply chain
o Walmart sued Amazon for trade secrets in 1998
 If you’re operating online, you want to go closer to the tail which
means supply chain optimisation needs to be done.
o Supply Chain is one thing – in addition, Amazon also controls
for Product Discovery Workflow (through Stores, Amazon
Recommends, Filters etc.)
o Equally critical to ensure that the product is discoverable
 Customer experience (or CX) are all from cart to sale to delivery to
return. Delivery is critical.
o CX (delivery) is important – a marketplace (you don’t have
that control)
o But this allows you to serve the tail – which may be critical in
online commerce
o “Long tail demand model”
 o Ability to serve the tail is what competitive advantage comes
from

Myntra

Is Myntra competing on price? No.

Competing on CX?

 Acquired Fit tech for virtual try-on

 Had professional stylists, lookbooks, fashion trends – we are the
fashion advisors
 Repeat customers contributed significantly to their total revenue –
this shows that CX investment was paying off

Myntra was madly unprofitable – being acquired was good for them. But
all-in-all, the one who benefited from this were investors who were
preparing for the Amazon attack. Amazon would’ve bled to win – so the
VCs made them consolidate. Then Walmart (the only other giant who can
take on Amazon) acquired Flipkart. Now – in the last year – Myntra has
hiked up its logistics spending and revenue (needs to control the last-mile
experience) – so the core is to serve the tail well.

There is no unit economics in quick commerce right now – it is

running on high funding. TATA never enters without unit economics – but
here the market is demanding even Big Basket to enter quick comm. Will
it succeed?

Class 12

Is MWH a ‘platform’?

Debriefing the simulation

- WTA flip happens super early

- Or if there is a huge strategic blunder (charging when the
competitor isn’t)
- Multihoming tends to increase if there is information transfer /
transparency – so it can be influenced and controlled through
information.

Melissa Wood

What is this business? (Is it a business?)

Answer three questions to deconstruct:

1. Who are the stakeholders?

 Followers
o Aspiration, trust, authenticity
 oTips, affiliate codes, discount codes
oBrand, lifestyle, content
oEngagement with Melissa
oContent (I think parasocial connection is actually the central
idea here)
 Subscribers
o Frequent content
o Personalised
o MWH Method
o Fitness/Meditation
++ Whatever is being offered to followers
 Brands/Affiliates
o Conversion
o Authentic exposure
o They get feedback and marketing from her: niche

Is this a platform business or a product business?

I think platform

 To see if this deal makes sense, calculate the LTV – with churn and
other numbers written in
 In creator businesses, you need to be able to segment users as per
social value, monetary value
 Find a way to monetise this: there is massive money in it

Different models:

1. Platform
2. Digital economy
3. Product

Class 14

Ganesh K.

Co-Founder, Promoter: Tutorvista, Blustone, BigBasket.

Business models North Star metric – in any business model, see what the
primary driver is for money. For instance, IPL: broadcasting rights, so need
the most number of people need to be able to see – need to be able to sell
ad rights to FMCG – convert non-consumers to consumers.

The objective is often to increase the TAM.

Changing non-consumers to consumers

Subscription model gives more touchpoints/data than a one-time sale
model so can be more useful. Even when it comes to things like milk
subscriptions etc. too this helps predict demand etc.

Logistics costs also come down – milk run (the variability costs reduce
because of this)

Meesho (take smaller cut from sellers) & Rapido (give money directly to
drivers) are examples of new players breaking into high WTA / network
effect duopoly markets – this is business model disruption in action.

Class 15

- Zoom also wanted to enter enterprise at some point – there is no

money in B2C.
- When Zoom was blowing up, Teams was competing with Slack on
collaboration and workspace movement. Zoom was booming on
video calling, specifically.
- Business Model: Teams – bundling/paid version
- If you’re in an allied segment and there is a higher tendency to
disrupt, then there is a decision to made on whether there should be
a response.
- Even in a WTA market, if a double-sided platform uses envelopment
– can take away customers. Same capabilities as the other platform,
but is capable of offering integration/bundling – this reduces the
double integration costs. This works for overlapping/same customer
bundle – because you don’t want to spend too much in acquiring
new customers.
- In business, a problem that can be solved through money is not a
problem.
- Google was worried when iPhone came out because we were
moving from a browser-based economy to an app-based economy.
Google enjoyed being people’s one-stop solution where they would
search for something.
o Google acquires and builds Android, then wins out on
distribution.
o Google responded to the iAd (tries envelopment (aggressive
strategy)) by investing 7 years in advance, figuring out
distribution to prepare for a threat that only materialised in
2014 – they get back at an aggressive enveloper
- Classic envelopment strategy – often adopted by Microsoft – bundle
it and give it for free – rife room for antitrust
o Microsoft doesn’t stand for anyone encroaching its value prop
o Incumbents fail to respond often because they face the
incumbent’s dilemma
GE Case

- Been reinventing for 125 years

- Specifically looking at 2001-2018:
- IoT: making each thing smarter (even in the physical world) and
then connect them to each other in the data space – capture the
data through instrumentation and then be able to capture/transit
the data.
- Predictable maintenance – can license this and sell it very widely –
will be highly valued by customers
- This can then be captured in a revenue-sharing model (outcome-
based pricing).
- Revenue/outcome-sharing models don’t have a ceiling – no measure
of WTP.
- Solutions model is highly personalised – allows for higher margins
too.
o Not scalable
o Goes to build a platform in order to scale
o They keep users/consumers on one side and specific software,
data-using companies on the other side (niches like defense
wind turbines). The companies wouldn’t know who’s who on
the other side.
- Predix (?) is now core business
- Incumbents are judged differently from new players /
disruptors – they are judged on their current period profits,
whereas younger players are judged on their future potential –
market is more optimistic about it too. This is what makes it difficult
for incumbents to decide to go after this. Lesser speculation for
smaller companies / disruptors.
- Not necessary that only one of disruptor / incumbent can
survive – often, co-existence is possible.
- WHSmith doubled down by going into markets where people
impulse-buy their books (to compete with Amazon who comes up)