Business Plan(ish)

Prelude: Why am I starting a new company?Why not continue focusing 100% on Loki?Why not join a fast growing company as a data guy?Why start a new company?What does the new company do?1. Data Acquisition 2. Data Cleansing, Augmentation, and Standardization 3. Data Storage 4. Data APIs 5. Data Augmentation 6. Algorithmic Insights 7. Data Visualization 8. Automated Reporting 9. Predictive Modeling What customers do I serve?Short-term Potential customers Long term How do I distribute the product/acquire customers?Phase I: Get usage and trial customers Phase II: Real developers using it post product-market-fit Phase III: Hiring sales people to sell this in the enterprise What are the headwinds and tailwinds that I face?How do I monetize?Who will I have to hire?How will I stay ahead of the competition?What am I excited and scared about?Scared about Excited about Feedback

💡

This is a work in progress, and contains several bad, mediocre, and/or not well thought-through ideas. Trying to work more in public so I can get more feedback – solicited and unsolicited – from others. If you have some, please email me at rishabhsriv@gmail.com or @ me on Twitter @rishdotblog

Prelude: Why am I starting a new company?

Why not continue focusing 100% on Loki?

I have been working on Loki for 4 years now (3.5 years as a registered company, 6 months as a freelancer before that). I have customers, am default alive (revenue > company + personal expenses), have achieved product-market-fit (no churn), and have loads of room to grow more and get to a nice $20-30k recurring revenue amount

But I have been turning down customers for Loki recently. Working with media companies is just not fun. It's labor intensive. I have to do loads of customization and integration work to get anything done, because things often move so slowly in these companies that it's hard for them to use my products otherwise¹

Selling to some of these companies is also quite painful. Many of them are currently losing money, and so negotiate a lot to bring the price down to ridiculously low levels

Most importantly, the future of these companies is not bright. Moving forward, most good companies will either in-house the functions that I'm currently doing for them (ala NYT/WashPo) or they'll go out of business. This turns selling to media companies into a short-term finite game instead of an infinite one – I should not be playing finite games this early in my career.

💡

Continuing to work on Loki feels like a cul-de-sac. I don't think I'm quitting during the dip – though I may be wrong here

¹ Example: they often struggle to convert a JSON file into an XML one, or to manipulate the JSON structure into something that works on their frontend

Why not join a fast growing company as a data guy?

This is definitely an option, but I think I can get far more freedom + more financial returns if I start my own thing. I currently have enough savings to last a few years. So I want to try launching and selling new products, and see how that works for me. If it gets no traction, I will become an employee at another company

Why start a new company?

I've realized 2 things while working on Loki:

I'm able to solve problems that others find difficult to solve. Like creating articles that are 100% automated (this, this, this), figuring out alternative forms of tracking economic growth in developing countries, and more I think this is because I have a fair bit of tacit knowledge in multiple fields (where to find what kind of data, how to create predictive models that work, how to create a dashboard that people actually want to use, how to create robust automated workflows, how to build cost effective data pipelines etc). I got this tacit knowledge by working on a very wide variety of jobs as a freelancer and while running Loki. This is hard to replicate for others

I have developed a bunch of software libraries and functions that help me become hyper-productive when doing anything data related. These include hacks around scraping data, functions that make data visualization and dashboard creation easy, functions that allow me to analyze text really easily, functions that allow me to get insights from satellite imagery really easily etc (many many more)

These two things allow me to become a highly leveraged, full-stack data developer. Tacit knowledge + software I've built creates leverage. Experience in working across the stack (data scraping, cleaning, analysis, dashboards, modelling, server management, API creation) creates full-stack ability

Right now, the data world is super balkanized. Enterprises had a "big-data" hype fueled rush to just buy whatever software solved their needs at that point in time. And they are using bad, sluggish, limited software that hardly talks to other software to do their job. Better software can make things so much better for them

I have seen the amazing results enterprises get when they use smarter tools with people that know how to wield them. I built a dashboard for the Singapore government department for $25k which a Govtech team that they had paid $1.5M could not build properly. Over time, this way of working (highly leveraged developers that use tools that work well across the stack) will become the new normal

I think I can help accelerate this new normal, and create great value for both myself and the world by doing it. So starting a company around seems like the most rational thing to do.

What does the new company do?

💡

One-liner: Tools that supercharge data-driven decisions in organisations.

We provide tools that help companies perform tasks across the information supply chain more efficiently.

Below is a simplified view of how the information supply chain looks like for many use-cases:

This is how we intend to supercharge it:

1. Data Acquisition

We have 4 ways of acquiring data:

Scraping publicly information from the internet, and then observing changes over time. This includes listings on job sites, property and car sites, government portals, flights and rails, electricity production, and more This also includes scraping text, transcripts, and metadata from social media feeds, media reports, photos, videos, and audio recordings/podcasts

Using paid feeds for hard to acquire data, like data about financial markets and paywalled data about the real economy

Using satellite imagery to detect changes over time (both free and paid)

Monitoring first-party data (web and app analytics) through our Javascript embeds and SDKs

2. Data Cleansing, Augmentation, and Standardization

This includes 4 things:

Cleaning HTML data to extract relevant numerical and text insights, and converting both these kinds of data into a standardized format for downstream storage and processing

Tagging images, videos, and audio with specific metadata using machine learning (including transcription)

Geo-spatial analysis on satellite imagery using machine learning

Converting structured numerical data into a standardized form

3. Data Storage

This is a simple database storage layer, and consists of 5 kinds of databases:

BigQuery for data warehousing

A PostgreSQL database for all numerical data

An ElasticSearch cluster for all text data

Static storage buckets for static data (photos, videos, audio)

Cloudflare workers for all customer-facing data

4. Data APIs

Converting the data in data storage into REST APIs that can be easily used by developers

5. Data Augmentation

Adding more attributes to already available data, like adding the labels "Gender:Male", "Ethnicity:Chinese" to the name "Cedric Chin" and "Gender:Female", "Ethnicity:Bengali" to the name "Medha Basu"

6. Algorithmic Insights

This involves querying and aggregating data to create useful insights for the end user. Examples of this could include an regional or sectoral economic index for a country, a "state of discourse" that emerges from analyzing media data etc

7. Data Visualization

This involves dashboards where users can drill down to whatever granularity they want to, and where important emerging trends are automatically highlighted

These dashboards also allow users to download data or visualizations in any form they want – as CSVs, JSON, images, or SVGs

8. Automated Reporting

This involves automation creation and delivery of reports (web pages, PDFs, slide decks, and emailers) that are created either at a regular pre-defined frequency, or when interesting trends emerge

9. Predictive Modeling

This involves automated updates of predictive models — like election models, pollution prediction models, user-churn prediction models etc

What customers do I serve?

Short-term

Right now, it can really help analysts, investors, researchers, journalists, and business executives understand, model, and respond to situations quickly

💡

If I charge $3000/month and can get 10 customers, I'll make around $30k/month to start with. Doesn’t matter how many users or what the usage is like – can charge a fixed price at the start to the first 10 customers. Customers get a shit ton of value. And I get breathing room + cashflows to hire other people

I can sell to a number of industries right now. Have worked with a customer in all of these 6 industries in the past, so should be able to test things out very quickly

Potential customers

Media Companies

Times of India

Straits Times

Channel News Asia

News18

India Today

Universities

SUTD

The IIMs

INSEAD

Think tanks and lobbyists

NUS Think Tanks

Various Chambers of Commerce in Singapore

Geopolitical consulting companies in SG

Investment firms and banks

OCBC

Citibank

Blackrock

Blackstone

Sequoia

Lightspeed

HDFC

ICICI

Kotak

Government departments

Singapore MFA

Singapore MTI

Singapore MINDEF

India MEA

India Defence

Strategy/data teams in large corporates

Singtel

Grab

Airtel

Maruti

TATA

Essel Group

Long term

Literally every large and medium-sized company in the world. All of reporting is data-driven. If you ask people to create reports for you, you can use our product. A shitty product like Tableau (which only looks at the "dashboarding" part of the stack) had $1.2B in revenue. I can do a lot better than that if this really works

💡

At the start, our differentiation will be bundling of the data stack services into one. Basically, a cheap and good-enough buffet to serve large and very hungry customers. Once there is enough market penetration, can unbundle

Some specific use-cases that we can serve include:

A "country dashboard" used by foreign ministries and strategy teams to understand how the economy and politics in a country are shifting. Scrapes a shit ton of data (numerical, speech transcripts, press releases) and combines it all into a specific use case

A "market dashboard" used by teams in companies to help understand how the day-to-day realities of a market are changing (like Singtel monitoring the telco market in Singapore)

An investment dashboard used by investors to see how the data around an industry, company, or commodity is changing

How do I distribute the product/acquire customers?

(this is sparse, because I haven't thought deeply about this)

Initial set of customers will be old Loki clients. Beyond that, I can either open-source software tools to get more adoption, OR do an Intercom-styled content-marketing approach to get inbound leads

More of the approach here:

▶️Approach to B2B Growth and Sales

Phase I: Get usage and trial customers

Own network and old clients

Hacker News

Twitter Networks

Phase II: Real developers using it post product-market-fit

Content Marketing

Paid ads on Twitter and YouTube

Phase III: Hiring sales people to sell this in the enterprise

Wining and dining customers

What are the headwinds and tailwinds that I face?

Headwinds

Departments are currently balkanized and comfortable with balkanized tools. They may want to protect what they perceive as their turf/old decisions they've made and not want to use our tools

[if I open-source this] Asian companies don’t have a culture of paying for commercial licenses for open-source tools

Will be difficult to compete with copycats that will inevitably emerge [don't have to care much about this if I already have customers]

Employees are not empowered to ask companies to buy tools in Asia

People may perpetually be stuck on using our free tools, and we may never be able to monetise properly

Tailwinds

Open-source software is increasingly being favoured in the enterprise

Asia is increasingly following western practices and empowering employees — specially new tech-focused companies

JAMStack is rising, and is currently the most efficient way to create apps and services

Appreciation for data-driven search and decision making is growing everywhere

Public data is increasingly getting richer everywhere

How do I monetize?

(Haven't thought very deeply about this)

Option A: Free and open-source core. You pay for cloud products, live data feeds, our hosting and support. Flat $3,000/month for data feeds and support, pay as you go for hosting [my preferred option]

Option B: Monetization through courses and licensing. No recurring revenue, but also no maintenance or support

Option C: No monetization. Just user acquisition. Then I get acquired [bad idea]

Option D: Pay-as-you-go model for the enterprise [may be a hard sell]

Option E: Consultancy model where I essentially set up a data agency for anything data related [may not enjoy this]

Who will I have to hire?

Can start by myself. Once there is revenue to hire more people, I would have to hire:

Front End Devs for making better dashboards and interfaces

Sales Team

DevRel Person/Evangelist/Marketing person (Medha’s feedback: very important for non technical people — they don’t know what full stack data is. Talk about how this make things better for them. It’s not just about the money. What is allowing me to do better. What new things can I do now. Talk about why different layers of the stack talking to each other is important. What difference does is make to management?)

Full Stack Developers

How will I stay ahead of the competition?

Keeping adding data sources consistently

Keep the product in the public eye with excellent content marketing

Capture a large chunk of a niche but important field: business school students and data analytics students. Give the product away for free to them if needed

Create a LOT of free content & videos that demonstrate how how users can use our tech to impress their boss

Get other open-source projects to use your project, creating dependencies [if I chose to open-source everything]

What am I excited and scared about?

Scared about

This is a fairly big undertaking and is difficult to get right. Theoretically, a "full stack data" approach makes a ton of sense and is where the future of enterprise lies. But this may be a hard sell. It's much harder to explain than "here's a data visualization tool that non-programmers can use"

Also, it'll be really hard for me to polish the product (specially from a user-interface perspective) if I start out trying to do everything. I'm confident of creating kickass APIs that works really well across the data stack. But am less confident of creating a polished interface for all the jobs to be done with this product

Excited about

This will create a shit ton of savings for companies and can create a lot of impact in the world. Even if this fails to scale into a big product company, doing this will definitely help me become a much more equipped data guy. And I can always convert this into a data-science agency if all else fails

Feedback

From Cedric

There is already a full stack data company. It's called Palantir

In a lot of the stuff Loki has done, the initial engagement is brainwork. All subsequent engagements are grey hair work (just pattern matching things from before). That initial engagement is hard

If you don't require a consultant (including for the customers to realize they have a problem), only then you're a product company

One way to describe this is data journalism skills with some automation applied to business/research contexts (scraping, visualization, uncovering trends)

Potential positioning: "Intelligence augmenting data tools"?

Alternative business model: Provide constantly updating data + tools to people that make decisions that require a lot of analysis