Module 2, Lesson 1: Market Research

Where We Are in the Course

We're on Module 2, Lesson 1. This is where the over-the-shoulder work begins — no more slides or theory. We're at the computer.

Here's what we've covered so far:

• Module 1, Lesson 1: What You'll Need — tools installed, accounts set up, ready to go
• Module 1, Lesson 2: Understanding the Process — the seven-step pipeline from research to shipping code, and why each step exists

In this lesson, I'm picking up from the pipeline overview and kicking off Step 1 — Market Research. This is the first piece of real work we do before touching any product design whatsoever.

Before We Start

Here's what I'd expect you to have under your belt before this lesson:

From previous lessons:

• You should understand the seven-step pipeline we mapped in Module 1, Lesson 2 — this lesson is Step 1 of that pipeline
• You know why we do research before documentation, and documentation before specs — we covered the "why" in Module 1, Lesson 2

Tools / setup you'll need:

• Claude Desktop (the local desktop app — not claude.ai in your browser)
• A Claude Pro plan (or higher) — the free tier won't give you access to Deep Research
• An idea you want to validate — you're welcome to follow along with mine, but I'd encourage you to use your own

By the end of this lesson, you'll:

• Know how to use Claude's Deep Research feature to validate a market in under 20 minutes
• Understand what to look for in the research output — market size, competition, pain points, and the go/no-go signal
• Have a research document saved into your notes system, ready to feed into the next phase

About This Lesson

Duration: ~26 minutes video + ~20 minutes for your own research run
Skill Level: Intermediate
What You'll Build: A structured market research report saved to your Obsidian vault (or notes tool of your choice)

The whole point of this lesson is to show you that what used to take a team of consultants weeks — understanding a market, mapping competitors, validating a pain point — you can now do in about 15 minutes on your own. And we're doing it with real data, not gut feel.

Watch the Lesson

What We're Covering

Here's what I'm walking you through in this lesson and why it matters if you're trying to ship real things:

• Why most people skip research (and why that's a mistake) — I'll show you exactly what the old way looks like and why Claude changes the equation
• Claude Deep Research vs. standard chat — these are genuinely different things, and knowing which mode you're in matters
• Running the research session — the prompt structure I use, including the "ask me questions first" trick that makes the output dramatically better
• Reading the output — what to look for: market size, competitive landscape, pain points, unit economics, and the go/no-go verdict
• Saving the research with the Filesystem MCP — Claude writes the output directly into your Obsidian vault, ready for the next phase

1. Let's Set the Scene (~0:00)

Our goal in this section is simple: validate whether your idea is worth building before you invest any real time in it.

Before tools like Claude, you'd spend days doing this manually — clicking through a bunch of links, reading articles, pulling together spreadsheets, synthesising findings. Most people just skipped it. They'd have an idea, convince themselves it was good, and start building. And that's precisely how you end up six months in with something nobody wants.

Here's what I love about where we are now. Using Claude's Deep Research, you can get a structured, data-grounded market analysis — competitors, market size, pain points, monetisation models, go/no-go recommendation — in under 20 minutes. That changes everything. There's no excuse anymore to skip this step.

The idea I'm working with in this course is a co-parenting expense management app — a tool for separated or divorced parents to track and split shared costs for their kids. I came across this as an interesting problem. I'm not personally a co-parent, which is exactly why I want the data before I trust my intuition.

Use your own idea. That's the point of this exercise.

2. Claude Desktop vs. Claude.ai — Get This Right First

Before we run a single prompt, let's make sure you're set up correctly. This is a gotcha that trips people up.

2.1 Why Claude Desktop Matters Here

You need the Claude Desktop app — not the browser version at claude.ai. They look similar, but they're not the same. The desktop version is what allows Claude to use MCP extensions, including the Filesystem extension we'll use later to save the research document directly to your computer.

Here's how you know you have the right version installed:

• At the top of the app, you should see Chat and Cowork in the menu bar
• Open the sidebar, go all the way down to your name, click it, and go to Settings
• In Settings, you should see a Desktop App section — specifically a category called something like "Assistant Extensions"
• If you see that, you're good

If you're on the web version, the Desktop App section won't be there. That's your signal to go download the correct one.

2.2 The Model to Use

As of when I recorded this, the latest version is Claude Sonnet 4.6. Stick with that. Don't switch to Opus 4.6 — it's significantly more expensive in terms of token usage, and for a long research session, you'll burn through your limit before the research is even done.

3. Deep Research vs. Standard Chat (~2:00)

This is a teaching point I want you to really get, because it changes how you use Claude for this kind of work.

3.1 What Standard Chat Does

A normal Claude chat can answer questions, summarise things you paste in, brainstorm, write — it's drawing on what it already knows, and it can do a basic web search when you enable it. That's useful for a lot of things, but it's not what we want here.

3.2 What Deep Research Actually Does

Deep Research is different. When you enable it, Claude doesn't just look something up — it actively browses, reads multiple sources, cross-references findings, and synthesises everything into a structured output. Think of it like asking a very capable analyst to go away and do a proper literature review rather than just answering off the top of their head.

It's similar to tools like Perplexity if you've used those, but built into Claude with the same context window and reasoning capabilities you already know.

3.3 How to Enable It

In a new Claude Desktop chat, click the plus (+) icon inside the text input area. You'll see options — you want to enable Research, not just web search. Make sure the Research mode icon is active before you send your prompt. If it's not there, you're in standard chat mode and you'll get a very different (and much less useful) result.

4. Running the Research Session (~2:30)

4.1 The Prompt Structure I Use

Here's the structure of the prompt I ran for the co-parenting app idea. Adapt this for your own idea — the prompt is in your course resources:

I have an idea for an app: [describe your idea in 2-3 sentences].

Please do deep research on the best way to approach this towards getting to a go / no-go decision for this app.

Important: before you start the research, ask me all the clarifying questions you need. Don't begin the research until I've answered them.

That last bit — asking for clarifying questions first — is the most important part of the prompt. Here's why.

4.2 The "Ask Me Questions First" Trick

When I ran this, Claude came back with a set of clarifying questions before starting any research:

• What's your target market? (US only? English-speaking globally? Worldwide?)
• What's your intent with this research? (Validate the market? Understand competitors? Decide on pricing?)
• Who are you? (Technical co-founder, solo founder with budget, first-time builder, etc.)

These questions aren't just Claude being thorough — they're genuinely useful for you too. Answering them forces you to articulate things you might have left fuzzy. And the more specific and honest you are in your answers, the more targeted and useful the research output will be.

For the co-parenting app, my answers were:

• Target market: English-speaking markets globally
• Intent: Validate if the market exists and is worth entering, understand competitors, find the best monetisation model — give me a full go or no-go
• Who am I: Technical, with some budget to invest, starting from scratch

Be specific. If you say you're non-technical when you're actually technical-ish, the research will be calibrated differently. Claude adjusts its depth and framing based on your answers.

4.3 What Happens While It Researches

Once you've answered the questions, Claude will confirm what it's going to do, then go off and do the research. This takes time — for me it was about 14 minutes. That's normal. Walk away, get a coffee, come back.

You'll see it actively working in the background — it tells you what it's browsing, what it's finding. That's the agentic piece in action: it's not just querying a database, it's actually browsing and synthesising like a person would.

5. Reading the Output (~6:45)

After about 14 minutes, I had a full market research report. Here's how I walked through it — and what you should be looking for in yours.

5.1 Market Size and Opportunity

The first section to look at is what the total addressable market looks like. For the co-parenting app:

• Around 34 million separated parents across five English-speaking markets (US, UK, Canada, Australia, New Zealand)
• Less than 3% are using any dedicated app for this — huge untapped market
• The total market opportunity is in the range of $4 billion

You don't need to obsess over the exact numbers — they're estimates. What you're looking for is whether this is a meaningful size and whether it's underpenetrated. A large market with low existing adoption is a much better signal than a saturated one.

5.2 Competitive Landscape

Claude laid out the existing players, what they charge, roughly how many users they have, their key strengths, and — this is the useful bit — their key weaknesses.

For this market, two competitors were doing over $10 million in annual recurring revenue. That's a great signal. It tells you the market is real and people are paying. If the field were empty and the market were supposedly worth $4 billion, that would actually be a red flag.

One thing that jumped out: a competitor called OurFamilyWizard made a significant pricing change in early 2026 that's likely to push users out the door. That's an active acquisition window — a moment where people are already annoyed and looking for alternatives. That kind of intel would have taken a consultant weeks to surface.

Claude also identified where the market is crowded (communication features, shared calendars) versus where there's genuine white space. The white space here? Expense-only management. An app that had owned that niche — called Onward — shut down a few years back and nobody has filled the gap since. No existing competitor offers receipt scanning or auto-categorisation. That's the gap.

5.3 Validated Pain Points

Any idea you validate has to have a pain point that's real — not assumed. Claude's research pulled actual evidence of the problem:

• Co-parents tracking expenses manually through spreadsheets, Venmo, Zelle, text messages — fragmented and unverifiable
• Court-ordered payment disputes where one parent has no proof they covered a cost
• Apps that charge $50 for 24-hour access to your own financial records (genuinely wild)
• Rigid expense-splitting that doesn't support anything other than 50/50

These aren't things I assumed. These are documented frustrations from real users in the market. That's what "validated" means.

5.4 Monetisation Model

Claude recommended a three-tier model — free basic, a mid tier at around $120/year, and a premium tier with AI features. It also laid out unit economics: cost to acquire a customer, lifetime value per family, the LTV:CAC ratio.

You don't need to take Claude's monetisation recommendation as gospel. Use your professional judgment. But having a data-grounded starting point — rather than guessing — is exactly the point.

One thing that stood out to me: if the app ever got mandated by a family court in any US state, customer acquisition cost drops to zero. The court mandates it. That's an unusual GTM angle worth holding on to.

5.5 The Verdict

At the end of the report, Claude gave a clear recommendation. For this idea: very strong go.

The reasoning: large market, low penetration, validated pain points, two competitors proving people will pay, an active market disruption moment (the OurFamilyWizard pricing change), and a clear gap in the expense-only space.

That's what you're looking for at the end of your own report. Not a maybe — a clear signal, with evidence.

6. Saving the Research to Obsidian (~9:30)

Once the research is done, you need to save it somewhere you can reference it later. Everything we produce from this point forward — the PRD, the GTM document, the user journey, the specs — is going to be grounded in this research. Don't let it live only in the Claude chat window.

6.1 Setting Up the Filesystem Extension

Here's how this works. In Claude Desktop, go to Settings → Desktop App → Assistant Extensions. You should see an extension called Filesystem — if it's not installed, click Browse and install it. It's usually near the top of the list.

Once it's installed, click Configure and add the file path to your notes folder. For me, that's my Obsidian vault saved inside Dropbox. Whatever your path is, add it here. Claude needs to know where it's allowed to write files on your computer.

6.2 Saving the File with a Prompt

Go back to the chat where your research lives and type something like this:

Using the filesystem tool, save the research to [your file path]

Replace the path with the actual path to your notes folder — the same one you configured in the extension settings.

Claude will then:

1. Look at your file system
2. Navigate to the path you specified
3. Write a structured Markdown file with all seven sections from the research

This is the agents piece I keep mentioning. Claude isn't just answering a question — it's actually doing something: navigating your file system and writing a file. That's why the Desktop app and the Filesystem extension are required.

6.3 What You End Up With

After Claude saves the file, go open Obsidian (or whatever notes tool you're using). You'll find a new Markdown document with the full research output — structured, readable, with sections you can collapse and reference. This becomes the foundation everything else is built on.

I save mine inside my Dropbox so it's accessible and backed up. You can save yours wherever makes sense.

7. Try It Yourself

Exercise 1: Run Your Own Research Session

What to do: Open Claude Desktop, enable Deep Research, and run the market research prompt against an idea you have. Answer Claude's clarifying questions honestly and specifically. Let it run.

A nudge if you're stuck: If you don't have an idea yet, pick something you've complained about recently. A problem you've encountered yourself is often the best starting point — you already understand the pain.

How you'll know it's working: Claude will ask clarifying questions before it starts. If it just jumps straight into research without asking anything, check that Research mode is actually enabled — standard chat doesn't ask.

Exercise 2: Read the Output and Identify the Key Signals

What to do: Once the research comes back, find and note these five things before moving on: (1) total addressable market size, (2) top two or three competitors and what they charge, (3) the biggest gap or white space in the market, (4) the core validated pain point, (5) the go/no-go verdict.

What this is practising: Reading research critically — not just scanning it, but actually extracting what matters for a build decision.

Exercise 3: Save the Research to Your Notes

What to do: Use the Filesystem extension to have Claude write the research into your Obsidian vault (or your notes folder). Confirm the file is there and opens correctly.

How you'll know it's working: The file will appear in Obsidian, render as structured Markdown, and have all the sections from the research output.

8. You Should Be Able to Build This Now

Here's what you can put together with what we just covered:

• A market research report for any idea, grounded in real data, completed in under 20 minutes
• A competitive landscape overview that would have taken a team to produce manually

Check Yourself

• I ran Deep Research on my own idea (not just watched Dele run it on his)
• I know the difference between standard chat and Deep Research mode in Claude
• I can identify the five key things to look for in a research output
• My research document is saved in Obsidian (or my notes tool) and ready to reference
• I have a clear go or no-go signal for my idea

If Something's Not Working

The Short Version

Here's what I want you to walk away with:

• Deep Research is not the same as a Claude chat: It actively browses and synthesises. Enable it deliberately before every research session.
• Ask for clarifying questions first: That one line in the prompt makes the research dramatically more targeted. Don't skip it.
• You're looking for five things: Market size, competitive landscape, white space, validated pain points, and the go/no-go verdict.
• Save everything: The research document is the foundation. Module 2, Lesson 2 doesn't work without it.
• What you can do now: Validate any market idea in under 20 minutes with data you can actually trust.

Quick Reference

The Research Prompt Template

I have an idea for an app: [describe your idea in 2–3 sentences].

Please do deep research on the best way to approach this towards 
getting to a go / no-go decision for this app.

Important: before you start the research, ask me all the clarifying 
questions you need. Don't begin the research until I've answered them.

Clarifying Questions Claude Will Ask (expect these)

1. Target market — geography, language, specifics
2. Intent — what decision is this research informing?
3. Who are you — founder background, technical level, resources

Five Things to Find in the Output

1. Total addressable market size
2. Top competitors + what they charge
3. The gap or white space
4. Validated pain points (with evidence)
5. Go / no-go verdict + rationale

Enabling Deep Research

In the chat input → click (+) → enable Research mode
Look for the Research icon before hitting Enter

Resources

Tools Used

• Claude Desktop — the desktop app, required for this lesson
• Obsidian — the notes app used for saving the research output (any Markdown-compatible tool works)

Questions I Get Asked

Q: Can I use Perplexity instead of Claude Deep Research?

You can, and the output will be reasonably similar for the research phase. The reason I use Claude here is that we're going to stay in Claude Desktop for the next several phases — feeding the research into documentation, then into specs — so it makes sense to keep everything in the same tool. If you run the research in Perplexity, you'll need to paste the output manually into Claude for the next steps.

Q: The research came back as a no-go. What do I do?

Honestly? Be glad. That's the point of this step — to find out before you spend months building something. Go back to the idea stage, pick something else, and run the research again. A no-go in 15 minutes is infinitely better than a no-go after six months of work.

Q: My research is taking longer than 14 minutes. Is that normal?

Yes — research time varies depending on how complex the market is and how busy Claude's servers are at that moment. Give it up to 20–25 minutes before worrying. If it times out or errors, just run it again.

Q: Do I need to use Obsidian specifically?

No. Obsidian is my choice — it saves as Markdown, it's offline, and it works well with the Filesystem MCP. But any notes tool with a local file path will work. Notion local files, a plain folder on your desktop, whatever. The key is that the path you give Claude actually exists and is writable.

Q: How do I know I'm ready for the next lesson?

If you've run the research on your own idea, you can identify the five key outputs in the report, and your research document is saved and accessible — you're ready for Module 2, Lesson 2.

💬 Stuck? Come Talk to Us

Glossary

Deep Research: A specific mode in Claude that goes beyond standard chat — it actively browses multiple sources, synthesises findings, and produces a structured report. Different from a regular Claude conversation with web search enabled. (first used in Module 2, Lesson 1 — Market Research)

Total Addressable Market (TAM): The full potential revenue available if you captured every possible customer in your target market. Used as a ceiling to understand whether a market is worth entering — not a target you'll realistically hit. (first came up in Module 2, Lesson 1)

LTV:CAC Ratio: Lifetime Value to Customer Acquisition Cost. A unit economics metric — how much revenue a customer generates over their lifetime compared to what it costs to acquire them. You generally want this at 3:1 or better. (first came up in Module 2, Lesson 1)

Filesystem Extension (MCP): A Claude Desktop extension that lets Claude read from and write to your local file system. This is how Claude saves the research document directly into Obsidian without you having to copy-paste anything. (first used in Module 2, Lesson 1; we go much deeper on MCP in Module 2, Lesson 3 — User Journey Mapping)

Go/No-Go: The binary decision this research phase is designed to produce — is this idea worth investing further time and money in, or not? (introduced in Module 1, Lesson 2; first applied in Module 2, Lesson 1)