Research workflows

Claude is a useful research assistant for early-stage side projects: fast at synthesizing information, good at identifying patterns across multiple sources, and able to structure findings in whatever format you need. It’s not a replacement for primary research but it handles a lot of the preliminary work.

Competitor analysis

What works: Claude can survey what competing products exist, summarize their positioning, features, and pricing, and identify patterns. With Research mode on claude.ai, it can do this in a structured pass across multiple sources.

I'm building a multiplayer card game platform (HOKM, PASUR — Persian
card games). Survey the competitive landscape:

1. Name the top 5 multiplayer board/card game platforms (mobile + web)
2. For each: monetization model, real-time backend technology
   (if known), mobile app quality (App Store rating), primary audience
3. Identify any that target Middle Eastern / Persian-language markets
4. What's missing from all of them that a focused Persian card game
   platform could own?

Cite sources. Be specific — no generic "consider your competition" advice.

What doesn’t work: Real-time pricing data, app store rankings, recent funding news. Claude’s knowledge has a cutoff; for current data, it needs Research mode or you need to provide sources.

For technical evaluation (e.g., Nakama vs Colyseus): Claude is good at technical comparisons when you provide the dimension list:

Compare Nakama and Colyseus for a real-time multiplayer card game:
- Self-hosting vs managed hosting options
- Language for server-side logic (Go/TypeScript etc.)
- WebSocket support and reconnection handling
- Matchmaking primitives
- Community and maintenance status

I'm hosting on a VPS. My priority is real-time turn management
with bot substitution. I know Go and Dart.

Customer interview synthesis

If you’re doing customer discovery, Claude is good at synthesizing transcripts into findings. The workflow:

1. Conduct interviews yourself (5–10 people)
2. Paste transcripts (or notes) in batches
3. Ask Claude to extract patterns

These are notes from 8 interviews with Persian expats in North America
about how they stay connected with family culture. Extract:

1. The specific activities they miss most
2. What they currently use (if anything) for multiplayer card games
3. Friction points with existing solutions
4. Willingness to pay indicators (what they said or implied)

Here are the notes: [paste]

Claude handles the synthesis; you judge the patterns. Don’t ask Claude to draw conclusions you haven’t validated — use it to surface patterns you then verify.

ICP definition

Once you have interview data and market research, Claude is useful for drafting an ICP (Ideal Customer Profile):

Based on the interview synthesis and competitor research above,
draft an ICP for the HOKM platform. Include:

- Demographic and geographic specifics (not generic "tech-savvy")
- The specific pain they have that we're solving
- What they're currently doing instead
- How they'd describe the value in their own words
- One sentence: why this person buys vs. why they don't

Be specific. If the data doesn't support a claim, say so.

The “if the data doesn’t support a claim, say so” instruction is important — it stops Claude from inventing specifics when you haven’t provided them.

What Claude is bad at for research

• Primary source verification — Claude can’t call people, run surveys, or access paywalled data
• Real-time market data — pricing, rankings, recent news without Research mode
• Knowing your specific users — it can synthesize research you give it but can’t replace talking to users
• Predicting market outcomes — it can frame risks but not tell you if your idea will work