The complete startup research guide: how to validate your idea before building in 2026

TL;DR

• 42% of startups fail because there's no market need for what they built - they skipped research
• Research ≠ validation (most founders confuse these two critical phases) - deep dive: research vs validation
• Proper research takes 1-2 weeks with Perplexity and structured prompts - not months of analysis paralysis
• The 6 research categories you must cover: ICP, pain points, jobs-to-be-done, competition, market size, and GTM
• Research creates hypotheses to test; validation tests those hypotheses with real users
• Free download: Complete Startup Research Prompt Library (50+ prompts for Perplexity/ChatGPT)

Why 42% of startups build something nobody wants

CB Insights analysed 483 startup post-mortems (CB Insights) and found something brutal: 42% of startups fail because there is no market need for their product or service (RevliEdition). Not because the product was bad. Not because the team was incompetent. Because they built something nobody wanted.

Here's what actually happens. Founder gets excited about an idea. Maybe they experienced the problem themselves, or a friend complained about it, or they saw a gap in the market. They immediately jump to solutions. "I'll build an app that does X!" They spend 3-6 months building. They launch. Crickets. The problem they thought existed either wasn't painful enough for people to pay for a solution, or it was painful but their solution wasn't the right one, or the problem only existed for 12 people in the world and none of them have money.

Many startups prioritize building their product over deeply understanding their audience, relying on assumptions rather than verified insights (Edition). This results in solutions that fail to resonate with customers, wasting time, money, and effort that could've been saved with two weeks of proper research.

And here's the thing that kills me: 70% of startups fail between years 2-5 (DemandSage). They survive the initial launch, get some traction, maybe even raise money, and then die slowly because the foundation was wrong from the start. They built on sand instead of rock because they skipped the research phase.

I've built 20+ apps now. The ones that worked all started with deep research. The ones that struggled? We either skipped research entirely or did surface-level "I asked 5 friends" research that confirmed our biases instead of challenging our assumptions.

Micro-takeaway: 90% of startups fail eventually (Revli), but the 10% that succeed almost always start with understanding the market before building anything.

Research vs validation: the confusion killing MVPs

Most founders think research and validation are the same thing. They're not. Confusing them is like confusing dating and marriage - both involve commitment, but they're fundamentally different stages with different goals.

Research answers: Is there a market? Who are they? What problems do they have? How do they solve them now? Who else is solving this? How big is the opportunity? What's my go-to-market strategy?Validation answers: Will people in my target market actually want MY specific solution? Will they sign up for a waitlist? Will they complete a survey telling me more about their needs? Will they agree to interviews? Will they eventually pay for this?

Research happens BEFORE you have a solution. Validation happens AFTER you have a hypothesis but BEFORE you build the full product.

Here's the sequence that actually works:

Phase 1 - Research (1 day): Deep dive into the problem space. You're not validating YOUR idea yet. You're understanding the entire landscape. Who has this problem? How painful is it? How do they solve it today? What do they hate about current solutions? Who are the competitors (direct and indirect)? What does the market look like? This phase ends with clear hypotheses: "I believe [ICP] has [problem] because [reason] and they would pay for [solution] because [value proposition]."Phase 2 - Validation (1-2 weeks): Now you test those hypotheses. You build a landing page that describes your solution. You drive your target ICP to that page through marketing channels. You measure conversion rate. You ask them to complete a survey. You invite them to interviews. This phase ends with data: "We got 50 signups at 8% conversion rate, 30 survey responses, and 5 interview bookings. The data shows our hypothesis about [X] was correct but our assumption about [Y] was wrong."Phase 3 - Build MVP (4-6 weeks): Only NOW do you build the actual product - lean MVP - first version, informed by both research insights and validation data. In order to stay smart make use of new no-code tools for building products 5-8x faster than custom code - the way we build applications for founders.

Most founders skip Phase 1 entirely and go straight to Phase 2. Or worse, they skip both and jump to Phase 3. Then they wonder why nobody uses their product.

Micro-takeaway: Research tells you WHAT to validate. Validation tells you WHETHER to build. Building without both is gambling with your runway.

What happened when I spent one full day researching Test AI Models

I recently launched Test AI Models (our tool for comparing LLM performance, speed, and cost). Before writing a single line of code, I spent one full day on deep research using Perplexity and Claude. Here's what happened and what I learned.

Morning: ICP and pain points research

I started with what I thought I knew: "Developers building AI features need to compare different LLMs." That was my assumption. So I asked Perplexity to search Reddit, developer forums, and review sites for complaints about choosing AI models.

- First surprise: The pain wasn't just "I don't know which model to choose." The real pain was "I integrated GPT-4, got an unexpected $3,000 API bill, and now my CFO is asking questions." API cost spikes were a MAJOR driver I hadn't considered. People weren't just losing hours on integration - they were losing money on wrong model choices. - Second surprise: I was targeting solo developers and startups. But the research showed agencies were even better suited for this tool. Agencies work on multiple client projects simultaneously, each potentially needing different LLMs. They have more pain points per product than a startup with one product. - Third surprise: AI agent builders and agentic AI developers were the power users I should target first. They're not just using one LLM per product - they're using multiple LLMs within a single product, each for different tasks. Their pain is 10x what a regular developer experiences.

By lunchtime, my entire ICP had shifted. I went from "solo developers building AI features" to "AI agent builders, agencies, and agentic AI developers dealing with multi-LLM workflows and unpredictable costs."

Afternoon: Competition and market analysis

I searched for direct competitors (other LLM comparison tools) and found a few.

But then I dug into INDIRECT competitors - how are people solving this problem TODAY without a dedicated tool? The answer: Excel spreadsheets. Manual testing across multiple platforms. Asking on Reddit. Guessing based on marketing materials. Or just picking ChatGPT because it's the one they've heard of. This was valuable because it told me two things: First, the problem is real enough that people are creating manual solutions. Second, the current solutions are painful enough that there's room for a better tool.

I also found out how users were actually talking about the problem. The language they used. The specific complaints. The features they wished existed. This became my marketing copy later - I literally used their words because those are the words that resonate.

Evening: GTM strategy and hypothesis formation

Based on everything I learned, I created my go-to-market plan. Not what I WANTED to do, but what the RESEARCH told me would work.

- Primary channels: Reddit (where developers complain about API costs), X/Twitter (where AI builders share their work), and LinkedIn (where agencies discuss client work). - Primary message: Not "compare LLM performance" but "stop wasting money on the wrong AI model." - Primary ICP: AI agent builders first (power users who will push the product and give best feedback), then expand to agencies, then solo developers.

The hypothesis I'd test in validation: "AI agent builders experiencing API cost issues will sign up for a tool that lets them test multiple LLMs simultaneously without separate API setups, with at least 5% conversion rate from landing page to signup."

Total time investment: 8 hours of focused research. Total cost: $0 (Perplexity free tier + Claude already using on subscription). Value: Completely reshaped my product positioning, ICP, and GTM strategy before spending a dollar on development.Micro-takeaway: One day of research saved me months of building for the wrong audience. The ROI on research is insane when you actually do it properly.

The 6 research categories you can't skip

Based on 15+ app builds and working with dozens of founders, these are the six research categories that separate successful launches from expensive failures.

1. Ideal Customer Profile (ICP)

This isn't "who might use this." It's "who has this problem so badly they'll pay to solve it within 30 days of discovering you exist."

Questions to answer:

• Who specifically experiences this problem? (Job title, industry, company size, geography)
• What's their day-to-day reality? (Tools they use, workflows they follow, constraints they face)
• What's their decision-making process? (Do they have budget authority? Do they need approval?)
• Where do they hang out online? (For B2B: LinkedIn, X, industry Slack groups. For B2C: Instagram, TikTok, Facebook)

Why this matters: I've seen founders build amazing products for "small business owners" and fail because "small business owners" is 33 million different people with completely different problems. But "solo construction contractors in Texas who manually track job costs in Excel" is 50,000 people with the same painful problem you can actually reach.

2. Pain points and current solutions

Questions to answer:

• What exactly is the problem? (Not what you think - what THEY say)
• How painful is it on a scale of "annoying" to "keeps me up at night"?
• How do they solve it today? (The real answer, not the ideal answer)
• What do they hate about current solutions?
• What would make them switch to something new?

Where to find this: Reddit is gold for complaints. People don't go to Reddit to say "everything is working great." They go to complain, ask for help, and share frustrations. Search your problem space on Reddit and read 100+ threads. The patterns will emerge. Review sites like G2, Capterra, TrustPilot - read the 1-star and 2-star reviews of competitors. People tell you EXACTLY what they hate.Why this matters: If the current solutions are "good enough," you're building a painkiller for a headache people tolerate. You need to find problems people are actively trying to solve but failing.

3. Jobs-to-be-done (JTBD)

This isn't about features. It's about what job the user is hiring your product to do.

Questions to answer:

• What are they trying to accomplish? (The outcome, not the task)
• What progress are they trying to make in their life/work?
• What triggers them to look for a solution?
• What does success look like from their perspective?

Example: People don't hire Uber to "get a ride." They hire Uber to "get somewhere without the hassle of parking, or when they've been drinking, or when they're in an unfamiliar city." The job varies by context. For Test AI Models, the job isn't "compare LLMs." The job is "choose the right AI model for my specific use case without wasting money testing all of them separately."Why this matters: Understanding the job helps you position your product correctly. You're not selling features. You're selling outcomes and progress toward a goal.

4. Competition (direct and indirect)

Most founders only look at direct competitors. That's a mistake.

Questions to answer:

• Who else solves this exact problem? (Direct competitors)
• Who solves related problems or overlapping problems? (Indirect competitors)
• What are people using TODAY before they find better solutions? (Excel, manual processes, free tools)
• What are competitors' strengths and weaknesses?
• Why do customers choose them or leave them?
• What market gaps exist that nobody is serving well?

Why this matters: Indirect competition is often more dangerous than direct. Notion didn't kill Evernote by being a better note-taking app - they redefined the category entirely. Understanding the full competitive landscape helps you position differently. Also, analysing competition tells you what's already been validated. If three well-funded companies are attacking the same space, the problem is real. Your job is differentiation, not validation.

5. Market size and opportunity

Questions to answer:

• How many people/companies have this problem? (TAM - Total Addressable Market)
• How many could you realistically reach? (SAM - Serviceable Addressable Market)
• How many could you acquire in Year 1? (SOM - Serviceable Obtainable Market)
• Is this market growing, shrinking, or stable?
• What's the typical price point customers pay for solutions in this space?
• What's your potential revenue if you capture X% of the market?

Why this matters: You need to know if you're building a $10M business or a $100M business. Both are valid, but they require different strategies and investor expectations. Also, some markets sound huge but are impossible to penetrate. "Everyone needs email" is true but doesn't help you compete with Gmail. Finding a big ENOUGH market that you can actually reach is the goal.

6. Go-to-market (GTM) strategy

Questions to answer:

• Where does your ICP hang out online and offline?
• What channels work for reaching them? (Paid ads, content, partnerships, communities, sales)
• What messaging resonates based on your research?
• What's the customer acquisition cost (CAC) in this space?
• What's the typical sales cycle length?
• Who are the influencers or champions in this space?

Why this matters: You can build a perfect product for a real problem and still fail because you can't efficiently reach your customers. GTM research during the research phase prevents you from discovering this AFTER you've built the product. For B2B products, GTM is often LinkedIn + content + sales outreach. For B2C, it's often social media + paid ads + influencer partnerships. But specifics matter. Selling to developers? Reddit and X work. Selling to executives? LinkedIn and conferences work. Research tells you which specific channels will work for YOUR ICP.Micro-takeaway: Skip any of these six categories and you're building with blind spots that will kill you later. Do all six and you have a complete picture of the market before you build anything.

How to actually do research (the Perplexity framework)

Everyone says "do research" but nobody explains HOW. Here's my actual framework using Perplexity.

Why Perplexity over ChatGPT or Claude?

Perplexity is specifically built for research. It cites sources automatically, searches Reddit and review sites, and gives you the most recent information available. Claude is better for ideation and analysis, but Perplexity is better for "tell me what actually exists and what people actually say." Also, Perplexity's citations mean you can verify everything. When it says "developers complain about API costs on Reddit," it links to the actual Reddit threads. You can read them yourself and get even more context.

The structured prompt approach

Don't just ask vague questions like "tell me about the AI market." That gets you generic summaries. Use structured prompts that force specific, actionable answers.

Example prompt for ICP research:

Research where developers who are choosing AI models and building with LLMs actively participate online. I need specific communities, platforms, and spaces where they seek advice, share experiences, and discuss AI model selection.

For each platform/community found, provide:
- Platform name and URL
- Member count / monthly active users  
- Engagement level (high/medium/low)
- Content that performs well
- Self-promotion rules
- Key influencers or active moderators
- Best times to post
- Examples of successful posts about AI tools

Focus on Reddit communities, X/Twitter hashtags, Discord servers, and Slack communities where my ICP congregates.

This prompt is specific enough to get actionable answers but open enough to discover things you didn't know existed.

Example prompt for pain point research:

Search Reddit, review sites, and developer forums for complaints and frustrations related to choosing and integrating AI models (ChatGPT, Claude, Gemini, etc.) in applications.

I want to understand:
- What specific problems do developers mention?
- What language and phrases do they use to describe their pain?
- What triggers them to complain or seek help?
- What solutions have they tried that didn't work?
- What would their ideal solution look like?

Prioritize recent discussions (last 12 months) and include direct quotes with sources.

See the difference? You're not asking Perplexity to summarize. You're asking it to find specific evidence and patterns.

My complete research process (full day)

Morning: ICP + Pain Points

• Run 3-5 prompts about your target customer
• Run 3-5 prompts about pain points and complaints
• Read the actual sources Perplexity links to (don't just read the summary)
• Take notes on patterns, unexpected insights, language people use

Noon: Competition + Market

• Run 3-5 prompts about direct competitors
• Run 3-5 prompts about indirect competitors and current solutions
• Run 2-3 prompts about market size and trends
• Create a simple competitive matrix (who does what, strengths, weaknesses)

Afternoon: JTBD + GTM

• Run 3-5 prompts about jobs-to-be-done and customer outcomes
• Run 3-5 prompts about marketing channels and communities
• Run 2-3 prompts about pricing and business models in your space
• Map out potential GTM strategies based on findings

Evening: Synthesis + Hypothesis Creation

• Review all your notes and findings
• Identify the 3-5 biggest insights that surprised you
• Document your hypotheses to test in validation
• Do your Problem Reality Check (more on this below)
• Build your validation plan based on research insights

Total time: 6-8 hours in one day Total cost: $0 if using Perplexity free tier, $20/month for Pro if you want unlimited searches Output: A complete understanding of your market, clear hypotheses to test, and a validation plan

The non-negotiable rules

Rule 1: Make it as unbiased as possible. Don't ask "Why do developers love AI model comparison tools?" Ask "What do developers say about choosing AI models?" The first question assumes they love comparison tools. The second discovers what they actually think.Rule 2: Read the sources, not just summaries. Perplexity gives you links for a reason. Click through to Reddit threads and read 20-30 comments. The nuance matters.Rule 3: Look for patterns, not one-off complaints. One person complaining about API costs might be an edge case. Twenty people complaining about API costs is a pattern worth exploring.Rule 4: Document everything. Use Notion or Google Docs to capture your findings. You'll reference this during validation and marketing.Rule 5: Be willing to be wrong. If research contradicts your assumptions, trust the research. Your gut isn't better than data from 50 people in your target market.Micro-takeaway: Research isn't rocket science. It's systematic curiosity with structured prompts and honest analysis of what you find.

When research is "done" and validation begins

One of the most common questions I get: "How do I know when I've researched enough and should move to validation?" Here's the honest answer: You're done with research when you can confidently answer these questions without guessing.

The research completion checklist

✓ Can you describe your ICP in one sentence? Good: "AI agent builders who manage multiple LLM integrations and have experienced unexpected API cost spikes" Bad: "Developers who use AI"✓ Can you list 3-5 specific pain points with evidence? Each pain point should have:

• Clear description of the problem
• Evidence from research (Reddit threads, reviews, interviews)
• Explanation of how people currently solve it (or fail to)
• Assessment of pain level (annoying vs. critical)

✓ Can you name your top 3-5 competitors and explain their positioning? For each competitor, you should know:

• What they do well
• What customers complain about
• Who their target customer is
• How you'd differentiate from them

✓ Can you explain the market opportunity? You should have rough numbers for:

• How many people have this problem (TAM)
• How many you can realistically reach (SAM)
• What they pay for solutions today
• Whether market is growing, stable, or shrinking

✓ Do you have documented hypotheses to test? Format: "I believe [ICP] has [problem] because [reason] and would pay [price] for [solution] because [value proposition]." Example: "I believe AI agent builders have the problem of wasting hours testing multiple LLMs separately and experiencing unpredictable API costs because they're building complex workflows with multiple models. They would pay $29/month for a tool that lets them test all major LLMs simultaneously with clear cost breakdowns because it would save them 5+ hours weekly and prevent cost surprises."✓ Do you have a validation plan? You should know:

• What landing page you'll create
• What survey questions you'll ask
• What channels you'll use to drive traffic
• What success metrics you're targeting
• What would make you proceed vs. pivot

Signals you're done vs. signals you're stuck

Good signals you're done:

• You've found patterns across multiple sources
• You have evidence for your hypotheses
• You can explain your market to a stranger in 2 minutes
• You've discovered things that surprised you
• You know where your ICP hangs out online
• You feel confident about what to test in validation

Bad signals you're in analysis paralysis:

• You're reading the 50th article on market trends
• You're comparing market size estimates from 10 different sources
• You're researching sub-topics that don't affect your core hypotheses
• You're avoiding moving to validation because research feels safer
• You can't articulate what you're still trying to learn

How to break analysis paralysis: Set a deadline. Research for one day max. On day 2, you move to validation whether you feel "ready" or not. Perfect information doesn't exist. Good enough information plus real-world testing beats perfect research that never launches.

The Problem Reality Check tool

At Eterna, we created a simple tool called Problem Reality Check. It's a questionnaire where founders input what they know about their problem, market, and solution. The tool gives them a score from 1-10 and tells them what they still need to research.

How it works:

• Answer 5 questions about your problem, market, and solution
• Each answer is scored based on specificity and evidence
• Tool identifies gaps in your knowledge
• Suggests specific research areas to focus on
• Gives you a "research readiness score"

If you score below 7/10, you're not done with research. Keep digging until you hit 7-8/10, then move to validation.

Want the tool? It's still in BETA and free, but you can use it after signing up here:Problem Reality Check ToolMicro-takeaway: Research is done when you have enough information to create falsifiable hypotheses and a concrete validation plan. Not when you know everything - when you know ENOUGH.

Common research mistakes founders make

I've made every one of these mistakes. I've also watched dozens of founders make them. Here are the five research mistakes that kill startups before they even get to validation.

Mistake 1: Asking friends and family instead of the actual market

What it looks like: "I asked my wife/coworker/friend and they said it's a great idea! I should definitely build this." Why it's deadly: Your friends want to support you. They'll say nice things. They're not representative of your actual market. They won't tell you the brutal truth that your idea is solving a problem nobody has. Real example: Founder I worked with built a "social network for dog owners" because all his friends with dogs said they'd use it. Spent $8K on development. Launched. His friends signed up. Nobody else did. Why? Turns out Instagram + dog hashtags already solved everything dog owners needed. The research would've shown him this in one day. The fix: Talk to strangers in your target market. Not people who love you. People who will brutally tell you if your idea sucks. Reddit, X, LinkedIn - find your ICP and ask them questions. Their honesty is more valuable than your friend's support.

Mistake 2: Only researching direct competitors, ignoring indirect ones

What it looks like: "There are only 2 other apps doing exactly what I'm planning, so there's room in the market!" Why it's deadly: You're not competing with direct competitors only. You're competing with indirect competitors, manual workflows, and "doing nothing." Real example: With Test AI Models, my direct competitors were other LLM comparison tools. But my REAL competition was: developers manually testing each LLM separately, developers just picking ChatGPT by default, developers asking on Reddit, and developers not testing at all and hoping for the best. If I'd only looked at direct competitors, I would've missed 80% of my actual competition and failed to position correctly. The fix: Research three layers of competition:

1. Direct: Who does exactly what you're planning?
2. Indirect: Who solves the same problem differently?
3. Substitute: What do people do TODAY before finding a solution?

All three matter. Sometimes substitute competition (manual Excel tracking) is harder to beat than direct competition.

Mistake 3: Not adding kill-switch criteria at the beginning

What it looks like: "I'll do research, but I'm building this regardless of what I find." Why it's deadly: If you're going to build no matter what research tells you, you're wasting time on research. Research only helps if you're willing to pivot, adjust, or kill based on what you learn. Real example: Founder came to me with idea for "Uber for lawn mowing." Research showed: lawn mowing is highly seasonal, low margin, requires scheduling weeks in advance (not on-demand), and local companies with established crews dominate. Every piece of research said "don't build this." Founder built it anyway. Failed within 6 months. The fix: Before you start research, define your kill-switch criteria. "If I discover X, I will not build this. If I discover Y, I will pivot to Z instead." Example criteria:

• If fewer than 10,000 people have this problem → kill
• If current solutions score above 8/10 satisfaction → kill
• If CAC is higher than $500 for a $29/month product → kill
• If I can't identify 3 distribution channels → kill

Having criteria in advance prevents emotional attachment from overriding data.

Mistake 4: Confusing interesting trends with real problems

What it looks like: "AI is hot right now, so any AI product will succeed!" Why it's deadly: Trends don't equal viable businesses. Crypto was hot. NFTs were hot. Metaverse was hot. Lots of startups died chasing trends instead of solving real problems. Real example: 2021-2022 saw hundreds of "ChatGPT wrapper" startups. Most failed. Why? They were chasing the AI trend instead of solving specific painful problems. The few that succeeded solved real problems (Jasper for marketers needing content, Copy.ai for specific copywriting tasks) not just "AI for everyone." The fix: For every trend you identify, ask: "What specific problem does this trend create or solve for my ICP?" If you can't articulate a clear problem with evidence, you're trend-chasing, not problem-solving. Trends are useful for understanding "why now" but they don't replace problem validation.

Mistake 5: Spending 3 weeks on research instead of one day

What it looks like: "I want to be thorough, so I'm going to research for a few months before I make any decisions." Why it's deadly: Analysis paralysis kills momentum. The goal isn't perfect information - it's ENOUGH information to make smart decisions and test hypotheses. Research that doesn't lead to action is procrastination wearing a business suit. Real example: One founder spent 4 months "researching" the project management space. Read 100 blog posts, analyzed 50 competitors, created detailed market reports. By the time they were ready to build, two new competitors had launched and captured the exact positioning they wanted. The fix: Set a hard deadline. One day for research, max. Focus on the six categories I outlined earlier. Get good-enough answers, create hypotheses, move to validation. You'll learn more from one day of real structured research than from one month of scattered searches. Research is about reducing uncertainty, not eliminating it. You can't eliminate it.Micro-takeaway: Research mistakes are expensive because they waste time and money on the wrong path. Avoid these five and you'll save months of wasted effort.

The complete research-to-launch sequence

Let me show you the complete sequence from idea to launch, with research as the critical first step.

Phase 1: Research (1 day)

Goal: Understand the market, problem, and opportunity before proposing any solution Activities:

• Deep research on ICP using Perplexity and structured prompts
• Pain point discovery through Reddit, reviews, forums
• Competitive analysis (direct, indirect, substitute)
• Market size estimation
• Jobs-to-be-done research
• GTM channel identification

Deliverables:

• ICP definition document
• Pain points list with evidence
• Competitive landscape map
• Market opportunity assessment
• Documented hypotheses to test
• Validation plan with success criteria

Success criteria:

• Can explain market in 2 minutes to stranger
• Have evidence for all major claims
• Know where ICP hangs out online
• Score 7+/10 on Problem Reality Check
• Have 3-5 falsifiable hypotheses

Decision gate: Do we have enough information to test hypotheses? If yes → Validation. If no → Research another day.

Phase 2: Validation (1-2 weeks)

Goal: Test hypotheses with real users before building the product Activities:

• Create landing page describing the solution
• Write survey with 5 key questions (like we did with Test AI Models: role, team size, use case, building for, goals)
• Set up email capture + survey flow
• Drive traffic through identified channels (for Test AI Models: Reddit, X, LinkedIn)
• Conduct user interviews with survey respondents
• Analyse conversion rates and feedback

Deliverables:

• Live landing page with email capture
• Survey results from 50-300 people (B2B needs 50+, B2C needs 100+)
• 5+ user interview transcripts
• Conversion rate data
• Updated hypotheses based on real data
• Go/no-go decision with evidence

Success criteria for B2B:

• 50+ waitlist signups minimum (100+ ideal)
• 2-10% conversion rate from visit to signup
• 30%+ survey completion rate
• 5+ people agree to interviews
• Clear evidence hypothesis was correct (or clear evidence to pivot)

Success criteria for B2C:

• 100+ waitlist signups minimum (300+ ideal)
• 5-15% conversion rate
• 40%+ survey completion rate
• Consistent feedback themes across responses
• Product/market fit signals in qualitative feedback

Decision gate: Did validation confirm our hypotheses? If yes → Build. If partially → Pivot and re-validate. If no → Kill or completely rethink.

Phase 3: Build (4-8 weeks)

Goal: Create the minimum viable product that solves the core problem for your validated ICP Activities:

• Build core features only (resist feature creep)
• Weekly demos with beta users from waitlist
• Iterate based on feedback
• Focus on solving the main job-to-be-done
• Prepare for launch (marketing site, documentation, onboarding)

Deliverables:

• Working MVP with core features
• Beta user feedback integrated
• Launch plan ready
• Marketing materials prepared
• Analytics tracking set up

Success criteria:

• MVP solves the core problem validated in Phase 2
• Beta users confirm it's valuable
• No major blockers to launch
• Team confident in public release

Decision gate: Is the MVP good enough to launch? Not perfect - good enough to get feedback from real paying users.

Phase 4: Launch & Iterate (ongoing)

Goal: Get users, collect feedback, iterate, grow Activities:

• Public launch on validated channels
• Monitor analytics and user behavior
• Conduct user interviews
• Prioritize features based on usage data
• Iterate quickly on feedback
• Build in public and share learnings

Deliverables:

• Growing user base
• Product improvements based on real usage
• Clear understanding of what works vs. doesn't
• Roadmap informed by data
• Path to revenue/growth

Success criteria:

• Users actually using the product (not just signing up)
• Retention rates meeting benchmarks (40%+ Month-1 for B2B, 25%+ for B2C)
• Clear product/market fit signals
• Users referring others or requesting features
• Path to sustainable growth visible

The key insight: Each phase informs the next

Research tells you WHAT to validate. Validation tells you WHETHER to build. Building creates the product to test at scale. Launch tells you what to iterate on.

Most founders skip Phase 1 and 2. They jump straight to Phase 3. Then they wonder why Phase 4 doesn't work.

Real example - Razmeni (our marketplace): Phase 1 (Research): Discovered parents need to exchange baby/kid items but existing solutions were too complex or scammy. Found our ICP: parents of kids 0-5 years in Serbia. Phase 2 (Validation): Created landing page and survey, asked parents to send to friends, got 300 signups. Survey revealed they wanted simple, trusted marketplace with verification. This shaped our MVP scope. Phase 3 (Build): Built basic marketplace with verification, categories for kids items, and trust features. Phase 4 (Launch): Launched to waitlist first, got 1,000+ users in first 3 months. Iterated based on usage patterns.Micro-takeaway: This sequence works for online business, offline business, B2B, B2C, products, services - everything. The specifics change but the sequence stays the same.

Your research action plan starts now

Research isn't glamorous. It doesn't feel like progress the way building features does. But 42% of startups fail because there is no market need for their product (RevliEdition) - and research is what tells you whether market need exists BEFORE you waste months building.

Here's what to do next: If you're pre-idea: Download the Startup Research Prompt Library (50+ prompts for Perplexity) and start exploring problem spaces that interest you. Research before you commit to a specific idea. If you have an idea but haven't researched: Block out a day this week. Follow the framework in this post. Use Perplexity with structured prompts. Cover all six research categories. Create your hypotheses. Then move to validation. If you've already built something: Do the research now, even though it's late. You might discover you need to pivot your positioning, change your ICP, or adjust your features. Better late than never. If you're about to hire someone to build your MVP: STOP. Do research and validation first. I've seen founders spend $50K+ building something nobody wants. Save that money. Spend 2 weeks and $0 on research instead.

The founders who win aren't the ones with the best ideas. They're the ones who understand their market better than anyone else before they build. Research is your competitive advantage.

Want the complete toolkit?

🎁 Free Download: Startup Research Prompt Library 50+ structured prompts for Perplexity/Claude organized by research category. Use these to run your own research phase in one day. Link: Download Startup Research Prompt Library

🔧 Free Tool: Problem Reality Check Score your research readiness and identify gaps in your knowledge before moving to validation. Link: Problem Reality Check Tool

📅 Want us to do it for you? Book free strategy session We did this many times and we can do it faster and better - if you don't want to invest the time. Link: Book 15 min strategy session here