Opportunity Discovery: A Practical Guide for When You're Stuck Looking for Problems Worth Solving
The Problem in Concrete Terms
You're sitting at your desk with a notebook open, or staring at a blank Notion page, or walking the dog trying to force an insight. You want to build something—product, service, business, side project—but every idea feels either obvious and crowded, or vague and unvalidated. You've read "scratch your own itch" but your itches are either too niche (nobody else has this problem) or too generic (everyone has this problem, and Google already solves it). You've tried "look for complaints on Reddit" but the signal-to-noise ratio is terrible. You've tried "talk to customers" but you don't know who they are yet because you don't know what you're building. The circular dependency is the trap: you need a problem to find customers, but you need customers to validate the problem.
Concretely, you're experiencing one or more of these:
- Idea paralysis: 47 half-baked concepts in your notes, none developed past a paragraph
- False positives: Ideas that seem great until you spend 20 minutes researching and find three well-funded competitors
- Validation theater: Running surveys or landing pages for ideas you don't actually care about, getting meaningless data
- Scope creep in reverse: Starting with a real observation, then abstracting it until it's a "platform for X" with no entry point
- Motivation decay: The longer you search, the less you believe any opportunity is real, so you stop looking
Why the Usual Fixes Fail
1. "Scratch your own itch" assumes your itch is representative. Most builders have atypical problems. You're technical, early-adopter, optimization-oriented. The mass market isn't. Building for yourself often yields tools for developers—fine if that's your market, dead end if it's not.
2. "Talk to customers" presumes you have a hypothesis. Customer discovery without a directed hypothesis is just chatting. You'll hear polite encouragement for bad ideas and miss the uncomfortable truths that signal real pain. You need a specific, falsifiable claim before a conversation becomes useful.
3. "Look at trending topics / Product Hunt / Twitter" finds saturated markets. By the time an opportunity is visible in public discourse, the easy entry points are taken. You're seeing the survivors, not the gaps. The real opportunities look boring or confusing from the outside.
4. "Build an MVP and iterate" burns months on the wrong problem. Iteration optimizes a solution. It doesn't tell you if the problem exists, if people will pay, or if you're the right person to solve it. You can iterate a bad idea into a polished bad idea.
5. Frameworks (Jobs-to-be-Done, Value Proposition Canvas, etc.) become procrastination tools. Filling out canvases feels like work. It produces artifacts, not evidence. The map is not the territory.
Step-by-Step Solution
Phase 1: Collect Raw Observations (Week 1-2)
Action: Keep a "friction log" for 14 days. Every time you notice yourself or someone else:
- Manually copying data between tools
- Complaining about a process ("why do I have to...", "it's so annoying that...")
- Building a workaround (spreadsheet, script, Notion template, Zapier chain)
- Paying for a tool but using 10% of its features
- Saying "there should be a better way" without naming it
Constraint: Write it down immediately. One line per observation. No judgment, no ideation, no research. Just capture. Aim for 50+ entries.
Why this works: You're not hunting for ideas. You're documenting reality. Your own friction is biased (see above), but other people's friction observed in the wild is data. The workaround is the strongest signal—someone already values the outcome enough to build a bad version of the solution.
Where to look beyond yourself:
- Support forums for tools you use (filter for "feature request" and "workaround")
- Reddit threads asking "how do you handle X?"
- Sales calls or customer success tickets at your current job (anonymized)
- Industry-specific Facebook groups, Discord servers, Slack communities
- Government procurement notices, RFPs, compliance mandates (boring = underserved)
Phase 2: Cluster and Filter (Week 2-3)
Action: Group observations by workflow, not by tool or industry. Example clusters:
- "Getting client feedback on creative assets"
- "Onboarding new employees to internal systems"
- "Reconciling bank transactions across accounts"
- "Scheduling recurring meetings across time zones with external parties"
Filter criteria (apply ruthlessly):
- Frequency: Does this happen daily/weekly for the person? Monthly is too low.
- Consequence: What happens if it goes wrong? Lost revenue, compliance risk, client churn, missed deadline = high consequence. Mild annoyance = discard.
- Workaround exists: Someone has built a spreadsheet, script, or manual process. This proves budget (time/money) is already allocated.
- No dominant incumbent: Search "[workflow] software" and "[workflow] tool". If the top 3 results are established companies with >50 employees and >5 years in market, discard. You want fragmented, ugly, or missing solutions.
- You can reach the buyer: You know 3+ people who do this workflow, or you can identify a conference/forum/list where they congregate.
Output: 3-5 opportunity areas. Each described as: Who does what, how often, with what current workaround, and what breaks.
Phase 3: Validate the Problem, Not the Solution (Week 3-5)
Action: Run 10 problem interviews per opportunity area. Not solution interviews. Do not show mockups. Do not pitch.
Recruitment script: "I'm researching how [role] handles [workflow]. Not selling anything. 20 minutes. I'll share what I learn across all interviews."
Interview script (exact order):
- "Walk me through the last time you did [workflow]." (Listen for steps, tools, people involved)
- "What's the hardest part?" (Don't lead. Wait.)
- "How do you handle that today?" (Workaround = gold)
- "What happens if [hardest part] goes wrong?" (Consequence)
- "Have you looked for tools to help?" (Reveals awareness and budget)
- "What would make this 10x better?" (Their language, not yours)
- "Who else should I talk to?" (Snowball sampling)
Record and transcribe. (Otter.ai, Fathom, or manual notes.)
Analysis: After 10 interviews, you'll have:
- A ranked list of pain points (frequency × severity)
- The exact vocabulary buyers use (use this in copy later)
- A list of existing workarounds and their failure modes
- Names of 20+ more people to talk to
Kill criteria: If fewer than 6/10 interviewees describe the same acute pain with an active workaround, drop the opportunity area. Do not "pivot the idea." Drop it. Return to your filtered list.
Phase 4: Define the Wedge (Week 5-6)
Action: Pick one opportunity area. Define the smallest slice that:
- Solves the #1 ranked pain point from interviews
- Can be built in 2-3 weeks by you (or 1 week if you're non-technical using no-code)
- Requires zero behavior change from the user (fits into existing workflow)
- Has a clear "done" state you can demo
Example:
- Opportunity: "Agencies collecting client feedback on video edits"
- Pain: "Clients send timestamped notes via email/Slack/text, editors miss them, revisions loop 5x"
- Workaround: "Project manager manually consolidates into a spreadsheet"
- Wedge: "A single link the agency sends. Client clicks, watches video, clicks timestamp to add note. All notes auto-sorted by timestamp. Editor gets one clean list."
Not a wedge: "Collaborative video review platform with versioning, annotations, approval workflows, and client portal." That's a product. Build the wedge first.
Phase 5: Pre-Sell the Wedge (Week 6-8)
Action: Get 3-5 letters of intent (LOIs) or paid pilots before writing production code.
LOI template (one page):
"We, [Company], confirm interest in purchasing [Wedge Description] for $[price]/month per seat. We understand this is in development with estimated delivery [date]. This is non-binding but represents genuine purchase intent contingent on the solution meeting the following criteria: [3-5 bullet points from interview criteria]."
How to ask: At the end of each problem interview (or follow-up):
