Opportunity Discovery: A Practical Guide for When You're Stuck
The Problem in Concrete Terms
You have the skills to build things. You have time (or can make it). You want to start a project, a product, or a business—but you keep circling the same question: what should I actually work on?
This isn't "I have too many ideas." It's the opposite. You sit down to brainstorm and come up empty. Or you generate a list, but every idea feels:
- Already done better by someone with more resources
- Too small to matter
- Too vague to start
- Something you'd abandon in two weeks
You've read about "scratching your own itch," but your itches are either already solved (you use Notion, Linear, Figma daily) or too niche to build a product around. You've tried "talking to customers," but you don't know who they are or what to ask. You've looked at market trends, but AI, Web3, and SaaS feel saturated by the time you notice them.
The result: weeks or months of paralysis. You consume content about ideation frameworks but never apply them. You bookmark "problems worth solving" newsletters but never follow up. You're stuck in the gap between wanting to build and knowing what to build.
Why the Usual Fixes Fail
1. "Scratch your own itch" assumes you have unsolved, generalizable problems. Most developers' daily frustrations are either too specific (my CI pipeline is slow) or already commoditized (I want a better to-do app). The problems you can articulate are often the ones least likely to support a product.
2. "Talk to 50 customers" is circular advice. You need an idea to know who to talk to. You need conversations to refine an idea. Cold outreach without a hypothesis gets ignored. Warm intros require a network you're trying to build.
3. Trend-chasing puts you behind. By the time a trend appears in newsletters, Twitter, or Product Hunt, the early movers have 6–18 months' head start. You're not discovering opportunities—you're reacting to saturation signals.
4. Frameworks like Jobs-to-be-Done or Value Proposition Canvas are analysis tools, not discovery tools. They structure thinking after you have a direction. They don't generate the initial spark.
5. Waiting for inspiration is a luxury you can't afford. Inspiration is pattern recognition at speed. It comes from volume of exposure, not passive waiting.
Step-by-Step Solution
Phase 1: Build Your Input Pipeline (Week 1)
You cannot discover opportunities in a vacuum. You need raw material—observations, complaints, workflows, constraints—from domains you don't currently inhabit.
Action 1: Pick three adjacent domains. Not "healthcare" or "finance." Pick specific, observable workflows:
- How a freelance video editor manages client feedback loops
- How a small wholesale distributor handles inventory reconciliation
- How a property manager coordinates maintenance across 50 units
Choose domains where you can observe or interview within your existing network (former coworkers, friends, meetup contacts, subreddits, Discord servers). Avoid domains requiring cold calls to strangers.
Action 2: Set up 15-minute observation sessions. Not interviews. Observations. Ask: "Can I watch you work for 15 minutes while you do [specific task]? I'm studying workflow friction, not selling anything." Record (with permission) or take timestamped notes. Focus on:
- Workarounds (spreadsheets, copy-paste, manual checks)
- Handoffs between tools or people
- Moments of visible frustration or hesitation
- "I wish this would just..." statements
Action 3: Collect 30+ friction points. Across your three domains, document every workaround, complaint, manual step, and "there should be a better way" moment. Format each as: [Role] spends [time] doing [manual task] because [constraint], resulting in [pain/cost]. Example: Property managers spend 4 hrs/week manually matching vendor invoices to work orders because their PMS doesn't link them, resulting in delayed payments and vendor disputes.
Do not evaluate yet. Just collect.
Phase 2: Filter for Buildable Opportunities (Week 2)
Now you have raw material. Apply ruthless filters to find the ones you can actually act on.
Filter 1: Frequency × Cost. Keep only frictions that occur at least weekly and cost measurable time/money. A monthly annoyance saving 10 minutes isn't a product. A daily 45-minute manual process costing $200/week in labor is.
Filter 2: Software-addressable. Can code reduce or eliminate this? Some frictions are process, policy, or people problems. Discard those. Keep: data entry, matching, routing, calculation, notification, reporting, scheduling, compliance tracking.
Filter 3: Budget owner identifiable. Who pays to solve this? Not "the company." A specific title: "Operations Manager at 10–50 unit property management firms." If you can't name the title and find 20 of them on LinkedIn in 10 minutes, discard.
Filter 4: You can build an MVP in 4–6 weeks. Be honest about your stack and scope. A mobile app with offline sync and real-time collaboration? No. A web dashboard that ingests CSV, applies rules, outputs matched pairs? Yes.
Filter 5: You have unfair access to 3+ potential users. From your observation sessions, who said "I'd pay for that" or "When is this ready?" If zero, go back to Phase 1 with new domains.
Output: A shortlist of 3–5 opportunities, each with: problem statement, target user (title + context), current cost, MVP scope, 3 warm contacts.
Phase 3: Validate with Evidence, Not Opinions (Weeks 3–4)
Stop asking "would you use this?" Start proving demand.
Step 1: Build a clickable prototype. Not code. Figma, Framer, or even a detailed Notion page with screenshots. Show the exact workflow: input → processing → output. Include realistic sample data from your observation sessions.
Step 2: Run 5 "concierge tests." For each shortlisted opportunity, find one warm contact. Say: "I'm building a tool that does [X]. It's not ready, but I can run the process manually for you this week using your real data. Interested?" Do the work yourself. Charge nothing. Measure:
- Did they give you real data? (Yes = pain is real)
- Did they use the output? (Yes = value is real)
- Did they ask "when can I get access?" (Yes = willingness to pay)
Step 3: Test pricing with a proposal. After the concierge test, send a one-page proposal: "Based on this week, here's what the automated version would do, what it would cost ($X/mo), and what it saves you ($Y/mo). Want to be a founding user at 50% off for 12 months?" Track yes/no/maybe.
Step 4: Kill or commit. If 0/5 say yes to the proposal, kill the idea. If 1–2 say yes with conditions, iterate the prototype and test 3 more. If 3+ say yes, you have a validated opportunity. Start building.
Phase 4: Build the Minimum Sellable Product (Weeks 5–10)
Scope to the "aha" moment only. Every feature that doesn't directly deliver the validated output gets cut. No settings, no onboarding flow, no integrations beyond what the concierge test required.
Build in public to your 3–5 founding users. Deploy weekly. Get feedback on output quality, not UI polish. Fix only what breaks their workflow.
Charge from day one. Even $50/mo. Free users give polite feedback. Paying users give demand signals.
Document everything. Your process—from observation to concierge to MVP—is now a repeatable discovery engine. The next opportunity will be faster.
Checklist
- [ ] Identified 3 specific, observable domains within network reach
- [ ] Completed 5+ observation sessions (15 min each) across domains
- [ ] Documented 30+ friction points in standard format
- [ ] Applied 5 filters (frequency×cost, software-addressable, budget owner, MVP scope, warm access)
- [ ] Produced shortlist of 3–5 opportunities with user titles and contacts
- [ ] Built clickable prototype for top opportunity
- [ ] Ran 5 concierge tests with real user data
- [ ] Sent paid proposals to concierge participants
- [ ] Achieved 3+ "yes" commitments at
Top comments (0)