DEV Community

Alan Descure
Alan Descure

Posted on Originally published at smtm-store.pages.dev

Opportunity Discovery: A Field Guide for When You’re Stuck Guessing

Opportunity Discovery: A Field Guide for When You’re Stuck Guessing

The Problem in Concrete Terms

You have capacity—time, capital, or a team—but you don’t know where to point it. Every direction looks equally viable or equally risky. You’re not lacking ideas; you’re drowning in them. Or worse, you have no ideas that survive first contact with a spreadsheet.

The stuckness usually manifests in three specific ways:

  1. The "Shiny Object" Loop: You chase a new market signal every Monday (AI wrappers, vertical SaaS, marketplace models), build a prototype by Wednesday, lose conviction by Friday, and restart next week.
  2. The "Analysis Paralysis" Trap: You have a list of 20 potential problems. You build TAM models for all of them. The numbers are all "big enough" but none feel right. You optimize for spreadsheet elegance instead of market pull.
  3. The "Internal Echo Chamber": You validate ideas by asking colleagues, LinkedIn polls, or friends who want to be supportive. You mistake politeness for demand. You build something nobody pays for because nobody told you it was a bad idea—they just didn't buy.

The core issue isn't a lack of creativity. It’s a lack of a repeatable filter. You are treating discovery as a brainstorming exercise. It is a research discipline.


Why the Usual Fixes Fail

1. "Talk to Customers" (Generic Advice)
You book 15 calls. You ask, "Would you use a tool that does X?" They say "Yes." You build it. They don't buy.
Why it fails: People are terrible predictors of their own future behavior. They are polite. They want to be helpful. You asked for opinions, not evidence of pain or budget.

2. Keyword Research / SEO Volume Tools
You see "project management software" has 100k searches/month. You decide to build a Trello clone.
Why it fails: Volume indicates interest, not unsolved pain. High volume usually means high competition and mature solutions. You find crowded markets, not opportunities.

3. Trend Reports (Gartner, CB Insights, Twitter Threads)
You read "AI Agents are the future." You start building an agent framework.
Why it fails: By the time a trend is in a report, the alpha is gone. You are entering a red ocean with better-funded swimmers. Trends are lagging indicators; opportunities are leading indicators.

4. Hackathons / Weekend Sprints
You build a "MVP" in 48 hours. It works technically. Nobody uses it.
Why it fails: You validated buildability, not sellability. Code is the most expensive way to test a hypothesis.


Step-by-Step Solution: The Evidence-Based Discovery Loop

This process replaces guessing with evidence gathering. It takes 2–4 weeks per cycle. Do not write production code until Step 5.

Phase 1: Define the Hunting Grounds (Days 1–2)

Action: Pick one specific constraint to anchor your search. Do not search "everywhere."

  • Constraint A (Founder Fit): "I have deep domain expertise in [Industry X] and want to stay there."
  • Constraint B (Channel Access): "I have a newsletter/audience/network of [Specific Persona Y]."
  • Constraint C (Tech Advantage): "I have a unique capability in Tech Z."

Deliverable: A one-sentence "Search Mandate."
Example: "Find a workflow problem in commercial real estate property management where Excel is the current 'competitor' and I can reach 50 decision-makers via my alumni network."

Phase 2: Map the Workflow, Not the Market (Days 3–7)

Action: Find 3–5 practitioners doing the job right now. Not buyers, not executives—operators. Watch them work.

  • Method A (Shadowing): "Can I sit next to you for 90 minutes while you do [monthly close / tenant onboarding / compliance audit]? I’ll buy lunch."
  • Method B (Screen Record): "Record your screen next time you do [painful task]. Narrate what you're thinking. I’ll pay $100 for the hour of footage."
  • Method C (The "Frankenstack" Audit): Ask: "Show me the spreadsheet/Notion/Airtable/whiteboard you built because the software couldn't handle it."

What to capture (verbatim):

  1. The Trigger: What starts this workflow?
  2. The Steps: Every click, copy-paste, email, VLOOKUP, phone call.
  3. The Hacks: Where do they leave the "official" tool?
  4. The Cost: Time spent, error rate, rework frequency, regulatory risk.
  5. The Emotion: Where do they sigh, curse, or apologize to a colleague?

Deliverable: 3–5 annotated workflow diagrams (Miro, Excalidraw, paper photos). Highlight the "Hack Steps" in red. These are your opportunity candidates.

Phase 3: Quantify the Pain (Days 8–12)

Action: Take the top 3 "Hack Steps" from Phase 2. Validate severity with a new set of 10–15 practitioners (cold outreach, paid incentives $50–100).
Do not pitch. Run a Problem Interview Script:

  1. Context: "Walk me through the last time you did [Hack Step]."
  2. Frequency: "How often does this happen? (Daily/Weekly/Monthly)"
  3. Current Cost: "How long does it take? What happens if it goes wrong?"
  4. Current Spend: "What tools/services/people are you paying right now to mitigate this?" (This proves budget exists).
  5. Magic Wand: "If this step vanished tomorrow, what changes for you/your boss/your client?"
  6. Commitment Ask: "I’m exploring a solution for exactly this. If I built a prototype that solved only this step, would you be willing to test it for 2 weeks and give raw feedback? No cost, just your time."

Scoring Rubric (Per Problem):
| Metric | Weak (1) | Medium (2) | Strong (3) |
| :--- | :--- | :--- | :--- |
| Frequency | Monthly/Quarterly | Weekly | Daily |
| Time Cost | < 30 mins | 30–60 mins | > 1 hour |
| Error Risk | Annoyance | Rework/Client complaint | Revenue loss/Compliance fail |
| Budget Proof | "We make do" | "We pay for a bandaid tool" | "We pay consultants/agencies" |
| Test Commitment | "Maybe/Keep me posted" | "Send me a link" | "When can we start? / Who do I pay?" |

Deliverable: A ranked list. Kill any candidate scoring < 10/15. You only proceed with the top scorer.

Phase 4: The "Concierge" Validation (Days 13–21)

Action: Solve the problem manually for 3–5 pilot users. No code. You are the software.

  • They send you the input (PDF, CSV, email forward, API key).
  • You run the logic (scripts, manual entry, API calls, LLMs).
  • You deliver the output (Report, CSV, Dashboard, Email).

Rules:

  • Charge a nominal fee ($100–$500/mo) or a "deposit" refundable if unsatisfied. Free pilots yield fake data.
  • Set a hard 2-week timeline.
  • Measure: Did they use the output? Did they complain when it was late? Did they ask for features?

Deliverable: A "Service Log" documenting inputs, outputs, time-to-deliver, and user feedback. If they churn in 2 weeks, the problem isn't acute enough. Return to Phase 3.

Phase 5: The Thin Slice Build (Days 22–35)

Action: Build the smallest automatable slice of the Concierge service.

  • Scope: One input format → One core transformation → One output format.
  • Stack: Boring tech you know. No new frameworks. Host on Railway/Render/Vercel. Auth via Magic Link or Google OAuth. DB: Postgres/SQLite.
  • Billing: Stripe Billing Portal (handles invoices, cancellations, tax) from Day 1.

Goal: Onboard your 3 Concierge pilots

Start here

Top comments (0)