| name | startup-trend-prediction |
| description | Analyze 2-3 year historical trends in technology, market, and business models to predict 1-2 years ahead. Uses pattern recognition, adoption curves, and cycle analysis to identify timing windows and emerging opportunities. History is cyclical - products and markets follow predictable patterns. |
| metadata | {"globs":"**/*.md\n**/research/**\n**/trends/**\n**/analysis/**\n"} |
Startup Trend Prediction
Systematic framework for analyzing historical trends to predict future opportunities. Look back 2-3 years to predict 1-2 years ahead.
Modern Best Practices (Dec 2025):
- Triangulate: require 3+ independent signals, including at least 1 primary source (standards, regulators, platform docs).
- Separate leading vs lagging indicators; don’t overfit to social/media noise.
- Add hype-cycle defenses: falsification, base rates, and adoption constraints (distribution, budgets, compliance).
- Tie trends to a decision (enter / wait / avoid) with explicit assumptions and a review cadence.
When to Use This Skill
| Trigger | Action |
|---|
| "When should I enter this market?" | Run timing analysis |
| "What's trending in [technology/market]?" | Run trend identification |
| "Is this trend rising or peaking?" | Run adoption curve analysis |
| "What comes after [current trend]?" | Run cycle prediction |
| "Historical patterns for [topic]" | Run pattern recognition |
| "2-3 year trends" or "predict 1-2 years" | Full trend prediction workflow |
Quick Reference: Building a Trend View (Dec 2025)
1) Define the Decision
- What decision are we supporting: enter / wait / avoid?
- Horizon: {{HORIZON}}
- Buyer and market: {{BUYER}} / {{MARKET}}
2) Collect Signals (Leading vs Lagging)
| Signal | Type | What it indicates | Examples | Failure mode |
|---|
| Regulation/standards | Leading | Constraints or enabling changes | Sector regulation, privacy law, ISO standards | Misreading scope/timeline |
| Platform primitives | Leading | New capability baseline | API/OS/cloud releases | Confusing announcement with adoption |
| Buyer behavior | Leading | Willingness to buy | Procurement patterns, RFPs | Sampling bias |
| Usage/revenue | Lagging | Real adoption | Public metrics, cohorts | Too slow to catch inflection |
| Media/social | Weak | Attention | Mentions, posts | Hype amplification |
3) Hype-Cycle Defenses
- Falsification: what evidence would prove the trend is not real?
- Base rates: how often do similar trends reach mass adoption?
- Adoption constraints: distribution, budget, switching costs, compliance, implementation complexity.
4) Market Sizing Sanity Checks
- Bottom-up first: #customers × willingness-to-pay × realistic penetration.
- Explicit assumptions: who pays, how much, and why you can reach them.
Adoption Curve Framework
Rogers Diffusion Model
ADOPTION CURVE
│
│ ╭────────╮
│ ╭───╯Late │
│ ╭───╯Majority │
│ ╭───╯Early │
│ ╭───╯Majority │
│ ╭───╯Early │
│ ╭───╯Adopters │
│──╯Innovators ╰──────
│ │ │ │ │ │
│ 2.5% 13.5% 34% 34% 16%
└─────────────────────────────────────────▶
TIME
Position Identification
| Position | Market Penetration | Characteristics | Strategy |
|---|
| Innovators | <2.5% | Tech enthusiasts, high risk tolerance | Enter now, shape market |
| Early Adopters | 2.5-16% | Visionaries, want competitive edge | Enter now, premium pricing |
| Early Majority | 16-50% | Pragmatists, need proof | Enter with differentiation |
| Late Majority | 50-84% | Conservatives, follow herd | Compete on price/features |
| Laggards | 84-100% | Skeptics, forced adoption | Avoid or disrupt |
Gartner Hype Cycle Mapping
HYPE CYCLE
│
│ Peak of
│ Inflated ╭─────────────
│ Expectations ╭───╯ Plateau of
│ ╭────╯ Productivity
│ ╭────╯
│ ╭────╯ Slope of
│──╯ Enlightenment
│ Technology ╲_____╱
│ Trigger Trough of
│ Disillusionment
└─────────────────────────────────────▶
TIME
| Phase | Duration | Action |
|---|
| Technology Trigger | 0-2 years | Monitor, experiment |
| Peak of Inflated Expectations | 1-3 years | Caution, don't overbuild |
| Trough of Disillusionment | 1-3 years | Build foundations |
| Slope of Enlightenment | 2-4 years | Scale solutions |
| Plateau of Productivity | 5+ years | Optimize, commoditize |
Cycle Pattern Library
Technology Cycles (7-10 years)
| Cycle | Previous Instance | Current Instance | Pattern |
|---|
| Client → Cloud → Edge | Desktop → Web → Mobile | Cloud → Edge → On-device compute | Compute moves to data |
| Monolith → Services → Composables | SOA → Microservices | Microservices → Composable workflows | Decomposition continues |
| Batch → Stream → Real-time | ETL → Streaming | Streaming → Real-time decisioning | Latency shrinks |
| Manual → Assisted → Automated | CLI → GUI | Scripts → Workflow automation | Automation increases |
Market Cycles (5-7 years)
| Cycle | Previous Instance | Current Instance | Pattern |
|---|
| Fragmentation → Consolidation | 2015-2020 point solutions | 2020-2025 platforms | Bundling/unbundling |
| Horizontal → Vertical | Horizontal SaaS | Vertical platforms | Specialization wins |
| Self-serve → High-touch → Hybrid | PLG pure | PLG + Sales | Motion evolves |
Business Model Cycles (3-5 years)
| Cycle | Previous Instance | Current Instance | Pattern |
|---|
| Perpetual → Subscription → Usage | License → SaaS | SaaS → Usage-based | Payment follows value |
| Direct → Marketplace → Embedded | Direct sales | Marketplace → Embedded | Distribution evolves |
Signal vs Noise Framework
Strong Signals (High Confidence)
| Signal Type | Detection Method | Weight |
|---|
| VC funding patterns | Track quarterly investment | High |
| Big tech acquisitions | Monitor M&A announcements | High |
| Job posting trends | Analyze LinkedIn/Indeed data | High |
| GitHub activity | Stars, forks, contributors | High |
| Enterprise adoption | Gartner/Forrester reports | Very High |
Moderate Signals (Validate)
| Signal Type | Detection Method | Weight |
|---|
| Conference talk themes | Track KubeCon, AWS re:Invent | Medium |
| Hacker News sentiment | Algolia search trends | Medium |
| Reddit discussions | Subreddit growth, sentiment | Medium |
| Influencer adoption | Key voices tweeting about | Medium |
Weak Signals (Monitor)
| Signal Type | Detection Method | Weight |
|---|
| ProductHunt launches | Daily tracking | Low |
| Blog post frequency | Content analysis | Low |
| Podcast mentions | Episode scanning | Low |
| Media hype | TechCrunch, Wired articles | Low (often lagging) |
Noise Filters
Exclude from prediction:
- Single viral tweet without follow-up
- PR-driven announcements without product
- Predictions from parties with financial interest
- Old data recycled as "new trend"
Prediction Methodology
Step 1: Define Scope
Domain: [Technology / Market / Business Model]
Lookback Period: [2-3 years]
Prediction Horizon: [1-2 years]
Geography: [Global / Region-specific]
Industry: [Horizontal / Specific vertical]
Step 2: Gather Historical Data
| Year | State | Key Events | Metrics |
|---|
| {{YEAR-3}} | | | |
| {{YEAR-2}} | | | |
| {{YEAR-1}} | | | |
| {{NOW}} | | | |
Step 3: Identify Patterns
Step 4: Generate Prediction
## Prediction: [TOPIC]
**Thesis**: [1-2 sentence prediction]
**Confidence**: High / Medium / Low
**Timing**: [When this will happen]
**Evidence**: [3-5 supporting data points]
**Counter-evidence**: [What could invalidate]
Step 5: Identify Opportunities
| Opportunity | Timing Window | Competition | Action |
|---|
| {{OPP_1}} | {{WINDOW}} | Low/Med/High | Build/Watch/Avoid |
| {{OPP_2}} | {{WINDOW}} | | |
Navigation
Resources (Deep Dives)
Templates (Outputs)
Data
| File | Contents |
|---|
| sources.json | Trend data sources (analyst reports, market data, filings, etc.) |
Key Principles
History Rhymes
Past patterns repeat with new technology:
- Client-server → Web apps → Mobile → On-device
- Mainframe → PC → Cloud → Distributed
- Manual → Scripted → Automated → Autonomous
Timing Beats Being Right
Being right about a trend but wrong about timing = failure:
- Too early: Market not ready, burn runway
- Too late: Established players, commoditized
- Just right: Ride the wave
Multiple Signals Required
Never bet on single signal:
- Funding + Hiring + GitHub activity = Strong signal
- Just media coverage = Hype, validate further
- Just VC interest = May be speculative
Update Predictions
Predictions are living documents:
- Revisit quarterly
- Track accuracy over time
- Adjust for new data
- Document what changed and why
Do / Avoid (Dec 2025)
Do
- Use a decision horizon (enter/wait/avoid) and revisit quarterly.
- Track leading indicators and adoption constraints, not just hype.
- Write assumptions explicitly and update them when data changes.
Avoid
- Extrapolating from a single platform, influencer, or funding headline.
- Treating “attention” as “adoption”.
- Market sizing without assumptions and bottom-up checks.
What Good Looks Like
- Decision: one clear enter/wait/avoid call with horizon and owner.
- Evidence: 3+ independent signal types (not just media) and explicit confidence (strong/medium/weak).
- Assumptions: TAM/SAM/SOM with assumptions + sensitivity ranges; falsification criteria documented.
- Constraints: adoption blockers listed (distribution, budget, switching, compliance, implementation) with mitigations.
- Cadence: quarterly refresh with “what changed” and accuracy notes.
Optional: AI / Automation
Use only when explicitly requested and policy-compliant.
- Topic modeling/clustering for large corpora; validate with primary sources and spot-checks.
- Summarization of reports; keep links and dates to avoid stale claims.
Integration Points
Feeds Into
Receives From
