Software TRIZ
Contradiction Matrix
40 inventive principles adapted for software engineering. 16 performance parameters. 240 mapped contradictions. The engineering trade-offs you fight every day are inventions waiting to happen.
Background
TRIZ was invented for physical engineering. We rebuilt it for software.
Classical TRIZ (Theory of Inventive Problem Solving) uses 39 engineering parameters and 40 inventive principles derived from studying 200,000+ patents. But “weight of a moving object” and “shape” don’t map to microservices. We adapted the entire framework for modern software systems.
Software parameters (latency, throughput, scalability, consistency, security, etc.)
Inventive principles rewritten with software examples (microservices, caching, streaming, etc.)
Contradiction cells mapped to specific inventive strategies for resolving software trade-offs
The IP Ramp Method
The principle is the compass.
The mechanism is the patent.
TRIZ tells you where to look. The Three-Layer Drill tells you how deep to go. The Alice Pre-Screen tells you if it will survive. Together, they form one continuous workflow from engineering trade-off to defensible patent.
Contradiction
Name the trade-off: what improves vs. what worsens
Principles
Look up the matrix cell → get 3 inventive directions
Alice Pre-Screen
4 questions to verify your Layer 3 is patent-safe
This methodology is derived from 200,000+ patent analyses. Each step builds on the last. Skipping steps produces weaker patents.
The Full Method
From trade-off to patent in four steps
The TRIZ matrix is Step 1. What you do after — the Three-Layer Drill and Alice pre-screen — is what separates a clever observation from a defensible patent.
Identify your contradiction
Every non-trivial engineering decision involves a trade-off. Find the parameter you want to improve (the ROW) and the parameter that worsens as a result (the COLUMN). Be specific — “performance” is too vague; “latency” or “throughput” is precise.
"We need to improve Latency, but every approach we've tried worsens Consistency."
Read the cell and look up principles
The cell at the intersection contains 3 inventive principle IDs, ordered by relevance. Look each up in the Principle Quick Reference. Each principle is a general strategy — a compass heading, not GPS coordinates.
Latency (row) vs. Consistency (column) → Principles: 16 (Partial Action), 4 (Asymmetry), 13 (The Other Way Around)
Apply the Three-Layer Drill
This is where most teams fail — and where patents are born
Use the principle as a starting direction, then describe your solution at three levels of depth. Layer 1 is the obvious description (not patentable). Layer 2 adds architectural detail (maybe patentable). Layer 3 describes the inventive mechanism — the specific, novel, non-obvious way you resolved the contradiction. Layer 3 is the patent.
Layer 1: "We use eventual consistency." Layer 2: "We use optimistic locking with version vectors in a CRDT-based store." Layer 3: "Each partition maintains a causal dependency graph that prunes stale writes using vector clock compression, reducing conflict resolution from O(n²) to O(log n) while preserving causal ordering guarantees across geo-distributed replicas."
Run the Alice / Section 101 pre-screen
Four questions that predict whether your patent survives
Before you spend $15–25K filing, ask four questions about your Layer 3: Does it improve a technical process (not just a business outcome)? Is the improvement tied to a specific mechanism (not “using AI to optimize”)? Would it require a specific implementation to work? Is there something unconventional about how components interact?
Good: "Reduces p99 latency by 40% using per-key learned eviction." Bad: "Uses AI to optimize caching for better performance." The first is Alice-safe. The second gets rejected.
“The principle is the compass. The mechanism is the patent.”
TRIZ principles give you a direction. Your specific, non-obvious implementation — the Layer 3 mechanism — is what gets granted.
Interactive Matrix
Explore the matrix
Click any cell to see which inventive principles resolve that specific contradiction. Toggle to the principles catalog to browse all 40.
Software TRIZ Contradiction Matrix
16 parameters -- 40 inventive principles -- adapted for software & cloud engineering
| IMPROVE ↓ / WORSEN → | Latency | Throughput | Memory Usage | Storage | Scalability | Consistency | Availability | Security | Accuracy | Debuggability | Infra Cost | Complexity | Reliability | Maintainability | Real-time Perf | Data Freshness |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Latency | -- | 5,20,38 | 7,10,21 | 10,7,34 | 8,12,1 | 16,4,13 | 7,26,11 | 8,39,24 | 21,16,31 | 32,7,30 | 27,8,37 | 7,10,30 | 11,9,26 | 30,7,33 | 20,38,5 | 10,7,20 |
| Throughput | 5,20,29 | -- | 5,34,19 | 5,34,19 | 1,12,37 | 16,4,5 | 26,37,15 | 24,39,5 | 16,5,19 | 32,33,5 | 27,37,38 | 5,6,33 | 9,11,23 | 33,6,30 | 20,38,29 | 20,5,13 |
| Memory Usage | 21,34,31 | 34,21,5 | -- | 34,31,7 | 1,34,27 | 34,16,7 | 34,11,27 | 39,34,31 | 31,21,16 | 32,34,31 | 27,34,31 | 34,31,21 | 11,34,7 | 34,30,33 | 21,34,38 | 34,19,21 |
| Storage | 34,7,10 | 34,5,19 | 34,31,7 | -- | 34,1,17 | 34,4,7 | 34,26,27 | 34,39,17 | 34,7,17 | 34,32,17 | 34,27,19 | 34,30,6 | 34,9,26 | 34,33,30 | 34,20,38 | 34,20,13 |
| Scalability | 1,8,12 | 1,37,12 | 1,27,17 | 1,17,34 | -- | 16,12,4 | 26,12,37 | 39,17,1 | 16,1,12 | 32,17,33 | 27,37,1 | 1,6,33 | 12,9,26 | 33,1,6 | 1,8,37 | 13,20,1 |
| Consistency | 4,16,10 | 4,16,5 | 4,7,34 | 4,34,7 | 4,12,16 | -- | 16,4,36 | 39,4,24 | 4,23,33 | 32,4,33 | 4,27,16 | 4,33,30 | 4,26,9 | 4,33,30 | 4,20,16 | 4,20,13 |
| Availability | 26,11,7 | 26,37,15 | 26,27,11 | 26,34,27 | 26,12,37 | 16,26,36 | -- | 39,26,24 | 16,26,11 | 32,26,33 | 27,26,37 | 26,11,6 | 26,9,22 | 26,33,25 | 26,11,20 | 26,20,13 |
| Security | 39,24,8 | 39,24,5 | 39,31,34 | 39,34,17 | 39,17,1 | 39,4,24 | 39,26,11 | -- | 39,23,33 | 39,32,30 | 39,27,28 | 39,30,6 | 39,9,26 | 39,33,30 | 39,8,24 | 39,24,13 |
| Accuracy | 10,21,16 | 5,16,19 | 31,21,7 | 34,7,17 | 1,16,12 | 4,23,33 | 11,26,16 | 39,23,33 | -- | 32,23,33 | 27,16,31 | 23,33,30 | 23,9,11 | 33,23,30 | 10,38,21 | 20,10,13 |
| Debuggability | 32,30,7 | 32,33,5 | 32,34,31 | 32,34,17 | 32,17,33 | 32,4,33 | 32,26,33 | 32,39,30 | 32,23,33 | -- | 32,27,33 | 32,33,30 | 32,9,22 | 32,33,30 | 32,20,33 | 32,13,20 |
| Infra Cost | 27,8,37 | 27,37,38 | 27,34,31 | 27,34,19 | 27,37,1 | 27,16,4 | 27,26,11 | 27,28,39 | 27,16,31 | 27,32,33 | -- | 27,6,28 | 27,9,22 | 27,33,28 | 27,37,8 | 27,19,34 |
| Complexity | 30,6,33 | 6,33,5 | 30,34,6 | 30,34,6 | 6,33,1 | 33,4,30 | 6,11,33 | 30,39,6 | 33,23,30 | 32,33,30 | 6,28,27 | -- | 6,9,33 | 33,30,6 | 30,6,8 | 30,6,13 |
| Reliability | 9,11,26 | 9,11,23 | 9,26,27 | 9,26,34 | 9,12,26 | 9,4,26 | 26,9,22 | 39,9,26 | 23,9,11 | 32,9,22 | 27,9,22 | 9,22,6 | -- | 9,33,25 | 9,11,20 | 9,20,26 |
| Maintainability | 33,30,7 | 33,6,30 | 33,30,34 | 33,30,34 | 33,1,6 | 33,4,30 | 33,26,25 | 33,39,30 | 33,23,30 | 33,32,30 | 33,28,27 | 33,30,6 | 33,9,25 | -- | 33,30,8 | 33,30,13 |
| Real-time Perf | 20,38,5 | 20,38,29 | 20,38,21 | 20,34,38 | 1,8,37 | 20,4,16 | 20,11,26 | 8,39,20 | 10,38,21 | 32,20,33 | 27,37,8 | 20,30,6 | 9,20,11 | 33,20,30 | -- | 20,13,10 |
| Data Freshness | 13,20,10 | 20,13,5 | 13,34,20 | 13,34,20 | 13,1,20 | 13,4,20 | 13,26,20 | 13,39,24 | 13,20,10 | 13,32,20 | 13,27,19 | 13,30,6 | 13,9,20 | 13,33,20 | 13,20,10 | -- |
Select a contradiction
Click any cell in the matrix to see which inventive principles resolve the tradeoff between improving one parameter while the other worsens.
How to use this in a patent sprint:
1. Identify your core engineering contradiction
2. Find the row (what you're improving) and column (what worsens)
3. Read the suggested principles and examples
4. Ask: “Does my solution use one of these patterns in a novel way?”
Worked Example
A real contradiction, resolved
Scenario
Building an ML inference API with strict latency SLAs
The team needs to serve model predictions under 50ms p99, but the model accuracy improves significantly with ensemble methods that take 200ms+. Improving latency degrades accuracy. Improving accuracy worsens latency.
Matrix Lookup
Principles Applied
Skipping (Lazy evaluation, sampling, probabilistic data structures)
Instead of running all ensemble models, use a lightweight router that predicts which sub-model will dominate for each input and only runs that one.
Partial or Excessive Action (Optimistic execution, speculative)
Speculatively start all ensemble members but return the first response that exceeds a confidence threshold, cancelling the rest.
Porous Materials (Probabilistic sketches, sparse structures)
Use a distilled “sketch” model that approximates the ensemble output for 90% of inputs, only falling back to the full ensemble for edge cases.
The Invention
The team combined principles 21 and 31: a lightweight routing model (trained on the ensemble’s own disagreement patterns) that predicts input difficulty and routes easy inputs to a distilled fast model (<10ms) and hard inputs to the full ensemble. The router is retrained weekly using the ensemble’s confidence distribution as labels. Result: p99 latency dropped from 200ms to 35ms with only 1.2% accuracy degradation on the hardest 5% of inputs.
Patent-Rich Principles
Six principles that generate the most patents
Based on our analysis of software patent filings, these principles appear most frequently in granted patents.
Nested Doll
Layered caching (L1 in-process, L2 Redis, L3 CDN). Nested containers. Hierarchical config.
Patent angle
Multi-tier architectures where each layer adds value are rich patent territory. The novel part is how layers interact.
The Other Way Around
Push instead of pull. Event-driven instead of polling. Invert the dependency.
Patent angle
Inverting the conventional approach is one of the strongest signals of non-obviousness in patent examination.
Dynamics
Auto-scaling, adaptive algorithms, dynamic configuration, self-tuning systems.
Patent angle
Systems that adapt their own behavior based on runtime signals are highly patentable when the adaptation mechanism is specific.
Feedback
Observability, closed-loop control, adaptive rate limiting, PID controllers for autoscaling.
Patent angle
Closed-loop systems that use their own output as input to improve are patent goldmines when the feedback signal is novel.
Inert Atmosphere
Sandboxing, isolation, namespaces, security boundaries, zero-trust networks.
Patent angle
Security innovations that create novel isolation boundaries are consistently patentable and defensible.
Composite Materials
Polyglot persistence, hybrid architectures, combining multiple strategies for one goal.
Patent angle
Novel combinations of known technologies for a specific purpose. The combination is the invention.
Go deeper
Every trade-off is a patent
waiting to be filed.
IP Ramp combines the TRIZ matrix with AI-powered ideation, automatic Alice scoring, and claim generation. Turn your engineering contradictions into defensible IP.