Trade Studies: How to Choose
Stop guessing: When faced with alternatives, most teams either pick the familiar option or argue endlessly. Trade studies force objectivity. Define criteria upfront, weight them by importance, score honestly. The math doesn't lie—it surfaces hidden biases and makes your rationale defensible to stakeholders and reviewers.
Trade Study Example
| Criteria (Weight) | VxWorks RTOS | Zephyr RTOS | Bare Metal |
|---|---|---|---|
| Performance (25%) | 8 × 0.25 = 2.0 | 9 × 0.25 = 2.25 | 10 × 0.25 = 2.5 |
| Cost (20%) | 3 × 0.20 = 0.6 | 10 × 0.20 = 2.0 | 10 × 0.20 = 2.0 |
| Heritage/Risk (30%) | 10 × 0.30 = 3.0 | 6 × 0.30 = 1.8 | 4 × 0.30 = 1.2 |
| Dev Time (15%) | 9 × 0.15 = 1.35 | 8 × 0.15 = 1.2 | 3 × 0.15 = 0.45 |
| Maintainability (10%) | 8 × 0.10 = 0.8 | 9 × 0.10 = 0.9 | 4 × 0.10 = 0.4 |
| TOTAL | 7.75 | 8.15 ✓ | 6.55 |
Decision: Zephyr wins on weighted score. VxWorks close if heritage/risk valued higher. Bare metal eliminated due to development time and maintainability despite best performance. Document rationale for audits and future reference.
COTS vs Custom Decision
Default to COTS: Custom engineering is expensive, risky, and time-consuming. Every custom component is a potential failure mode and maintenance burden. But sometimes COTS doesn't exist, doesn't fit constraints, or costs more long-term. The question isn't "can we build it?" but "should we build it?"
Choose COTS When:
- Component is not core differentiation
- Mature technology exists
- Development time is limited
- Long-term support available
- Cost is lower than custom NRE
Choose Custom When:
- No COTS solution exists
- Requirements are unique/extreme
- COTS lifecycle cost > custom NRE
- Performance/size/power can't be met
- Core IP and competitive advantage
Innovative Problem Solving
Think outside the box, but use your head: Innovation means finding better solutions, not being different for its own sake. Question conventional approaches, but validate with analysis before committing.
Example: Gigabit Data Pipe
Conventional: "Use Gigabit Ethernet—it's standard."
First Principles: What are we solving?
- Need: 1 Gbps throughput, low latency
- Constraint: Limited CPU, low overhead
Alternative: Pipeline 3× USB 2.0 interfaces (480 Mbps each = 1.44 Gbps raw). Simpler framing than TCP/IP, lower CPU overhead, well-supported drivers.
Evaluation: Prototype both. Measure CPU utilization, latency, complexity. Choose based on data, not assumptions.
Balance innovation with pragmatism: Innovate when conventional solutions don't meet constraints and you can prototype quickly. Use standard approaches when proven solutions exist and risk tolerance is low.