QRNG buyer's guide, twelve questions to ask any vendor
Twelve questions to put to any entropy source vendor, covering the physical process, the entropy model and what it is conditional on, runtime evidence, certification scope, integration cost and supply continuity. The answers are more informative than any specification sheet.
Made to be pasted into a procurement document. Ask us the same twelve — a question set only one vendor can answer is a specification written backwards.
The physical source
- 1. Which physical process produces the unpredictability? Name the mechanism, not the product category.
- 2. What is the min-entropy per raw sample, before conditioning?
- 3. What model is that estimate conditional on, and which document states it?
Evidence in service
- 4. Which health tests run continuously, and which failure modes do they detect? Specifically: what detects a source producing two-thirds of its modelled entropy?
- 5. Are the physical quantities the entropy model depends on measured at runtime, or only during evaluation?
- 6. Can the device report entropy health to our monitoring, and what does it do when the source fails?
Certification
- 7. Which scheme, which document version, which laboratory, and what exactly was in scope — the entropy source, the DRBG, or the whole module?
- 8. Was the evaluated configuration identical to what we would deploy?
- 9. If a vendor uses the phrase "Level 4 certified", ask which body issues it. No body does; the levels are an assurance model.
Integration and supply
- 10. What is the integration path from evaluation to production, and does the entropy source change across form factors?
- 11. What is the expected lifetime, and what happens to the entropy claim as the device ages?
- 12. What is the supply position — lead times, second sourcing, and the roadmap from module to integrated silicon?
Reading the answers
Question 3 is the most diagnostic. A vendor who cannot state what their entropy estimate is conditional on has not written the model down, and the number is decoration. Question 4 is next: it separates sources that are tested from sources that are observed.
A supplier who answers honestly and has gaps is a better prospect than one who has no gaps at all. The entropy pillar is where these questions get their full treatment.