Q1
And what a demonstration could not settle, which is why it costs your customer time and access. Example: Whether ingestion holds 40,000 events a second on their own event mix, which no demo environment can show.
Q2
Each criterion with what is measured, the target, and where the target came from. A target with no basis is a number somebody chose. Example: Sustained 40,000 events a second, because that is their Black Friday peak plus 20 per cent, from their own capacity plan.
Q3
What each target is relative to. Where no baseline exists, establishing one becomes the first milestone and needs an owner. Example: Nobody has measured current throughput. Their platform lead would have to, and has not agreed to.
Q4
Method, tooling, who runs it, and the conditions which make a result representative. A result from the wrong conditions settles nothing. Example: Their own load harness, 4-hour run, production event mix, on the cluster size they would actually buy.
Q5
Evidence nobody agreed to capture is evidence nobody will have when the decision is made. Example: Harness output and dashboard exports, captured by their platform engineer, countersigned by our architect.
Q6
Milestones with owners and dates or triggers, and the prerequisites which block each one. Example: Network access by the 20th, owned by their security lead, blocks everything. Cluster build follows.
Q7
An evaluation with no stop condition runs until somebody loses patience. Example: If access is not granted by the 20th we pause rather than compress the test window. Their sponsor or ours may call it.
Q8
The person who decides, not the team they sit in. A decision owned by a function is owned by nobody. Example: Their CTO, on the harness output and a joint readout. No date agreed.