_A UX test (outside-user) in the helm-expt test map._
This is a short test protocol for a Helm user who has not worked on helm-expt. The goal is to find out whether the public surface explains the product without inside context.
Run this test with a real person. Do not coach them except where the script says to answer a question.
Participant
Pick someone who knows Helm and Kubernetes, but does not know this repo.
Record:
name:
role:
Helm experience:
Kubernetes experience:
ConfigHub experience:
date:
repo commit:
Setup
Give the participant the repository URL and the generated public site path:
https://github.com/confighub/helm-expt
site/try.html
If they want to run commands, they need Node and a shell. They do not need an npm install step.
Fifteen-Minute Script
- Start at
site/try.html. - Find the Redis catalog entry.
- Identify the recommended Redis variant.
- Open the rendered objects for that variant.
- Open the proof status for that variant.
- Run one local verification command from Verify It Yourself.
- Read Why Synced Is Not Working.
- Open Top-100 Coverage.
Do not explain the model first. The page should carry the first pass.
Questions
Ask these after the participant finishes the script:
- What problem is helm-expt trying to solve?
- What is proven for Redis?
- What is not proven?
- Where would you look for a different chart?
- What would make you trust this enough to try it on one of your charts?
- Which part felt too complicated?
- Which part, if any, would be worth paying for?
Target Outcomes
| Task | Target |
|---|---|
| Explain the purpose | Under 2 minutes. |
| Find Redis variants | Under 2 minutes. |
| Find rendered objects and proof status | Under 3 minutes. |
| Run one verification command | Under 5 minutes after cloning. |
| State one current limitation | Participant can name at least one limitation without prompting. |
Results Template
participant:
commit:
time_to_purpose:
time_to_redis_variant:
time_to_rendered_objects:
time_to_verification:
wrong_turns:
unclear_words:
trusted_claims:
untrusted_claims:
would_try:
would_pay_for:
notes:
Follow-Up
File findings as repo issues when they change the product surface:
unclear first page
variant not findable
proof status hard to interpret
command failed
claim sounded too broad
paid/free boundary unclear