Trust Center · Phase 2 evidence
What we changed after Phase 1.
For each judging criterion: the judges' main concern, what we built in response, and the measured result. Numbers are read from result files written by the Phase 2 scripts; anything not yet measured says so.
Phase 1 scores
total 76.67 / 100
Bar = share of the points available for that criterion.
Criterion 1 of 7
Problem relevance
The judges' concern
“no evidence of interviews, deployment observations ... controlled customer testing”
What we built
- An on-site user study on participants' own phones, in Bangla and English: do people understand the Pause and the “not sure” state, and would they stop?
Criterion 2 of 7
AI/ML depth
The judges' concern
To be added from the Phase 1 feedback.
What we built
Being written up.
measuring… (no results file yet)
measuring… (no results file yet)
Criterion 3 of 7
Business impact
The judges' concern
Customer impact depends on an assumed 30–70% intervention rate.
What we built
- An economic model that asks what follow rate UVERA needs to break even, instead of assuming one.
measuring… (no results file yet)
Criterion 4 of 7
Prototype
The judges' concern
To be added from the Phase 1 feedback.
What we built
Being written up.
measuring… (no results file yet)
Criterion 5 of 7
Innovation
The judges' concern
“quantify whether linking and evidence generation actually improve analyst decisions beyond simple alert aggregation”
What we built
- Case Linker against naive alert aggregation on the same alerts, plus an A/B analyst task in the user study.
measuring… (no results file yet)
Criterion 6 of 7
Scalability
The judges' concern
To be added from the Phase 1 feedback.
What we built
Being written up.
measuring… (no results file yet)
Criterion 7 of 7
Responsible AI
The judges' concern
To be added from the Phase 1 feedback.
What we built
Being written up.
measuring… (no results file yet)
measuring… (no results file yet)