Each simulated case is scored 0 to 1 on quality, performance gain, mission relevance, and evidence, and rated for risk, cost, and complexity. Quality below 0.60 stops a case.
Each score puts one scorecard measure on a 0 to 1 scale so cases compare. Scores here are simulated; for a real case, rate your evidence against the anchors below.
Time per summary fell from 35 to 14 minutes, including review
A clear, measured gain on the primary outcome, between 0.50 and 0.90
0.75
Mission relevance
Internal team productivity, not a named institutional priority
Supports a priority indirectly
0.50
Evidence strength
Four weeks of before-and-after timing, no comparator
Matches the before-and-after anchor
0.50
Quality passes the gate, the gain is solid, and risk and cost are low, so the decision is Continue. Weak evidence and middling mission relevance are why it is not Expand.
Risk, cost, and complexity are rated low, medium, or high. Cost corresponds to the scorecard’s incremental cost measure; risk follows the governance and risk lens.
{{ iCount }}
No case matches all of these filters. Clear one and try again.
Explanation generated from the case’s scores and the framework’s gates.
Decisions across the inventory
Level
Risk
{{ filterLabel }}{{ total }} cases
{{ b.name }}{{ b.n }} · {{ b.pct }}
Risk
Continue
Improve
Expand
Restrict or pause
Retire
Cases
Restricted
Low risk
41
41
2
56
14
154
36%
Medium risk
36
48
0
64
18
166
39%
High risk
17
20
0
137
6
180
76%
What the inventory shows
Quality and risk gates work as intended
76% of high-risk cases were restricted or paused. All 257 below the quality bar were stopped, including 84 with strong productivity gains.
The framework is conservative
Only 0.4% reached Expand and 18.8% Continue. Most cases stay in pilot or improvement until they clear thresholds. Defensible, but it slows low-risk adoption unless thresholds vary by stakes.
Mission relevance and evidence act as secondary filters
They came into play after quality and performance, matching the staged progression: access, use, quality, performance, mission impact, sustainable value.
The multilevel structure held
Decisions spread reasonably across Individual, Team, Unit, and Institution without forced aggregation.
Practical improvements
Seven refinements drawn from the inventory and earlier reviews. Each is now applied in the framework.
1
Differentiate thresholds by risk and stakes
A low-stakes productivity tool faced the same bars as student-facing advising. Tie the bar to the Compass review bands, each with its own minimum quality, evidence, and approval.
In the dataImprove was the second most common outcome (109 cases). 34 of them scored like Continue cases.
3
Strengthen lifecycle and drift monitoring
First decisions are well covered; what follows is not. Set a monitoring cadence and triggers for model changes, vendor updates, quality drops, and shifting use.
Give each use-case type a few default equity checks: differences in access and in outcome quality by subgroup, and accessibility fix rates. No separate assessment system needed.
Redeployed capacity is hard to show. List acceptable evidence (reallocated hours, added services, backlog cut with stable staffing) so units go beyond time saved.
The core held: three domains, four levels, lenses as gates, the use case as the unit, and clear decisions. What needed work: differentiation by stakes, paths for improvement and monitoring, and concrete equity and capacity measures.
About the data
The 500 cases are simulated. Scores run 0 to 1; cost, risk, and complexity are low, medium, or high. Decisions come from applying the framework’s gates to each case.