Infrastructure investment and reporting platform

A production decision-support program that turned 16M+ data points and 60+ standards into recommendations used across five regional departments.

Center for Strategic Initiatives2021 — 2023Product Manager · matrix-led delivery, decision logic, adoption6 engineers · 2 data/ML specialists · 2 analysts/designersproduction across 5 regions

$40M+allocation decisions influenced

16M+infrastructure data points92%final-approved completion99%report accuracy

Context: unequal needs, one capital plan

More than 100 interviews informed a shared program combining predictive scenarios, 60+ standards, automated validation, ranked recommendations, and redesigned reporting workflows.

Manual prioritization was subjective, regional evidence was fragmented, and only 25% of reports reached final approval. Teams could not consistently defend why one investment ranked above another.

Turn standards into decision logic

evidence → scenario → recommendation

Pair predictive scale with traceable decision logic

evidence16M+ data points

Regional infrastructure and reporting inputs.

standards60+ decision rules

Comparable requirements across categories.

modelscenario + ranking

Monte Carlo and Random Forest signals.

decisionbudget-constrained priority

Recommendation reviewed by executives.

A real allocation decision

standard gap60+ rules

evidence16M+ data points

feasibilitydelivery + cost

ranked needexplainable priority

budget decision$40M+across 5 regions

The product combined the standards gap, population exposure, feasibility, and a real budget constraint—so a rank could be explained and defended.

The scoring choices

Model uncertainty instead of hiding it

Use Monte Carlo scenarios and Random Forest signals inside a traceable recommendation workflow. Executives needed scalable prioritization without losing the evidence behind a rank.

Alternative considered: Use a single linear score or present raw indicators without prioritization.

Trade-off: The model was more complex, so the product exposed standards, factors, and constraints alongside each recommendation.

Treat completion as a product problem

Add reminders, status badges, progress visibility, and targeted workflow changes. Drop-off data showed that complex submissions—not lack of demand—blocked completion.

Alternative considered: Rely on training alone or enforce completion administratively.

Trade-off: Adoption work expanded scope, but final-approved completion rose from 25% to 92%.

My product scope

Defined core features, technical requirements, metric logic, and validation rules
Translated 60+ standards into scoring and recommendation workflows
Matrix-led engineering, data, design, finance, policy, and regional stakeholders
Owned the roadmap, delivery scope, executive sign-offs, and rollout
Redesigned reminders, status visibility, and adoption flows after drop-off analysis

Outcome

Decisions

The platform influenced more than $40M in regional capital-allocation decisions.

Completion

Final-approved completion increased from 25% to 92%; overall completion rose from 40% to 75%.

Quality

Automated validation reached 99% report accuracy across 10K+ monthly reports.

measurement notes

Definitions and sources are shown so every result can be examined in an interview.

$40M+ influenced

Capital-allocation decisions in workflows where ranked recommendations informed executive prioritization.

16M+ data points

Infrastructure evidence processed by the shared regional decision-support program.

25% → 92%

Final-approved report completion measured over one quarter.

99% report accuracy

Validated report accuracy across a workflow processing 10K+ monthly reports.

A predictive model became useful only when executives could inspect its logic and regional teams could complete the workflow around it.