observed signalservice level compared with a documented standard
product decisiontranslate the gap into a maintainable rule
how it was checkedtechnical and regional stakeholder review
infrastructure prioritization and recommendation system
A recommendation system that converted standards, regional evidence, and budget constraints into explainable capital priorities.
4,000+
60+
$40m+
5+
overview
The product translated more than 60 standards and regional datasets into ranked, explainable recommendations. Users could filter by region and category, inspect the reasoning behind a score, and evaluate priorities against budget constraints.
the user — Government and regional finance teams prioritizing infrastructure investments.
the problem
Water, schools, healthcare, roads, and electricity were assessed with fragmented standards and uneven regional data. Teams struggled to compare needs consistently, explain priorities, and connect recommendations to available budgets.
what i owned
- defined the product model connecting standards, data, and capital decisions
- aligned government, finance, technical, and regional stakeholders
- prioritized scoring, ranking, filtering, and budget-allocation workflows
- led rollout and adoption support across multiple regions
- managed roadmap, delivery scope, and stakeholder sign-offs
evidence and decision sequence
observed signalseverity, population exposure, and delivery feasibility
product decisioncreate an explainable priority score
how it was checkedinspect the factors behind each rank
observed signalranked need plus available capital
product decisiontest recommendations against budget constraints
how it was checkedfinance review and regional sign-off
standard → rule → recommendation
make every rank traceable to a domain rule and a constraint
standardrequired service level
Documented threshold owned by the domain team.
gap signalcurrent state vs standard
Comparable evidence across regions and categories.
priority logicseverity + exposure + feasibility
Visible factors instead of an unexplained score.
decisionrank within the available budget
Recommendation connected to an actual planning constraint.
key decisions and trade-offs
make recommendation logic explainable
alternativesUse a black-box composite score or show raw indicators without prioritization.
what we choseCreate traceable scores with visible category inputs and recommendation explanations.
whyPublic-sector users needed to defend why one investment ranked above another.
trade-offThe model was less compact, but substantially easier to audit and adopt.
connect recommendations to planning constraints
alternativesShip a reporting dashboard or build a full financial-planning suite.
what we chosePair ranked recommendations with category filters and budget constraints.
whyInsight only mattered if it could carry into a real allocation decision.
trade-offThe workflow required more product design than a dashboard, but was more actionable.
product interface
ranked recommendationssanitized portfolio reconstruction
all regionswater + schools + roadsbudget constrained
ranked projectregionscore
01water pipelineregion a92
02rural schoolregion b87
03road rehabilitationregion c81
why #1high service risk
critical standard gap across 14 communities with feasible delivery inside the current budget.
$8.5m allocated · $4m remainingresults
decision support
Government teams compared priorities across more than 4,000 communities.
capital planning
The recommendation system supported more than $40 million in allocation decisions.
product logic
More than 60 standards became maintainable, explainable scoring rules.
measurement notes
Definitions shown so the claims can be discussed and challenged.
$40M+ decisions supported
Cumulative capital-allocation decisions in workflows where the recommendation product informed prioritization.
4,000+ communities
Communities represented in the regional datasets evaluated by the product.
60+ standards
Documented infrastructure standards translated into explicit, maintainable product rules.
reflection
Adoption depended as much on explainability and training as on the scoring model. A technically correct recommendation became valuable only when users could understand and defend it.