Data & Analytics for State Agencies
Data and Analytics for State Agencies
Data platforms, automated federal program reporting, and governed dashboards for state departments answering to funders, auditors, and the legislature.
A state agency answers to more audiences than almost any other part of government. A federal funder wants program data on its own schedule, in its own layout, using its own definition of who counts as served. The legislature wants a performance measure that was written into statute years ago and has never been revisited. The budget office wants something close to that number, calculated a different way, for the next hearing. The public wants a dashboard. Meanwhile the system holding the underlying records was built to process transactions, applications, cases, licenses, claims and payments, not to answer questions about them.
What fills that gap is usually people. An analyst pulls an extract, applies exclusion and counting rules that live partly in a spreadsheet and partly in their head, and produces the submission before the window closes. It works until that person retires, the funder revises its file specification, a program is reorganized, or an auditor asks how a published figure was derived and nobody can reconstruct it. The reporting burden is real, but the deeper problem is that the agency cannot prove its own numbers.
What these environments look like
- A system of record built for casework or transactions, often decades old, that offers scheduled extracts rather than any usable analytic interface
- Federal reporting tied to each funding stream the agency administers, with separate data dictionaries, validation rules, submission portals, and deadlines
- Single Audit and subrecipient monitoring obligations under federal uniform administrative requirements, which put grant and expenditure data in scope for review
- Statewide systems where the records are entered by county, district, or provider staff the agency does not supervise, so data quality depends on people outside the org chart
- Restricted categories sitting alongside general program data, including federal tax information in eligibility systems, protected health information, substance use treatment records, criminal justice data, and student records, each with its own access rules
- A central IT or data office setting enterprise standards, a shared cloud tenancy with chargeback, an architecture review board, and a security review gate before anything is provisioned
- Several partial warehouses and reporting tools accumulated over the years, each built by a different program or vendor, none of them retired
Where HQT helps
- Designing and building a data platform on the state's approved cloud, layered so that restricted categories stay separated and access is granted by program and role rather than by handing out the whole warehouse
- Automating federal program reporting: extract, transform, validate against the funder's published rules before submission, and retain a record of what was sent and how each figure was calculated
- Establishing a metric dictionary so that the measure reported to a funder, quoted in a budget hearing, and published on a public dashboard all trace back to one documented calculation
- Data quality monitoring with feedback loops to the local and provider staff entering records, since correction at the source is the only fix that lasts
- Dashboards and scheduled reporting for program managers, agency executives, and oversight bodies, plus open data publication with suppression and aggregation rules applied in the pipeline rather than by hand
- GIS work including statewide layers assembled from county sources, service area and access analysis, and the spatial reporting many federal programs expect
- Machine learning where it is defensible and worth the effort, with responsible AI practices around it: an inventory of uses, documentation of data and intended purpose, testing for performance across populations, human review of consequential decisions, and alignment with the NIST AI Risk Management Framework and the state's own AI policy
How an engagement usually starts
Work of this kind typically begins with a specific event rather than a general interest in analytics. A federal monitoring review or Single Audit raises a finding about reported data. A funder changes its file specification and the existing process cannot produce it. The legislature directs the agency to publish a public dashboard on a fixed timeline. A legacy system is being replaced, and someone realizes the replacement program has no plan for the reporting the old system fed. A newly appointed chief data officer needs to know what the agency actually has.
The first step is narrow on purpose. HQT traces one measure end to end, from the field where the record is entered, through every extract and transformation, to the number that appears in the submission or on the dashboard, and documents where it changes meaning and where it cannot be reproduced. That single trace usually reveals the structural issues clearly enough to scope the rest. HQT's people have professional experience doing this inside state program environments, and the work is planned with the agency's privacy counsel, records officer, and the state security office involved from the design stage, because those constraints determine the architecture rather than merely reviewing it.