Consulting that takes you a hop forward.

Four shapes of engagement.

Scope is written down before anything starts, and you keep everything that gets built.

1. 01 Make the numbers defensible.

   Reimbursement and benchmark models, price-transparency data, payer-facing analysis, outcomes and effect-size work. Sourcing you can hand to a skeptic without bracing for it. Where a figure has no anchor in a primary source, it gets listed as a gap rather than estimated into place.

2. 02 Build the platform underneath them.

   Ingestion, a layered warehouse or lakehouse, a semantic model the business can query without an analyst in the loop, refresh orchestration, row-level security, and the governance that keeps the whole thing from rotting in eighteen months. Those patterns are portable. The vendor is a decision, not a religion — and if you have already standardized on something, the answer is almost always to build well inside it rather than sell you a migration.

3. 03 Ship the tool people use.

   Next.js and Postgres on Vercel: auth-gated internal applications, importers, export packs, CRMs. Built by the same person who modeled the data, so the model and the interface never drift apart.

4. 04 Sit in the room for the decision.

   Fractional analytics leadership. Roadmap, governance, vendor evaluation, and the translation between data, operations, and leadership that most teams are actually short of. Most organizations do not need another report.

Where the bench is deepest: Microsoft Fabric and Power BI. I architected and administered a production Fabric environment end to end — medallion lakehouse, PySpark notebooks, OneLake shortcuts, capacity sizing, workspace and gateway governance, refresh orchestration — and owned the Power BI semantic layer above it: DAX, Power Query M, row-level security. If you are already on Fabric, that is the fastest start available to you. If you are not, it is not a reason to move.

Also built on: Postgres and Next.js on Vercel · Spark SQL and PySpark · Python pipelines · T-SQL · Excel and Power Automate where that is genuinely the right tool. Chosen against your licensing, your staffing, and what your data is allowed to touch — in that order.

You get the code, the models, and the documentation. Not just a slide deck. Those four are also the ground the next section stands on — the AI work is only ever as good as the platform and the definitions underneath it.

Where AI actually pays.

The fifth thing people ask for, and the first one that fails when the four above are missing. AI work is a layer, not a project — this is how I sequence it onto a platform that can hold it.

1. The roadmap, not the pilot.

   A sequenced plan tied to what your data can support today: which use cases are ready, which are blocked on the model or the pipeline underneath them, and what has to be true before the blocked ones move. Programs stall on data readiness far more often than on model choice, so the roadmap starts there and says plainly which items are not ready yet.

2. Workflows, not chatbots.

   The automations that hold up are the unglamorous ones: the document somebody retypes every week, the queue a person sorts by hand, the summary written the night before the meeting. Each gets measured against the manual baseline it replaces, with a human in the loop wherever the output carries clinical or financial weight.

3. The governance question first.

   In regulated work the first question is not which model. It is what data leaves the building, under whose BAA, retained how long, and trained on by whom. Vendor terms get read before the pilot rather than after it, and anything that cannot run inside your existing agreements gets named as such while it is still cheap to change.

If the honest answer is that your data is not ready for the use case you want, you get that in writing, with the sequence that would make it ready.

The terms, before you ask.

These are the terms on the signed statement of work, not a posture written for this page.

Basis

Time and materials, billed against a written estimate.

Cap

A written not-to-exceed on every engagement. The invoice cannot surprise you.

The 80% call

When billed hours reach 80 percent of the estimate, I tell you in writing before continuing. That is a clause, not a courtesy.

Invoicing

Itemized by workstream with the hours attached, not a lump sum. Net 15.

Ledger

A running hours ledger you can read start to finish, with every hour tagged to the invoice it was billed on.

IP

Assigns to you on full payment. All of it.

Exit

Either of us can end the engagement on seven days' written notice.

Changes

Scope changes by written change order only.

The budget position is printed on the face of every invoice I send — hours worked against the estimate, dollars billed against the cap — so you never have to ask where you stand. Including on the invoices where the position is uncomfortable.

Ren Rooney.

Ten-plus years building data products in healthcare and consulting. Most recently, Director of Analytics at a multi-site behavioral health provider, where I architected and administered the Microsoft Fabric environment — medallion lakehouse, PySpark notebooks, OneLake shortcuts, capacity sizing, workspace and gateway governance, refresh orchestration — and owned the Power BI semantic layer above it.

Before that, four years in management consulting in Chicago: staffing and spend analysis across managed IT services engagements, technology scorecards benchmarking client operations against their peers, and the BI layer behind an annual industry technology benchmark. I started in healthcare revenue-cycle analytics at an academic medical center, building recurring dashboards across several revenue-cycle systems and presenting fiscal-year trend findings at quarterly reviews.

Data-Roo is the independent practice. It runs on my own accounts and my own infrastructure and has never touched an employer's.

Published

Co-author, a peer-reviewed study on the effectiveness of brief partial hospitalization for depression and anxiety. the Behavior Therapist, Vol. 49 No. 1, January 2026.

Spoke

Four Becker's Healthcare panels across 2025 and 2026: behavioral health data strategy, EHR data silos and integration, and health IT tooling.

Degree

B.B.A., Business Analytics and Information Systems, University of Iowa.

Microsoft Fabric · Power BI · DAX · PySpark · T-SQL / Spark SQL · Python · Next.js · Postgres · Vercel

Tell me what is broken.

The report nobody trusts, the platform decision you are stuck on, the number two systems disagree about. Send the problem, who else is involved, and what "done" would look like.