Percentry Data & AI Services
Data engineering, analysis, and practical AI for healthcare price and payment data.
We load and clean hospital price files, payer Transparency in Coverage files, and 835 remittances, then analyze them or build tools on top. Percentry is one person: James Tudor, a .NET developer who has worked hands-on with these files, payer contracts, and the employer-benefits side, scopes, builds, and checks every project himself.
Who this is for
- Hospitals and health systems
- Payers and TPAs
- Employers and benefits brokers
- Price-file vendors
- Healthcare consultancies
- Health law firms, on matters about posted prices or payments
What we do
Pricing and reimbursement analysis
Posted rates, percent of Medicare, and comparisons across hospitals or payers. We show the source file and the method behind each figure. Medicare comparisons cover hospital facility services for now; physician fee schedule prices are not built yet.
Example projects
- Compare a market's hospitals by posted rates for common outpatient services
- Show a hospital's negotiated rates as a percent of Medicare, with Medicare adjusted for that hospital's wage index
- Line up the negotiated rates several payers post for one hospital, code by code
Price-transparency data engineering
Hospital machine-readable files and payer Transparency in Coverage (TiC) files are very large and inconsistent. We load, clean, and standardize them into tables you can query.
Example projects
- Load every hospital price file in a state or market into one database
- Pull the in-network rates for your hospitals from the TiC files of the payers you name
- Set up a scheduled refresh that flags when a hospital's file changes
835 and claims data pipelines
We parse 835 remittances and replay final adjudication, so each claim counts at its final amount after reversals and corrections. 835s contain protected health information, so this work runs inside your environment with our tools and the files stay there. Percentry doesn't receive patient data.
Example projects
- Turn a year of 835s into paid amounts by payer, code, and claim, with patient names and member IDs removed
- Check paid amounts against your contract rates for the payers and services you choose
- Match what payers paid you to the rates they post in their TiC files
Dashboards, reports, and custom tools
Dashboards and reports for managed care, finance, and benefits teams, built in your BI tool or as a .NET web app. Custom .NET tools for jobs that run on a schedule or inside your own systems.
Example projects
- A market-rate dashboard for a managed care contracting team
- An employer report showing the plan's negotiated rates at local hospitals as a percent of Medicare, from public TiC files
- A tool that fills in the CMS allowed-amount fields of a hospital's price file from the hospital's own 835s, inside the hospital's or its vendor's network
Practical AI on pricing data
AI for narrow jobs on pricing data, never on patient data. One example is drafting plain-language service descriptions for your staff to check and approve.
Example projects
- Sort free-text rate notes and methodologies into consistent categories
- Match codes and service descriptions across files that label them differently
- Flag unusual rates for a person to review
What's already built
Percentry built these tools in-house, and client projects can start from them. The sample price file review comes from the same code.
Hospital price file checker
Reads CMS v3 price files in JSON and CSV, including tall and wide CSV layouts. On 133 Georgia files it flagged every error type that CMS's own validator (version 1.10.8) reported, and it checks more. Both agreed on which 10 files had errors. Passing these checks doesn't show that a file's prices are right. Percentry isn't affiliated with CMS.
Georgia statewide price file check
A crawler finds each hospital's cms-hpt.txt on its public website, follows its links, and checks NPIs against the public NPPES registry. It names itself, follows robots.txt, pauses between requests, and doesn't get around a refusal. In September 2026 it covered the websites of Georgia's 166 non-federal hospitals; we reviewed 133 price files from 129 of them. For most of the other 37, the site refused the request or had no cms-hpt.txt.
835 remittance parser
Replays final adjudication, including reversals and corrections. It drops patient names and member IDs and replaces claim numbers with salted hashes. Its output is still treated as patient data.
Allowed-amount calculator
Fills in the allowed-amount figures CMS requires in hospital price files (the count, median, and 10th and 90th percentiles) from the hospital's own 835 data. It changes only those fields and leaves the rest of the file as it was. In a stress test it filled all 677,493 percentage and algorithm rates in a 903 MB price file in 7.6 seconds. It runs on the hospital's or vendor's own servers, so the 835 data stays there.
Medicare reference prices
Built from public CMS outpatient (OPPS) and inpatient (IPPS) files and adjusted for each hospital's wage index. A second, full-Medicare figure adds each hospital's IME, DSH, and uncompensated care payments. Both are used to show hospital prices as a percent of Medicare.
How we use AI
- AI handles narrow tasks under set rules, such as sorting rate notes into a fixed list of categories or matching descriptions.
- AI output can be wrong. A person checks it before it feeds a figure or finding we deliver, and for large batches we hand-check a sample and report the error rate.
- Calculations such as percent of Medicare use fixed formulas in code, so each number can be rerun and checked. AI doesn't produce figures.
- AI tools never see patient data. 835 work runs in plain code.
- We use AI only through services whose terms bar training on customer data, and we name them before we start.
- Before we use AI on your files, we tell you which models and vendors we would use, and you can say no.
How an engagement works
- Talk through the question. A call about what you need to know, what data you have, and who will use the result.
- Agree on scope. A written scope with data sources, deliverables, timeline, price, and where your data will be processed. Patient data stays in your environment; Percentry doesn't receive it. A first project can be small, such as a percent-of-Medicare comparison for one market on codes you choose.
- Build and check. Where public data can answer part of the question, we start there. Then we load your data, build the analysis or tool, and review results with you as we go. Your data is used only for your project.
- Hand over. You get the deliverables, a written note on the method, and the code or tool where the scope includes it. We return or destroy the data you sent within an agreed period and confirm it in writing.
Kept separate from Price File Review
Percentry's main service is an independent Price File Review, and consulting must not compromise it. We never review a price file we helped build, including any file whose allowed amounts our calculator produced. If we consult for a hospital on its price file, we won't review that hospital's file. If we have worked for the vendor that produced a hospital's file, the review says so. We don't build or sell tools that produce a hospital's full price file.
Outside our scope: work outside healthcare price and payment data, legal advice, and negotiating contracts for you.
Have a pricing data question?
Tell us what you're trying to answer and what kinds of data you have. Please don't send patient data or files. James Tudor will reply with whether Percentry can help and what a first project could look like.