01
Date logic
Reporting Window Builder
Create fiscal, rolling, custom, and like-for-like comparison periods without hidden date arithmetic.
- Inclusive day counts
- Fiscal quarter alignment
- Leap-year warnings
Local CSV change audit
Catch changed records and schema drift before they reach a healthcare report.
Compare two CSV extracts locally in your browser. No uploads, accounts, or analytics.
Runs locally in your browser. No uploads, accounts, or analytics.
Calculation record
Reviewable
- Inputs
- Visible
- Assumptions
- Stated
- Outputs
- Exportable
- Data transfer
- None
Purpose
Decision support, not a black box
New: save a monthly check and explain its count changes
Open toolkit
Six local, reviewable workflows
Define the period, verify what changed, then test whether capacity can absorb the work. Each utility can stand alone.
02
Data quality
Extract Change Auditor
Compare two CSV extracts in your browser and separate schema drift from added, removed, and modified records.
- Composite record keys
- Duplicate-key detection
- Safe CSV export
03
Operations
Waitlist Capacity Planner
Compare current capacity with a proposed change and see the implied backlog and queue-wait trajectory.
- Current versus planned path
- Transparent fluid-queue model
- Required-capacity estimate
04
Conformance
Reporting Results Checker
Compare aggregate pipeline outputs with versioned synthetic edge-case contracts without uploading the files.
- Exact key and integer checks
- Actionable mismatch diagnostics
- Privacy-safe analysis receipt
Deliberate boundary
Useful without pretending to be certified.
The toolkit is not a submission engine, clinical system, queueing simulation, or replacement for local policy. It makes a narrow set of calculations inspectable and reproducible.
For deterministic reporting test cases, see the companion Health Data Edge Cases project.
Tool 01 · Date logic
Reporting Window Builder
Turn an as-of date into an explicit current period and a defensible comparison period.
Best for
Operational dashboards, fiscal reporting, and partial-period comparisons
Result
Awaiting inputs
Aligned reporting windows
Your dates will appear here
Choose a period and comparison method. The result will show exact inclusive boundaries and flag calendar differences.
Current period
— —Comparison period
— —| Period | Start | End | Inclusive days |
|---|---|---|---|
| Current | — | — | — |
| Comparison | — | — | — |
Check before using
Calculation rules and limitations
Dates are calculated as calendar dates in UTC to avoid daylight-saving shifts. Day counts include both the start and end date.
Like-for-like comparison subtracts one calendar year from each boundary and clamps February 29 to February 28 when required. It does not adjust for holidays, weekdays, fiscal calendars with 4-4-5 periods, or local reporting exclusions.
Tool 02 · Data quality
Extract Change Auditor
Compare two CSV snapshots and distinguish schema drift from record-level change.
Local processing
Selected files stay in this browser tab until cleared and are never uploaded by the toolkit.
Result
Awaiting extracts
Change summary
Load two CSV snapshots
Use your own files or the synthetic example. Shared non-key columns are compared as text after optional trimming.
Schema changes
- Added columns
- None
- Removed columns
- None
- Type shifts
- None detected
Key integrity
- Ambiguous duplicate keys
- 0
- Rows with blank keys
- 0
- Unchanged records
- 0
Excluded from record comparison
Analysis receipt
Includes source fingerprints and aggregate findings. It
omits filenames, key-column names, row keys, and cell values.
Changed-record preview
Showing material differences
| Key | Status | Changed fields | Detail |
|---|
Comparison rules and limitations
Added or removed schema columns are reported separately and do not make every record appear changed. Only columns present in both files are compared at record level.
Duplicate and blank keys are excluded rather than guessed. Values are compared as text; inferred type shifts are screening signals, not a formal schema. CSV exports prefix spreadsheet formula characters to reduce formula-injection risk.
Comparison runs in a same-origin browser worker. The interface retains only aggregate findings, a 100-record preview, source fingerprints, and—when it fits—a complete change-log download. Downloads above 250,000 detail rows or 48 MB are refused rather than silently truncated.
Tool 03 · Operations
Waitlist Capacity Planner
Make the backlog consequence of demand and service capacity visible before committing to a plan.
Planning proxy
This is a deterministic fluid-queue model, not a patient-level wait prediction.
Result
Calculating
Capacity outlook
Projected waitlist
People remaining after each week
Current capacity (dashed)
Proposed plan (solid)
Target line
Milestone view
Target assessment
| Week | Current backlog | Planned backlog | Planned wait |
|---|
Model assumptions and limitations
The model adds a constant average number of referrals each week, serves up to the available capacity, and carries remaining work forward. The wait proxy is backlog divided by that week's service capacity. Sustainable capacity is never reported below average weekly arrivals under these assumptions.
It assumes a single queue, no prioritization, no cancellations, no seasonality, no capacity variation, and no differences in case complexity. Validate any decision with local distributions, staffing constraints, and patient-level flow.
Tool 04 · Conformance
Reporting Results Checker
Test aggregate outputs from another reporting stack against one versioned, synthetic Health Data Edge Cases contract.
Local processing
Selected result files stay in this browser tab until cleared and are never uploaded by the toolkit.
Result
Awaiting results
Contract comparison
Load two result exports
Use your own aggregate outputs or one of the synthetic examples. The checker requires the complete expected key set and exact integer values.
Exact diagnostics
Showing result differences
The detailed CSV can contain operational keys and values. Review it before storing or sharing it.
| Type | Key | Status | Expected | Actual |
|---|
Privacy-safe reproducibility record
Includes source hashes, contract provenance, case ID, and
aggregate counts—not filenames, keys, IDs, values, or row diagnostics.
Selected synthetic contract
Edge case
What plausible logic gets wrong
Expected resolution
Five-minute tutorial: fail, diagnose, and correct
- Select any bundled edge case.
- Choose Use deliberate mismatch and review the single wrong-value diagnostic.
- Read the selected case’s naive failure and expected resolution below the checker.
- Choose Use matching example to replace the changed value with the contract value.
- Confirm the match, then download the aggregate receipt. Use the detailed CSV only when row-level diagnostics are appropriate to retain.
A pass proves only that these two exports match this synthetic case. It does not certify the pipeline, source data, or untested logic.
Comparison rules, privacy, and limitations
Column names and order must match the public external-results contract exactly. Keys cannot be blank or duplicated. Values must be whole-number text without decimals, exponents, NaN, or infinity. Missing, unexpected, and incorrect results all prevent a match.
The checker reads only aggregate result exports. Do not select patient-level source data. Detailed diagnostics can contain operational identifiers and values; analysis receipts deliberately omit them. Source hashes are linkable fingerprints, not anonymization.
Audit utility · Reproducibility
Receipt Inspector and Replay
Check an existing toolkit receipt for internal consistency, replay supported calculations, and optionally match its source fingerprints.
Important boundary
Internal consistency is not proof of identity, authorship,
approval, or time of creation.
Result
Awaiting receipt
Receipt assessment
No receipt loaded
Select a toolkit analysis receipt or load the synthetic example. The file stays in this browser tab.
Consistency checks
—
Source fingerprints
—
| Role | Recorded bytes | Selected file | Result |
|---|
What this result means
“Internally consistent” means the receipt structure, recorded digest, and any supported replay agree. It does not authenticate the receipt, its creator, or its timestamp, and it does not prove that unselected source files match their fingerprints.
Compatibility and verification boundaries
The inspector recognizes published receipt profiles from toolkit versions 0.2.0 through 0.8.0. Reporting-window and capacity receipts are recalculated; extract and conformance receipts are checked by schema, digest, and optional exact source-file fingerprints.
Source verification compares the selected file’s exact byte count and SHA-256 digest. A match does not establish where the file came from or whether its contents are appropriate for a real decision.