Patent Pending

Tableau to Power BI with verified semantic equivalence

Navis converts your Tableau workbooks to Power BI projects with a formally proved composition theorem, multi-context validation, and a tamper-evident evidence capsule. Zero data access. Every conversion auditable.

Designed for Regulated Industries
Financial Services · Healthcare · Government · Defence · and other compliance-driven environments

Zero data access. Airgapped by design.

Navis never connects to Tableau Server, Tableau Cloud, Power BI Service, or any database. It operates exclusively on exported .twb and .twbx files — on your premises, in your VPC, with no network egress.

For financial services, healthcare, government, and classified environments, this is non-negotiable. We don't see your data. We don't need to.

  • Parses local Tableau XML only — no server connectivity
  • Zero outbound network calls
  • Runs in fully airgapped environments
  • No credentials stored or transmitted
  • No telemetry, no phone-home
  • No LLMs, no AI, no randomness — deterministic translation

Formal grammar parser — Navis parses Tableau's calculation language using a formal grammar, not regex or heuristics. Every calculation is decomposed into semantic atoms, matched against a pre-validated library, and reassembled using proved composition rules.

Multi-context validation — Tableau and Power BI evaluate the same expression differently depending on context: row-level, aggregate, filtered, totals. Navis validates every mapping across multiple evaluation contexts. A mapping that works in one context but fails in another is rejected.


Proved, not promised.

Navis relies on a formally proved composition theorem and a library of pre-validated semantic atoms — not LLMs, not heuristics, not best guesses.

Semantic atoms

Every Tableau expression is broken into minimal, indivisible patterns — each individually validated by executing both the Tableau original and the DAX equivalent against a reference dataset, across multiple evaluation contexts, until results match exactly.

Composition theorem

Formally proved: if atoms A and B are individually validated, and they're connected by a pure operator (arithmetic, scalar function), then the composed expression is guaranteed correct — without testing the specific combination.

This extends a finite library of validated patterns to an unbounded space of composed expressions. Competitors without a composition theorem need individual validation for every possible combination — which is combinatorially infeasible.

Example

// Tableau expression ZN(SUM([Sales])) / WINDOW_AVG(SUM([Sales])) // Decomposes into: Atom 1: ZN(SUM([Sales])) ← validated Atom 2: WINDOW_AVG(SUM([...])) ← validated Operator: division ← pure // Composition theorem guarantees // the DAX equivalent is correct

Four confidence levels

Exact

Individually validated against reference data across all evaluation contexts.

Composed

Derived from validated atoms via the composition theorem. No individual test, but correctness guaranteed.

Deterministic

Structural elements (sets, groups, parameters) with direct, non-expression mappings. No validation needed — the mapping is definitional.

Degraded

Cannot be fully translated. Semantic loss explicitly documented in the degradation report.

What Navis converts: Calculated fields (arithmetic, logical, string, date), LOD expressions (FIXED, INCLUDE, EXCLUDE — up to 5 levels of nesting), table calculations (WINDOW_AVG, RUNNING_SUM, RANK, LOOKUP), parameters, sets, groups, filters, hierarchies, field descriptions, and dashboard layout. Full list under NDA.

Calculations: verified. Visuals: best-effort, human review required.

Navis provides verified semantic equivalence for calculations, data structures, and data model elements. Dashboard layout, chart types, and visual structure are converted on a best-effort basis. Visual formatting — colours, fonts, padding, reference line styling, tooltip configuration, conditional formatting, etc. — requires human review and adjustment in Power BI Desktop. Automated conversion is a starting point for visuals, not a finished product. We strongly recommend a human-in-the-loop review for all visual elements before production deployment.


Tamper-evident evidence capsule.

Every conversion produces a Conversion Evidence Capsule — a self-contained, cryptographically sealed package that lets your team and regulators verify correctness independently.

📥
Input artefact
Original Tableau workbook (TWB/TWBX)
📤
Output artefact
Generated Power BI project (PBIP/PBIR)
Semantic equivalence records
Per-element: source expression, DAX, confidence level, mapping rule
⚠️
Degradation report
Every unmappable element listed with triage category
📋
Execution trace
Every mapping decision in deterministic order
🔐
Cryptographic seal
SHA-256 hash chain + digital signature

Deterministic

Same input + same converter version + same rule set = identical capsule hash. Always. A third party can re-run the conversion and verify the hash matches.

Self-contained

Verification requires no access to Navis, your systems, or any external service. The capsule contains everything an auditor needs.

Tamper-evident

Every file in the capsule is hashed. The hashes are chained. The chain is digitally signed. Modify any file — the signature breaks. No one can strip degradation entries, inflate confidence levels, or alter the conversion results without detection.

Assured boundary

The capsule explicitly defines what is auditable (conversion engine, rules, proof) and what isn't (UI, scheduling, auth, storage). Only artefacts inside the capsule are considered audit evidence.

"Our auditors required proof that every Tableau calculation was correctly translated. Navis's capsule gave them exactly that — without granting access to our migration tooling or systems."

— VP Data Governance, Fortune 500 bank


Know what won't migrate before you commit.

Navis produces a degradation report before conversion. Every Tableau feature that cannot be automatically translated is listed with a triage category.

Category Meaning Example
Implementation gap
Addressable by updating the application WINDOW_CORR with custom partitioning
Platform limitation
No semantic equivalent in Power BI MODEL_QUANTILE
Manual required
Stub generated, scope of work identified RAWSQL_BOOL with custom SQL

Silence is not allowed. Every successfully converted element is explicitly marked. Every unconverted element is explicitly listed. The absence of degradation is never implied by omission. Delivered as part of the evidence capsule — you decide whether to proceed or remediate first.


Days, not months.

Conversion runs in seconds per workbook. The time you spend is on review and validation — not waiting for the tool. A 200-workbook estate converts in minutes; the migration completes in weeks.

No human-in-the-loop (except where needed)

Only genuinely unmappable features require manual attention. The degradation report tells you exactly where — and exactly how much work remains. Everything else is automatic.

Iterate with confidence

Need to adjust mappings or re-run after a rule update? Same inputs + same converter version = same output, every time. Diff to see what changed.

One workbook, one conversion, zero rework

Navis converts calculated fields, LOD expressions, table calculations, parameters, sets, groups, filters, hierarchies, field descriptions, and dashboard layout in a single pass. Connect your data in Power BI Desktop and you're running.

Looker target under development

The same composition theorem and evidence capsule framework is being extended to Tableau → Looker conversions. The validation methodology is platform-pair specific; the correctness guarantees are universal.

Deploys in minutes. Python 3.9 and pip install. No infrastructure, no servers, no configuration. Runs on your laptop, your build server, or your VPC — wherever your Tableau exports live.


Already migrated? Find out if it's right.

Migrated to Power BI using another tool, an LLM, or by hand? Run consult against your original Tableau workbook. In seconds, you'll know exactly which calculations are correct and which need attention.

Every calculation, one clear answer

The consult report covers every Tableau calculation in the workbook: correct, equivalent (written differently but producing the same results), needs review, or missing entirely. No ambiguity, no guesswork.

Know what kind of review you need

When something differs, the report tells you why — a column was renamed, the formula takes a different approach, extra logic was added, or the wrong aggregation was used. Your team spends time on real issues, not chasing false alarms.

Fix it yourself, or let Navis fix it for you

The report tells you what differs. If your team can fix it, great — you've saved a full conversion licence. If you'd rather have the validated DAX, licence a conversion for just the affected workbooks. Pay for the diagnosis and the cure separately.

Your Power BI project stays yours

Consult reads your PBIP folder (text files — open it in Power BI Desktop and save as a Project first if you have a .pbix). Nothing is modified, uploaded, or sent anywhere. Same airgapped operation as the rest of Navis.

Start free, go deeper when you need to. analyse shows your estate's complexity at no cost. consult validates an existing migration at a fraction of conversion cost. convert produces the full Power BI project with signed evidence.


Ready to prove your migration?

Install Navis on your own machine. Run it against your own Tableau workbooks. See the Power BI project, the evidence capsule, and the degradation report — before you talk to us.

pip install cepheus-navis cepheus-navis analyse your_workbook.twbx cepheus-navis convert your_workbook.twbx

No data access. No network calls. No sign-up. Evaluate on your terms.

When you're ready to convert at scale, request a licence.

For regulated industries, financial services, healthcare, and government.