VECTR is not AI.
It's code.
VECTR is New Zealand's economics platform: scenario analysis across the whole economy, run as a deterministic program your team licenses in-house — or commissioned from ours. It isn't a model that guesses. Ask the same question twice and you get the same answer twice — and every run carries an ID anyone can check weeks later.
A program, with conveniences around it.
At the root, VECTR is a standalone command-line program — the CLI. That is the product: the engine that does the economics. Everything else exists to make that engine easier to reach.
VECTR CORE™ — the CLI
The engine. A standalone command-line program that does the actual economic modelling. Deterministic, scriptable, and the same code every other way in eventually calls.
The Console — a GUI
A graphical interface for building a run without the command line: set the parameters, explore the map, read the result.
The API — hosting
A REST API so VECTR can run inside your own pipeline: self-hosted, or through the official danalytics.co.nz API, which is live today. Same code, reachable over the network.
An AI assistant — optional
The smallest layer. Bring your own AI and your own API key — Danalytics Consulting Ltd supplies the surrounding software, never the model itself. The assistant turns a scenario described in plain English into model settings; it never touches the economics. Switch it off and VECTR loses no capability at all.
It's code — and code can be tested.
A deterministic program can be repeated, checked and audited in ways a model that guesses cannot. Because the code is checkable, so is every number it produces.
Deterministic
The same scenario returns the same result, every run. No temperature, no drift, no surprise.
Inspectable
Every matrix, coefficient, and calibration parameter can be read. Source-available under licence — not a black box.
Tested
Convergence is measured and reported; the maths is specified; the pipeline is reproducible. See how we test accuracy, precision and validity →
VECTR is software, end to end — a Python engine with a TypeScript/React interface wrapped around it. Not one number in a result is produced by AI. Every figure is computed by deterministic code you can read. And every run mints its own ID, printed on the report as a 2D barcode: when a stakeholder asks about that number weeks later, the run is still there to open. Prove It is coming next — ask any figure on the report where it came from and VECTR returns the full derivation behind it, back to checkable public data.
Trust is welcome, but not required. Every result rests on proof and evidence you can test, not on faith.
The smaller part.
Danalytics is not anti-AI. AI belongs in three places here, and the engine is not one of them.
An optional assistant
An AI assistant can turn a plain-English Scenario (“a 20% cut to dairy exports in Southland”) into the right model settings. Setting model parameters is the whole job: the economics are untouched. You can ignore the assistant entirely.
Refined context, not our own model
Danalytics does not train or own the model behind the assistant. What Danalytics supplies is refined context: the New Zealand economic detail and the model settings the engine expects. The model stays yours, running under your own API key.
AI helped build the code
The codebase was reviewed and refactored with AI coding tools and modern engineering methodology, and this site was written with AI assistance under human review. Disclosed in full on the data & legal page.
If you never want AI anywhere near your analysis, VECTR is still the whole tool.