Business modelling
beyond reporting —
without the complexity and cost.
Not just dashboards that report on your business — living reports and data surfaces that model it. Interactive, always current, and yours to run.
In practice
What that means
for you.
Before the technical detail, here's what a live, modelled view of your business actually does for you.
Your numbers are never out of date
Change a price, a target, an assumption — and everything that depends on it updates on the spot. No rebuilding a report, no overnight refresh.
We call this a reactive model: your reports behave like a spreadsheet, so you can try a scenario and watch the whole picture respond at once.
Ask in business terms, get an answer
Slice by region, product or month and the totals roll up by themselves. The questions you'd ask in a meeting are the ones the model answers — no query to write.
Underneath is a dimensional model — what the industry calls MOLAP — so your data keeps its natural shape instead of being flattened into rows and columns.
Plan ahead, don't just look back
Type in new figures, build budgets and what-ifs, and let the model recalculate around them. Reporting on the past and planning the future finally live in one place.
This is write-back: the surface isn't read-only, so planning and reporting share one model.
Foresight, built in
Forecast next quarter, surface the outliers, and see what's driving the change — without bolting on a separate data-science tool.
A library of analysis blocks — forecasting, regression, clustering, anomaly detection — runs directly over your data.
Try anything, safely
Spin up a private what-if over your shared numbers, change whatever you like, then keep it, merge it back, or throw it away. The shared data isn't touched until you say so.
We call this stacking: your scenario lives in its own layer above the committed data, so exploring never risks the source.
It's yours, end to end
Runs on your own servers — your data never leaves the building, and the bill doesn't grow with every seat. Start as a quiet engine behind today's tools, or grow into a full workspace. Serving many customers? Each sees only their own governed data.
Ships self-hosted with a standard SQL interface, so it drops into the stack you already have.
Build & extend with AI
Describe what you want. AI builds it.
Every capability here is declarative — plain, structured text rather than a maze of menus and dialog boxes. Describe the report, model, or scenario you have in mind in ordinary language, and an AI assistant turns it into a working model. Need a change? Ask for it, and watch it update.
Because the whole language is explicit and AI-readable — and the compiler answers mistakes with plain-language hints — assistants write, extend, and repair your models reliably. Built for the agentic era.
Where it fits
A gap between
reporting and planning.
SQL-based BI tools are built for querying relational data. They're fast to set up, but weren't designed for write-back, real dimensional semantics, or reactive computation. Enterprise planning suites like Oracle Hyperion and Anaplan have that depth — at six-figure implementation cost, months of onboarding, and a GUI-first design that resists automation. Most teams building serious data products fall between the two.
BI & Reporting
Looker · Power BI · Metabase · Sigma
- SQL-native, relational queries
- Excellent for historical reporting
- No dimensional hierarchies or rollups
- No reactive rules or write-back
- Can report; can't model
psimulang
Reactive MOLAP · Developer-native
- Dimensional hierarchies with rollups
- Reactive rules and write-back
- Interactive reports and dashboards
- Integrated analytics and forecasting
- Developer-native · AI-readable · SQL-integrated
Enterprise Planning
Oracle Hyperion · Anaplan · SAP BPC · IBM TM1
- Multi-year contracts, six-figure licensing
- Specialist implementation partners required
- Months to first results
- GUI-first; resists automation and version control
- Purpose-built for large finance-team rollouts
psimulang is designed for the team with real modelling requirements and the engineering maturity to build and own the layer themselves.
What you get
Six capabilities
in one coherent design.
REACTIVE MOLAP
Declare dimensions, orthos, and rules. Polytope derives the dependency graph and recomputes exactly what changed — in order, every time, as an atomic surface readers never see half-updated.
INTERACTIVE REPORTS
Tables, graphs, slicers, and drill paths declared in PML. Channels propagate state; sections and panels compose into live dashboards — no event wiring, no JavaScript.
INTEGRATED ANALYTICS
Twenty analysis block types built into the language — forecasting, regression, clustering, anomaly detection, and more — working directly over your dimensional data without leaving PML.
SQL-INTEGRATED
Load from your warehouse with SOURCE SELECT.
Expose any ORTHO back through the PostgreSQL foreign data wrapper.
Heavy computation stays where the data lives.
AI-NATIVE SYNTAX
Block-structured, keyword-explicit PML is generated reliably by language models — and compiler feedback closes the loop when they make a mistake. Write it yourself or let an agent write it for you.
LLVM-COMPILED
JIT execution via LLVM ORC for interactive REPL and browser IDE speed. Generate a native binary from the same source for deployment or embedding.
The language
Explicit by design.
psimulang is block-structured and keyword-explicit — it reads like a specification, not a program. Every structure has a named introducer and an END terminator. Nothing is implicit.
That's deliberate: it makes models auditable, versionable, and straightforward for language models to generate and verify. The same source compiles to LLVM IR or emits SQL depending on where the data lives.
-- Dimensions describe the shape of your data
DIMENSION Region
FIELDS "North America", Europe, "Asia Pacific"
END DIMENSION
ORTHO Sales
DIMENSIONS Region, Product, Quarter
END ORTHO
-- Rules recompute automatically when their inputs change
RULE Margin (Product.*, Measure.Net)
ON Sales
Net = Revenue - Cost
END RULE
Built for teams that
own what they run on.
If you're building a data product, a planning layer, or a semantic model that your team needs to own and evolve — psimulang is worth a conversation.