Decision-support for paid media under uncertainty.
Upload Meta & Google exports, validate signal quality, and receive a clear decision outcome — SCALE / HOLD / REDUCE / BLOCK — with structured rationale, guardrails, and an audit snapshot.
What MDU Engine does
- Validates whether your data window is decision-ready (typically 7–30 days).
- Normalizes exports into an engine-ready daily schema.
- Evaluates outcomes under downside-risk and uncertainty constraints.
- Explains every decision outcome and explicitly blocks unsafe action.
- Logs a reproducible audit snapshot (versions, thresholds, outcome).
The problem MDU Engine addresses
Capital allocation decisions in high-variance environments are often made on unstable, incomplete, or noisy signals.
Optimisation systems tend to push action without making uncertainty, downside risk, or decision irreversibility explicit. This leads to premature scaling, difficult-to-reverse budget changes, and weak accountability when outcomes degrade.
Who MDU Engine is for
Business owners & founders
Protect capital when mistakes are costly and signals are noisy.
CFO & finance teams
Add governance, loss-protection, and auditability to media spend decisions.
Analysts & growth operators
Work with explicit confidence tiers and rationale instead of dashboard pressure.
Technical teams
Prefer deterministic, explainable logic with reproducible decision records.
Who this is not for
Automated optimisation • novice experimentation • growth hacks • execution-first systems without risk governance
From instinct to defensible decisions.
Most paid media failures come from reacting too quickly to noise, or acting too late while losses compound. MDU Engine is intentionally conservative, explainable, and repeatable.
Decision scenarios
Explicit conditions such as insufficient data windows, unstable signals, volatility spikes, or negative value drift — each mapped to a consistent outcome.
Loss-protection gates
Safety rails that block decisions when downside risk dominates or confidence thresholds are not met. Defaults to restraint.
Signal quality indicator
A single quality indicator reflecting data sufficiency and stability — not an optimisation score or performance promise.
What ships next (without destabilising the engine)
Benchmarking (planned)
Start with a simple baseline: compare current behaviour to your own trailing period and flag abnormal variance.
Decision memory UI
A decision history panel showing recent runs, inputs, versions, outcomes, and explanation trails — built for audit and review.
Stability over speed
The landing page remains public at mduengine.com. The decision engine operates independently at app.mduengine.com.
Trust, transparency, and safety by design.
Explainable outputs
No black-box advice. Every outcome includes the primary constraint, supporting factors, and confidence assessment.
Validation gates
Decisions are blocked when data is insufficient, unstable, or fails defined safety thresholds.
Audit snapshot
Each run records versions, thresholds, data window, and outcome for reproducibility and review.
Product principles
Loss-first decisions • Explainability over persuasion • Human-in-the-loop • Explicit refusal under uncertainty • Versioned and auditable outcomes