Decision-support (not auto-execution) • Explainable • Risk-aware

Capital-intelligent budget decisions for paid media.

Upload Meta & Google exports, validate signal quality, and generate a clear “HOLD / SCALE / REDUCE” recommendation with reasons, guardrails, and an audit snapshot.

Built by Satish Saka • Public Decision Engine • MIT License • Asia-South1

What MDU Engine does

  • Validates whether your data window is decision-ready (7–30 days recommended).
  • Normalizes exports into an engine-ready daily schema.
  • Simulates outcomes with downside-risk protection.
  • Explains every recommendation and blocks unsafe actions.
  • Logs an audit snapshot (versions + settings + decision summary).
MDU Engine provides decision support based on uploaded data. It does not execute changes automatically and should not be considered financial or investment advice.

From “gut feel” to defensible decisions.

Most ad accounts fail from two extremes: reacting too fast to noise, or waiting too long while losses compound. MDU Engine is designed to be conservative, explainable, and operationally usable.

Decision Scenarios

Defined situations like low data window, tracking instability, volatile spend spikes, or negative net-value drift — each produces a consistent action with rationale.

Loss Protection Rules

Hard safety rails that block risky recommendations when downside risk exceeds your selected threshold. Defaults to cautious behavior.

Health Score

A single quality indicator showing whether your signal is stable enough to act on today. Improves with clean daily data.

What you’ll ship next (without disturbing the current app)

Benchmarking (stub)

Start with a simple baseline: compare current performance to your own trailing period and flag abnormal variance. Later, expand to anonymized peer bands.

Planned

Decision Memory UI

A clean “Decision History” panel that shows last 10 runs, inputs, versions, outcome, and the explanation trail — built for audit and repeatability.

Shipping

Keep the engine stable

Landing page stays public on mduengine.com. The app remains on app.mduengine.com exactly as it is today.

Trust, transparency, and safety by design.

Explainable outputs

No black-box “increase budget” advice. Every decision includes what happened, what could go wrong, and what to do next.

Validation gates

Engine blocks decisions when the data window is too short or malformed. Safety > speed.

Audit snapshot

Each run records settings, confidence, risk thresholds, and engine/ruleset versions.