11.09 · Research / Impact Engine MVP brief
The Impact Engine, specified.
The product brief that turns the Impact Engine from a story into a build: a unified scoring and attribution system connecting media performance across Linear TV, Streaming/CTV, YouTube, and Retargeting to real business outcomes such as orders, revenue, leads, and site traffic. This page is the internal record of what the MVP is, how the scoring works, and what "done" means.
Purpose
The purpose of the Impact Engine MVP is to provide a single, interpretable view of how all media channels work together each day: helping teams understand what drove results, why it happened, and how to improve performance going forward. Interpretability is a requirement, not a preference. No black boxes.
What already exists · the V1 prototype
The V1 prototype scope as of July 2026. The prototype identifies independent signals of media quality, such as audience alignment, network and platform strength, day-of-week patterns, and daypart performance, then normalizes and combines them into a single daily score: the MIS, the Media Impact Score.
- Stable daypart effects. Time of day matters, and it matters consistently.
- Channel and network differences. Performance varies measurably by where the spot runs.
- Lagged response patterns. Linear especially moves outcomes across multiple days, not just day-of.
- Faster digital responses. Digital channels answer quickly; the model treats the two speeds differently.
Scope note. What counts as "existing functionality" is a group decision, not a documentation default: the list above is the draft for that discussion. The rule going forward: this section is reviewed and restamped at every major release, so the version named here always matches what is actually running.
The MIS process, in eight steps
The daily loop the production system formalizes.
MVP requirements
Data integration
Vendor connects to Diray's AWS environment and unified Postgres DB. Diray provides schemas and documentation.
Daily processing
- Runs once daily
- Required missing data triggers skip
- Late-arriving data triggers automatic reprocessing
Channel scoring (MIS)
- Interpretability
- Component independence
- Normalization
- Stable recalibration
- Client-specific learned weights
Attribution
Lag-aware attribution using channel-specific lag profiles and MIS-driven proportional contribution.
Recommendations
Scenario blocks and recommendation cards with simple predictive estimates. Generated daily and on-demand.
Anomaly detection & benchmarking
Automatically detect unusual spikes and dips, flag anomalies clearly, apply smoothing where appropriate. Internal benchmarking normalized for spend, campaign size, and outcome volume, displayed as percentiles.
Two dashboards, two audiences
To be precise about internal versus external: the system is internal-first, but not internal-only. The working tool is the internal Admin and Analyst dashboard. The client view is a controlled, summary-only surface on the same system, exposed by permission, never by default. (Separately, this page itself is internal documentation; "internal record" in the intro refers to the document, not the product.)
Internal dashboard · Admin and Analyst
- High-end UI
- MIS trends
- Channel breakdowns
- Attribution
- Component explanations
- Benchmarking
- Anomaly indicators
- LLM insights
- PDF + CSV exports
- Admin: config, retraining, backfills
Client view · summary-only
A simplified premium UI with MIS overview, trends, basic attribution, LLM summary, and light recommendations. The client sees clarity, never the machinery.
Acceptance criteria
The MVP is done when every box below is checked in production, not in a demo.
After the MVP
The roadmap the foundation is being prepared for.
- Predictive budget optimization
- Multi-outcome modeling
- Cross-client benchmarking
- Federated learning
- Creative scoring expansion
- Real-time updates
Source: the Impact Engine MVP Product Brief (2026), archived above. The brief's note to the vendor stands as the page's closing rule: the prototypes and white paper are conceptual guides; the vendor must validate ideas, enhance rigor, and deliver a scalable, interpretable production system.