05
Marketing Intelligence

PRISM™

Unified Marketing Intelligence & Decision System

PRISM turns marketing data into causal truth, governed intelligence, and business action. It unifies attribution, MMM reconciliation, causal intelligence, decision simulation, and governance in one auditable platform.

MMM + MTACausal AttributionDecision OSGoverned
View Platform → Request Demo
3
Integrated products: PRISM AIP, PRISM MTA, PRISM vNext.
7
Architecture layers from data ingestion to governed decision narrative.
MMM + MTA
Macro and person-level intelligence in one governed environment.
Causal
Attribution grounded in incrementality logic — not last-click assumptions.
L1
Data Ingestion & Semantic Mapping
Media dataChannel signalsBusiness KPIsAnalytical contract
L2
Identity & Journey Construction
Deterministic matchProbabilistic stitchSession assemblyTouchpoint ordering
L3
Attribution & Causal Intelligence
Hybrid attributionIncrementality logicMMM reconciliationExperiment signals
L4
Validation & Confidence Index
Model comparisonCI scoringHoldout testingStability checks
L5
Decision & Governance
Budget simulatorScenario plansRecommendationsAudit trail
L6
Executive Narrative
Channel rolesInsight summaryEvidence exportBoard-ready output
L7
Optimisation Loop
Efficient frontierAllocation modelSaturation curvesContinuous update
Platform Overview

A unified system — not a collection of disconnected tools.

PRISM eliminates the fragmentation between attribution, MMM, and budget decisions. Three integrated products. One governed intelligence environment.

01
Foundation

PRISM AIP

The Attribution Intelligence Platform defines the full operating pipeline: data ingestion, identity resolution, journey construction, modelling, validation, decisioning, and governance. Seven layers, one auditable environment.

7 LayersPipeline ArchitectureEnterprise Foundation
02
Truth Engine

PRISM MTA

The person-level intelligence layer. Reconstructs customer journeys, resolves identity, measures channel contribution, and estimates true incremental impact using hybrid attribution — Markov chains, Shapley values, and MMM reconciliation.

Hybrid AttributionIdentity StitchingConfidence Index
03
Decision OS

PRISM vNext

The evolved decision layer. Model tournaments, Bayesian reasoning, budget simulators, allocation optimisers, governed recommendations, and evidence-grade export packs — all within a single operating environment.

Decision OSModel TournamentGoverned Action
Attribution Intelligence

How PRISM MTA finds causal truth.

Markov Chain Attribution

Journey-level removal effect

Markov attribution models each channel as a state in the conversion journey. It estimates each channel's contribution by measuring the probability of conversion with and without that channel — the removal effect.

  • State transition matrices from journey data
  • Removal effect for each touchpoint
  • Handles long, non-linear paths
  • No arbitrary rule assumptions
Shapley Value Attribution

Cooperative game theory allocation

Shapley values treat channels as players in a cooperative game. Each channel receives its average marginal contribution across all possible orderings — a provably fair, game-theoretically grounded attribution rule.

  • Marginal contribution across all permutations
  • Efficiency: credits sum to total conversions
  • Symmetry: equal channels receive equal credit
  • Weighted by MMM-calibrated priors
Hybrid Attribution & MMM Reconciliation

The ensemble that resolves macro-micro conflict

PRISM blends Markov and Shapley outputs using MMM-calibrated weights — so person-level MTA is constrained to be consistent with aggregate media mix findings. This reconciliation eliminates the classic conflict between attribution models and MMM.

  • Weighted ensemble: α·Markov + (1-α)·Shapley
  • α calibrated from MMM offline spend elasticities
  • Confidence Index: model agreement × data coverage × stability
  • Geo holdout experiments for incrementality validation
Confidence Index

Knowing how much to trust the attribution

Attribution without uncertainty quantification is false precision. PRISM's Confidence Index (CI) scores each attribution output on three dimensions — making it safe to act on the results.

  • Model agreement: Markov vs. Shapley consistency
  • Data coverage: journey completeness ratio
  • Temporal stability: variance across periods
  • CI > 0.75 = high confidence; act directly
  • CI 0.50–0.75 = moderate; validate with experiment
  • CI < 0.50 = low; audit data pipeline first
PRISM vNext — Decision OS

From attribution to governed action.

PRISM vNext is where attribution intelligence becomes actionable. Model tournaments, Bayesian simulation, and allocation optimisation in one governed environment.

01

Model Tournament

Multiple attribution models compete on held-out data. The winner's weights feed the Bayesian scenario engine — no single model lock-in.

02

Budget Simulator

Simulate any spend allocation and see projected outcomes — conversion, revenue, ROAS — with saturation-aware diminishing returns modelling.

03

Allocation Optimiser

Find the optimal budget split subject to channel minimums, maximums, and total budget constraints. The efficient frontier surfaces all Pareto-optimal solutions.

04

Governed Recommendations

Recommendations are issued with confidence thresholds, evidence references, and audit trails. Every output is auditable from data to decision.

05

Bayesian Scenario Engine

Ask "what if?" questions under uncertainty. The Bayesian engine propagates uncertainty through the causal graph, returning probability distributions over outcomes.

06

Evidence Export Pack

Board-ready PDF exports containing attribution analysis, model comparison, scenario results, and recommended allocation — fully audit-traceable.

PRISM — unified marketing intelligence
PRISM Philosophy
Attribution is a causal problem. Treat it causally.

Last-click winners are not the same as long-term growth drivers. PRISM is built on the conviction that marketing intelligence requires causal reasoning — not correlational shortcuts — and that every attribution output must be auditable, validated, and uncertainty-aware.

Incrementality Over Attribution

Attribution tells you who got credit. Incrementality tells you who created the outcome. PRISM ties both together through geo holdout experiments and MMM reconciliation.

Uncertainty by Design

The Confidence Index makes model reliability visible. You know not just what the attribution says, but how much to trust it — before you act.

One Source of Truth

PRISM unifies MTA, MMM, and experiment signals in one governed environment. Marketing, finance, and leadership work from the same validated intelligence.

Decisions, Not Reports

Every PRISM output connects to an action: rebalance budget, cap a channel, shift to brand building. Attribution that does not drive decisions is not intelligence.

Ready to unify your
marketing intelligence?

Start a conversation — we'll walk you through PRISM's capabilities and show you where it fits your environment.

Request a Demo →