Marketing ScienceADA Asia2022 – 2024Production

Multi-Touch Attribution System

Probabilistic multi-touch attribution using Markov Chain removal-effect modeling — crediting each channel based on its actual influence on conversion probability across the full customer journey.

Daily

Attribution Scoring

2-Tier

Model Architecture

6

Funnel Stages Covered

The Problem

Marketing teams were using first-touch and last-touch attribution — both systematically over-crediting one touchpoint and ignoring the influence of others. They needed statistically-grounded attribution to justify channel investment and understand which audience-campaign combinations drove conversion at each funnel stage.

What Was Built

Built a Markov Chain MTA system — customer journeys modeled as Markov states, transition matrix built from historical paths, attribution computed via removal effect (drop in conversion probability when a channel is removed). Enriched with firmographic attributes (region, industry, company size, persona) and campaign-level metadata (channel type, content format, offer type, message theme), enabling attribution of not just which channels work but why specific audience × campaign combinations are effective at each funnel stage.

Business Impact

Replaced rule-based first/last-touch attribution with daily probabilistic multi-touch attribution. Enabled stage-by-stage conversion credit by channel, audience segment, and campaign type — giving marketing teams evidence-backed answers to budget allocation questions.

Tech Stack

PythonMarkov Chainscikit-learnBigQuery

Domain Tags

Multi-Touch AttributionMarkov ChainMarketing AttributionFirmographic EnrichmentFunnel Analytics

Details

Role
Primary Owner
Status
Production
Tier
Tier 1
Period
2022 – 2024
Employment
ADA Asia