Markov chain attribution
Markov chain attribution models customer journeys as a probabilistic graph of transitions between channel states, ending in conversion or null. Each channel's credit is derived from its 'removal effect' — how much the overall conversion probability falls if that channel (and its transitions) are removed from the graph. It is a leading algorithmic alternative to Shapley-based attribution.
What this means
A Markov chain represents the journey as states (start, each channel, conversion, null) with transition probabilities estimated from observed paths. The model assumes the next state depends on the current state (the Markov property). Once the transition matrix is built, you compute the baseline conversion probability of the whole graph.
Then, for each channel, you compute the 'removal effect': delete that channel's state and re-route its transitions, recompute the conversion probability, and measure how much it dropped. Channels whose removal causes a large drop are credited more.
Strengths and caveats
Markov attribution captures sequence and interaction effects without arbitrary positional weights, and the removal-effect framing maps naturally to 'what would we lose without this channel?' It is widely implemented in open-source tooling and marketing-science workflows.
Its caveats mirror other data-driven models: it needs sufficient path volume to estimate transitions reliably, the first-order Markov assumption can oversimplify long journeys, and undertracked channels (offline, walled-garden, consent-blocked) bias the graph. As with Shapley, correlation-based credit is not proof of causation — pair it with incrementality testing for causal claims.
- Journey modeled as channel-state transitions
- Credit = removal effect on conversion probability
- Needs path volume; assumes first-order memory
How it appears in analytics and logs
Credit derived from how much conversions drop when a channel is hypothetically removed indicates a Markov model; high removal-effect channels are bottlenecks the journey depends on.
Diagnostic use case
Use Markov chain attribution when you want a data-driven credit split that captures transition patterns between channels and quantifies each channel's importance by removal effect.
What WebmasterID can help detect
WebmasterID's first-party path data — ordered source/medium touchpoints — is the transition input a Markov model consumes, letting you reason about channel removal effects on your own journeys.
Common mistakes
- Running Markov attribution on too few paths to estimate transitions.
- Forgetting the first-order assumption can flatten long journeys.
- Reading removal effect as proven causal lift.
Privacy and accuracy notes
Markov attribution works on aggregated transition counts between channel states, not on individual identity. It estimates probabilities over paths rather than profiling people.
Related pages
- Shapley value attribution
Shapley value attribution applies a concept from cooperative game theory: it treats channels as players in a coalition and assigns each one credit equal to its average marginal contribution across all possible orderings of channels. The result is a principled, order-independent way to split conversion credit. It underpins data-driven attribution in several analytics products.
- Data-driven attribution: promise and caveats
Data-driven attribution (DDA) assigns credit using a model trained on a site's own conversion paths rather than a fixed rule like last-click. Done well it credits assist touches more fairly. Its caveats are real: it needs enough conversion volume, it is a model not a measurement, and it cannot see touches that were never tracked.
- Incrementality testing: what attribution cannot tell you
Incrementality testing measures the lift a channel actually causes by withholding it from a control group and comparing outcomes. It answers the question every attribution model dodges: would this conversion have happened anyway? It is causal where attribution is merely correlational, but it requires deliberate experiment design.
- Attribution analytics
Ordered first-party touchpoints as transition input.
Sources and verification notes
- Google Analytics Help — Data-driven attribution (algorithmic family)Context for algorithmic attribution; Markov-chain removal-effect attribution is a documented marketing-science method.
- Markov chain — conceptDefinition of state-transition modeling underlying the method.
Last reviewed 2026-06-24. Facts are checked against primary/official sources where available; uncertain specifics are marked “Data not yet verified” rather than guessed.