Baseline path parameters
A path scenario describes one aggregated ordered advertising journey. It preserves order, but the finalized schema has no event timestamps and cannot distinguish every impression from every click.
identifier
Meaning. Internal English name for a reusable path scenario.
Example. display_brand_product describes demand-side-platform display followed by Sponsored Brands and then Sponsored Products.
Why it makes sense. A name helps reviewers discuss and modify one journey without copying the full normalized path.
Confidence. High.
touchpoint_identifiers
Meaning. Ordered list of touchpoint catalog identifiers forming the path.
Example. Display, then brand advertising, then product advertising represents a possible movement from awareness to comparison to purchase-oriented search.
Why it makes sense. Multi-Touch Attribution depends on the sequence of contacts. Reversing the list creates a different observed path.
Relationship. Each identifier must exist in identifier, and each resulting key must exist in the performance table.
Confidence. High for matching; medium for the chosen journey stories.
base_users
Meaning. Neutral daily number of users assigned to this aggregated path.
Example. 620 means the display_brand_product path starts with 620 users before marketplace scale, seasonality, and path-audience noise.
Why it makes sense. Some journeys should be common and others rare. One explicit count makes that distribution visible.
Relationship. Scaled by traffic_multiplier, the daily seasonal multiplier, and path_audience_noise_standard_deviation.
Confidence. Low; scenario assumption.
adjacent_synergy_log_odds
Meaning. Extra conversion log-odds effect for each neighboring pair that is active together.
Example. 0.20 multiplies the odds by:
for one active adjacent pair.
Why it makes sense. An awareness advertisement followed by a search advertisement may reinforce memory and intent. The combined response can differ from treating each contact independently.
Relationship. Added to conversion_log_odds_effect inside the response function. It also changes exact Shapley marginal contributions because a touchpoint may complete a reinforcing pair.
Limitation. The exact values are not estimated from randomized experiments.
Confidence. Medium that cross-channel interactions can exist; low for configured values.