: It aggregates individual touchpoints into sequential "paths." This involves grouping all interactions a user had leading up to a specific conversion event [4].
In the context of Multi-Touch Attribution (MTA) models, the feature or step within a script like strongmta.sql is designed to transform raw, event-level marketing data into a structured format suitable for attribution modeling. Core Functions of the "Prepare" Feature
: It standardizes timestamps, user identifiers (UIDs), and channel names across different platforms (e.g., Google Ads, Facebook, Organic Search) to ensure a unified view of the customer journey [1, 3]. strongmta.sql
: A concatenated string or array of channels (e.g., Social > Search > Email ).
: In many MTA workflows, the "prepare" step separates paths that ended in a conversion from those that didn't, allowing the model to analyze "null" paths for more accurate probability calculations [4]. Typical Structure of the Prepared Data : A concatenated string or array of channels (e
Without this preparation step, MTA models cannot handle the high cardinality of raw clickstream data. It ensures that the input is and linearly ordered , which is a prerequisite for calculating the incremental lift of specific marketing channels [3, 5].
: The script applies logic to filter out interactions that occurred outside a defined lookback window (e.g., 30 days) and identifies which touchpoints belong to a single conversion cycle [2, 5]. It ensures that the input is and linearly
Once the "prepare" feature executes, the output table usually contains: : A unique identifier for the customer.