Build multi-touch attribution models
attribution-modelingskillsetup L2★4
ashutoshsrivastava17/skill-library ↗Causal-lift measurements
marketing-analytics+12pp vs no-skill baselinewith-skill 76% · baseline 64%
Measured by running the task with and without this artifact, K=5, graded by deterministic checks — no LLM judging.
What it does
Build and validate attribution models to understand how channels drive conversions
Best for
When you need to move beyond last-touch attribution to understand which channels truly drive conversions.
Inputs
- · Conversion event definition and lookback window
- · Touchpoint log (timestamp, channel, campaign, content)
- · Conversion log (timestamp, value, type)
Outputs
- · Multi-model comparison (first-touch, last-touch, linear, time-decay, position-based, data-driven)
- · Channel-by-model credit allocation table
- · Budget optimization recommendations
- · Incrementality test plan
Requires
- · Analytics platform (GA4, Mixpanel, custom warehouse)
- · SQL or Python for modeling
Preconditions
- · 1,000+ conversions for statistical confidence
- · User identity is consistent across channels
- · Touchpoint and conversion logs are deduplicated
- · Attribution window is appropriate for sales cycle
Failure modes
- · Confuses correlation with causation
- · Uses last-touch model and ignores awareness contribution
- · Attribution window is too short for long sales cycles
- · Low conversion volume (<500) limits model choice
- · Cross-device tracking is missing
Trust signals
- · 6-model comparison framework
- · Model selection matrix by goal (awareness, consideration, conversion, retention)
- · Validation with incrementality tests (geo holdout, ghost ads, matched market)
- · Budget allocation recommendation table
- · Edge cases for B2B, subscriptions, offline conversions