Problem Statement
The workflow aims to analyze marketing attribution logic for bias, quantify true incremental impact of channels, and produce a structured plan for testing and budget reallocation. It helps identify distortions such as last-click bias, self-selection, missing touchpoints, and offline spillover. The output includes a risk audit, an incrementality test plan, decision thresholds, and suggested budget shifts.
Steps
Ingest Spend and Conversion Data
INGESTCollect raw channel spend, conversions, touchpoint logs, and any geo or time splits provided. Standardize formats and validate that spend and conversion units are coherent before analysis.
Retrieve Current Attribution Logic
EXECUTELoad or reference the client's current attribution model, including rules for credit assignment, lookback windows, and included or excluded channels. Cross-check with available data to identify mismatches.
Generate Attribution Bias Audit
EXECUTEUsing provided data and assumptions, produce an audit identifying sources of bias such as last-click distortions, self-selection, missing touchpoints, or offline effects. Summarize the reliability of current attribution outputs.
Generate Incrementality Test Plan
EXECUTECreate 2–4 recommended incrementality experiments such as geo-holdouts, time-based tests, PSA tests, or audience splits. Include sample size heuristics, risks, contamination concerns, and power expectations.
Validate and Apply Decision Framework
EXECUTETranslate test results into decision thresholds for when to increase, reduce, or maintain spend per channel. Produce a final recommended budget reallocation plan based on expected incremental lift.
Prompt Configurations
Attribution Bias Audit Prompt
You are a senior marketing attribution analyst. Analyze the provided spend logs, conversions, touchpoint data, and attribution rules. Identify any likely biases including self-selection, last-click distortion, sparse data, missing touchpoints, offline spillover, or channel over-crediting. Produce a concise but thorough attribution risk report. Use a professional but clear tone. Do not invent data; only reason from the provided information.
Incrementality Test Plan Prompt
You are an incrementality and experimentation strategist. Based on marketing performance data and attribution risks, propose 2–4 incrementality tests such as geo-holdout, time-based suppression, PSA tests, or audience split tests. For each test, include sampling heuristics, required duration, contamination risks, and expected interpretability. Keep the tone practical and actionable. Do not fabricate data.
Decision Framework and Budget Shift Prompt
Act as a marketing science lead. Using attribution risks and proposed incrementality tests, produce a decision framework defining thresholds for budget increases, decreases, or freezes based on measured incremental lift. Then propose a budget reallocation recommendation grounded in the logic. Use a structured, business-friendly tone. Do not invent any unavailable data.
Example Inputs/Outputs
Spend: Meta $120k, Google $80k, TikTok $30k. Conversions: 4,200 total. Last-click attribution model provided. Touchpoints: Sparse for TikTok, missing for offline channels. Geo data: US East vs West available.
Attribution Risk Report: Last-click bias overweights Google by ignoring upper-funnel touchpoints. TikTok under-credited due to missing event logs. Offline spillover likely in US East due to regional TV spend. Test Plan: Recommend US West geo-holdout for Meta, time-based suppression for Google weekends, and PSA test for TikTok awareness campaigns. Decision Thresholds: Increase spend if incremental lift exceeds 20% relative to baseline; reduce if below 5%. Budget Suggestion: Shift 10–15% from Google to Meta upper funnel pending test confirmation.