attribution_agent
Performs data-driven marketing attribution to optimize channel budget allocation.
Install
mkdir -p .claude/skills/attribution-agent && curl -L -o skill.zip "https://agentskills.codes/api/skills/download/17006" && unzip -o skill.zip -d .claude/skills/attribution-agent && rm skill.zipInstalls to .claude/skills/attribution-agent
Activation
This is the description your AI agent reads to decide when to run this skill — the better it matches your request, the more reliably it fires.
Analyzes marketing channel attribution using 8 models (Markov, Shapley, LSTM, first/last-click, linear, time-decay, position-based) and recommends budget allocation.Key capabilities
- →Analyze marketing channel attribution using 8 models
- →Recommend budget allocation based on model agreement
- →Explain model agreement scores
- →Present budget recommendations as actionable lists
- →Explain disagreements between attribution models
How it works
The skill executes a Python script that runs 8 attribution models on marketing data, then parses the JSON output to present a briefing with budget recommendations and explanations of model behavior.
Inputs & outputs
When to use attribution_agent
- →Optimizing advertising spend
- →Evaluating marketing channel performance
- →Calculating ROI of acquisition campaigns
About this skill
Attribution Agent
You are OpenClaw's Marketing Attribution capability. When the user asks about channel performance, budget allocation, or marketing attribution, run the attribution pipeline and present the results.
How to Execute
Run the following command, passing the user's question as the --query argument:
cd C:\Users\ahmad\Downloads\Hack2Skill\openclaw-marketing-agent; venv\Scripts\python scripts/attribution.py --query "USER_QUESTION_HERE" --output json
Replace USER_QUESTION_HERE with the actual user question (keep the quotes).
How to Present Results
Parse the JSON output and present a clear, business-friendly attribution briefing:
-
Data Context: Briefly mention the dataset size (journeys, conversions, revenue) so the user knows the analysis is grounded in real data.
-
Top Channel: Lead with the winning channel, its confidence level, and how many models agree.
-
Model Agreement Score: Explain what the agreement score means in plain English:
- 90-100 (Strong): "All models consistently point to the same channels. High confidence in these recommendations."
- 70-89 (Moderate): "Most models agree, with some variation. Recommendations are reliable but have nuance."
- 50-69 (Mixed): "Models give different answers depending on methodology. Consider testing before making big changes."
- Below 50 (Low): "Significant disagreement between models. The data may not clearly favor any single channel."
-
Budget Recommendations: Present as an actionable list:
- Which channels to increase spend on and why
- Which channels to maintain
- Which channels to review or reduce
-
Model Disagreements: When models disagree, explain WHY in plain English:
- First-click ranks a channel higher: "This channel is strong at starting customer journeys (awareness)"
- Last-click ranks a channel higher: "This channel is strong at closing conversions (bottom-funnel)"
- LSTM differs from statistical models: "The deep learning model sees non-linear patterns the simpler models miss"
- Markov differs from rule-based: "The Markov model accounts for channel interactions and removal effects"
-
Query-Specific Focus: Tailor the response to the user's question:
- Budget questions: Emphasize recommendations and ROI
- Comparison questions: Deep-dive into the specific channels mentioned
- General questions: Provide the full picture
Explaining the 8 Models
When the user asks about models or methodology, explain briefly:
| Model | What it Does | Best For |
|---|---|---|
| First-Click | Credits the channel that started the journey | Understanding awareness |
| Last-Click | Credits the channel before conversion | Understanding closing power |
| Linear | Equal credit to all touchpoints | Fair baseline comparison |
| Time-Decay | More credit to recent touchpoints | Recency-weighted analysis |
| Position-Based | 40% first, 40% last, 20% middle | Balanced view |
| Markov Chain | Measures removal effect of each channel | Understanding true channel impact |
| Shapley Value | Game-theory fair allocation | Cooperative contribution |
| LSTM Deep Learning | Neural network gradient attribution | Non-linear pattern detection |
Handling Errors
- If the script fails, inform the user and suggest checking Python dependencies.
- If
data_sourceis "sample", note that sample data was used for demonstration. - If
data_sourceis "p6_precomputed", the data comes from a previous BigQuery analysis run. - Shapley values may be zero if they failed to converge during the original computation.
Example Interaction
User: "Where should we spend our marketing budget?"
Agent: Runs the attribution script with the user's question, parses the JSON, and responds with a business-friendly report showing which channels drive the most conversions, how confident the models are, and specific budget reallocation recommendations.
When not to use it
- →If the script fails due to Python dependencies
- →If the user needs to perform tasks unrelated to marketing attribution
- →If the user needs to perform data collection
Prerequisites
Limitations
- →Shapley values may be zero if they failed to converge
- →Relies on pre-computed or sample data
How it compares
This skill automates complex multi-model attribution analysis and translates the results into actionable business recommendations, unlike manual data analysis or single-model approaches.
Compared to similar skills
attribution_agent side by side with the closest alternatives in the catalog.
| Skill | Installs | Updated | Safety | Difficulty |
|---|---|---|---|---|
| attribution_agent (this skill) | 0 | 4mo | No flags | Intermediate |
| quant-analyst | 103 | 3mo | No flags | Advanced |
| umap-learn | 6 | 2mo | Review | Intermediate |
| embedding-strategies | 8 | 3mo | No flags | Intermediate |
Try saying
Example prompts that trigger this skill in your AI assistant.
You might also like
quant-analyst
zenobi-us
Expert quantitative analyst specializing in financial modeling, algorithmic trading, and risk analytics. Masters statistical methods, derivatives pricing, and high-frequency trading with focus on mathematical rigor, performance optimization, and profitable strategy development.
umap-learn
K-Dense-AI
UMAP dimensionality reduction. Fast nonlinear manifold learning for 2D/3D visualization, clustering preprocessing (HDBSCAN), supervised/parametric UMAP, for high-dimensional data.
embedding-strategies
wshobson
Select and optimize embedding models for semantic search and RAG applications. Use when choosing embedding models, implementing chunking strategies, or optimizing embedding quality for specific domains.
building-automl-pipelines
jeremylongshore
Build automated machine learning pipelines, including feature engineering, model selection, and performance evaluation.
model-compare
rawwerks
Compare 3D CAD models using boolean operations (IoU, Dice, precision/recall). Use when evaluating generated models against gold references, diffing CAD revisions, or computing similarity metrics for ML training. Triggers on: model diff, compare models, IoU, intersection over union, model similarity, CAD comparison, STEP diff, 3D evaluation, gold reference, generated model, precision recall 3D.
matchms
davila7
Mass spectrometry analysis. Process mzML/MGF/MSP, spectral similarity (cosine, modified cosine), metadata harmonization, compound ID, for metabolomics and MS data processing.