AI Marketing Automation Assignment for Interns
Objective
To assess the intern's ability to use AI agents and LLMs to automate marketing campaign
management, analyze large datasets, and optimize campaigns by performing actions such as
pausing, increasing budgets, or decreasing budgets based on performance metrics.
Scenario
You are tasked with building an AI Marketing Automation Agent that processes campaign
performance data, evaluates key metrics, and makes decisions to optimize campaigns using
predefined rules and AI-powered insights.
Assignment Instructions
Step 1: Dataset Setup
● Use a simulated or real dataset that includes marketing campaign performance data.
The dataset should have the following fields:
○ Campaign ID: Unique identifier for campaigns.
○ Impressions: Number of times an ad was shown.
○ Clicks: Number of clicks the ad received.
○ Conversions: Number of successful actions (e.g., purchases, signups).
○ Spend: Total money spent on the campaign.
○ Revenue: Total revenue generated from the campaign.
○ Status: Current status (e.g., Active, Paused).
Step 2: Define Rules for Campaign Optimization
Your agent should use the following rules to decide actions:
1. Pause Campaign:
○ If the campaign's CTR (Click-Through Rate) < 1% for the last 3 days.
○ If the Cost per Conversion > 3x the Target CPA (Cost Per Acquisition).
2. Increase Budget:
○ If ROAS (Return on Ad Spend) > 4.
○ If Conversions increase by more than 20% week-over-week.
3. Decrease Budget:
○ If ROAS < 1.5 for 5 consecutive days.
4. Generate Insights:
○ Summarize reasons for pausing/increasing/decreasing budgets.
○ Suggest optimization ideas for ad creatives or targeting based on trends.
Step 3: Build the Automation Pipeline
1. Data Ingestion:
○ Load the campaign data from a CSV, database, or API.
○ Preprocess the data (e.g., handle missing values, normalize metrics).
2. Analysis Engine:
○ Calculate key metrics like CTR, ROAS, CPA, etc.
○ Use LLMs to analyze campaign text (e.g., ad copy) for improvements.
3. AI Decision Agent:
○ Implement decision-making logic to recommend or execute actions (Pause,
Increase Budget, Decrease Budget).
4. Action Execution:
○ Simulate API calls to apply changes (e.g., update campaign status or budgets).
○ Use placeholders or integrate with a mock-up API if actual access to ad platforms
is unavailable.
5. Reporting:
○ Generate daily summary reports highlighting the actions taken, key metrics, and
recommendations for improvement.
Step 4: Enhance with AI Features
1. LLM-Powered Insights:
○ Use an LLM to identify trends, anomalies, or opportunities in campaign data.
○ Example: Generate suggestions for improving ad copy based on top-performing
campaigns.
2. Automation Efficiency:
○ Implement batch processing to handle multiple campaigns simultaneously.
○ Use caching or database integration to store historical data for trend analysis.
3. Visualization:
○ Create dashboards to visualize campaign performance (e.g., graphs for ROAS,
CTR, budget trends).
Deliverables
1. Code Repository:
○ Submit the code on GitHub with a clear [Link] file.
○ Include setup instructions and sample input data.
2. Demo Report:
○
Share a report or video (3-5 minutes) demonstrating the working pipeline and
outputs.
○ Include examples of decisions made by the AI agent and their impact on
campaigns.
3. Documentation:
○ Explain the architecture, tools, and AI models used.
○ Describe how the system can be extended for real-world scenarios.
Evaluation Criteria
1. Functionality: Does the system correctly analyze data and take appropriate actions?
2. AI Integration: How effectively are LLMs and AI techniques applied?
3. Scalability: Is the system designed to handle large datasets or multiple campaigns?
4. Code Quality: Clean, modular, and well-documented code.
5. Innovation: Unique or creative solutions, such as advanced LLM use cases or
visualization techniques.
Example Use Case for Reference
● Input:
○ A CSV file containing 10 active campaigns with varying performance data.
● Actions:
○ Pause 2 campaigns due to low CTR.
○ Increase the budget for 3 campaigns with high ROAS.
○ Decrease the budget for 1 campaign due to poor performance.
● Output:
○ A JSON or CSV file with updated campaign statuses and budgets.
○ A report summarizing the actions and reasons.
This assignment combines practical AI, data handling, and decision-making skills, preparing
interns for real-world challenges in marketing automation. Let me know if you'd like adjustments!
