# Machine Learning Framing Record Worksheet

**Project Name:** _________________________________  
**Date Created:** _________________________________  
**Record Owner:** _________________________________  
**Version:** _____________________________________

---

## Purpose
This Machine Learning Framing Record bridges business objectives and technical implementation, ensuring your ML project delivers measurable value. Based on the [AWS Well-Architected Machine Learning Lens](https://docs.aws.amazon.com/wellarchitected/latest/machine-learning-lens/machine-learning-lens.html), it combines business understanding with technical problem framing to establish clear evaluation criteria before development begins.

---

## Section 1: Business Understanding

### 1.1 Business Problem Statement
*Describe the specific business challenge in clear, non-technical language*

**What is the problem?**
```
[Describe the current situation and pain points]




```

**Why does this problem matter?**
```
[Explain business impact and urgency]




```

**What is the cost of NOT solving this problem?**
```
[Quantify in terms of revenue, cost, time, or customer impact]
- Annual revenue impact: $__________
- Cost impact: $__________
- Time wasted: __________ hours/week
- Customer impact: __________
```

### 1.2 Stakeholders
| Stakeholder | Role | Interest/Concern | Success Criteria |
|-------------|------|------------------|------------------|
| | | | |
| | | | |
| | | | |
| | | | |

### 1.3 Business Success Vision
*Describe what success looks like in business terms*

**Success Statement:** "We will know we've succeeded when..."
```
[Complete this statement with specific, measurable business outcomes]




```

**Business Metrics That Will Improve:**
- [ ] Revenue (Target: ______)
- [ ] Cost Reduction (Target: ______)
- [ ] Time Savings (Target: ______)
- [ ] Customer Satisfaction (Target: ______)
- [ ] Process Efficiency (Target: ______)
- [ ] Other: __________ (Target: ______)

---

## Section 2: Technical Problem Framing

### 2.1 ML Problem Type Classification
*Select the type of ML problem that best matches your business need*

- [ ] **Classification** - Categorize inputs into discrete classes
  - [ ] Binary (2 classes)
  - [ ] Multi-class (>2 classes)
  - [ ] Multi-label (multiple classes per input)

- [ ] **Regression** - Predict continuous numerical values

- [ ] **Clustering** - Group similar items without predefined categories

- [ ] **Recommendation** - Suggest relevant items to users

- [ ] **Anomaly Detection** - Identify unusual patterns or outliers

- [ ] **Time Series Forecasting** - Predict future values based on historical data

- [ ] **Natural Language Processing** - Process and understand text
  - [ ] Sentiment Analysis
  - [ ] Entity Recognition
  - [ ] Text Generation
  - [ ] Question Answering

- [ ] **Computer Vision** - Process and understand images/video
  - [ ] Object Detection
  - [ ] Image Classification
  - [ ] Segmentation

### 2.2 Input-Output Specification

**Model Inputs (What data will the model observe?)**
| Data Type | Source | Format | Availability | Quality |
|-----------|---------|---------|--------------|---------|
| | | | Ready/Needs Work | High/Medium/Low |
| | | | Ready/Needs Work | High/Medium/Low |
| | | | Ready/Needs Work | High/Medium/Low |
| | | | Ready/Needs Work | High/Medium/Low |

**Model Outputs (What should the model predict?)**
```
[Describe the specific predictions/outputs the model should produce]




```

**Constraints and Requirements**
- **Latency Requirement:** Response must be within ______ milliseconds
- **Throughput Requirement:** Must handle ______ predictions per second
- **Accuracy Requirement:** Minimum ______% accuracy
- **Infrastructure Constraints:** _________________________________
- **Regulatory Constraints:** _________________________________

### 2.3 Initial Technical Approach
*Describe your hypothesis for how to solve this problem*

```
[Outline the proposed technical approach, algorithms to consider, etc.]




```

---

## Section 3: Evaluation Criteria

### 3.1 Technical Metrics
*Select and define thresholds for technical performance metrics*

**For Classification Problems:**
- [ ] **Accuracy** 
  - Minimum: _____%
  - Target: _____%
  - Stretch: _____%

- [ ] **Precision** (Of predicted positives, how many are correct?)
  - Minimum: _____%
  - Target: _____%
  - Stretch: _____%

- [ ] **Recall** (Of actual positives, how many did we catch?)
  - Minimum: _____%
  - Target: _____%
  - Stretch: _____%

- [ ] **F1 Score** (Harmonic mean of precision and recall)
  - Minimum: _____
  - Target: _____
  - Stretch: _____

- [ ] **AUC-ROC** (Area under ROC curve)
  - Minimum: _____
  - Target: _____
  - Stretch: _____

**For Regression Problems:**
- [ ] **RMSE** (Root Mean Square Error)
  - Maximum acceptable: _____
  - Target: _____
  - Stretch goal: _____

- [ ] **MAE** (Mean Absolute Error)
  - Maximum acceptable: _____
  - Target: _____
  - Stretch goal: _____

- [ ] **R²** (Coefficient of determination)
  - Minimum: _____
  - Target: _____
  - Stretch: _____

**Other Metrics:**
- [ ] ________________ (Target: _____)
- [ ] ________________ (Target: _____)

### 3.2 Business Metrics Mapping
*Connect technical metrics to business outcomes*

| Technical Metric | Threshold | Business Impact | Value |
|-----------------|-----------|-----------------|--------|
| | | | $ |
| | | | $ |
| | | | $ |

### 3.3 Evaluation Methodology

**Offline Evaluation Plan:**
- [ ] Train/Validation/Test split defined (____/____/____)
- [ ] Cross-validation strategy: _________________
- [ ] Baseline comparison method: _________________
- [ ] Error analysis approach: _________________

**Online Evaluation Plan:**
- [ ] A/B testing framework defined
- [ ] Canary deployment percentage: _____%
- [ ] Success criteria for full rollout: _________________
- [ ] Rollback criteria: _________________

**Business Impact Measurement:**
- [ ] Pre-implementation baseline captured
- [ ] Post-implementation measurement plan
- [ ] ROI calculation methodology defined
- [ ] Stakeholder feedback collection process

---

## Section 4: Risk Assessment and Mitigation

### 4.1 Data Risks
| Risk | Likelihood | Impact | Mitigation Strategy |
|------|------------|---------|-------------------|
| Insufficient data volume | High/Medium/Low | High/Medium/Low | |
| Poor data quality | High/Medium/Low | High/Medium/Low | |
| Data bias | High/Medium/Low | High/Medium/Low | |
| Privacy concerns | High/Medium/Low | High/Medium/Low | |
| Data drift over time | High/Medium/Low | High/Medium/Low | |

### 4.2 Technical Risks
| Risk | Likelihood | Impact | Mitigation Strategy |
|------|------------|---------|-------------------|
| Model doesn't meet performance targets | High/Medium/Low | High/Medium/Low | |
| Scalability issues | High/Medium/Low | High/Medium/Low | |
| Integration complexity | High/Medium/Low | High/Medium/Low | |
| Maintenance burden | High/Medium/Low | High/Medium/Low | |

### 4.3 Business Risks
| Risk | Likelihood | Impact | Mitigation Strategy |
|------|------------|---------|-------------------|
| User adoption challenges | High/Medium/Low | High/Medium/Low | |
| ROI not realized | High/Medium/Low | High/Medium/Low | |
| Stakeholder misalignment | High/Medium/Low | High/Medium/Low | |
| Regulatory compliance | High/Medium/Low | High/Medium/Low | |

---

## Section 5: Go/No-Go Decision Criteria

### 5.1 Proceed to Development If:
- [ ] Business problem is clearly defined and quantified
- [ ] ML problem type is appropriate for the business need
- [ ] Success metrics are specific and measurable
- [ ] Required data is available or obtainable
- [ ] Technical approach is feasible
- [ ] ROI justifies the investment
- [ ] Risks are acceptable and mitigable
- [ ] Stakeholders are aligned on success criteria

### 5.2 Reasons to Pause or Pivot:
- [ ] Business problem can be solved without ML
- [ ] Data quality/quantity insufficient
- [ ] Technical complexity exceeds capabilities
- [ ] ROI doesn't justify investment
- [ ] Unacceptable risks identified
- [ ] Stakeholder alignment cannot be achieved

---

## Section 6: Approval and Sign-off

| Role | Name | Signature | Date |
|------|------|-----------|------|
| Business Sponsor | | | |
| Technical Lead | | | |
| Data Owner | | | |
| End User Representative | | | |

---

## Next Steps

1. [ ] Schedule technical feasibility deep-dive
2. [ ] Conduct data quality assessment
3. [ ] Create project timeline and budget
4. [ ] Assign project team
5. [ ] Begin development phase
6. [ ] Other: _______________________

---

## Appendix: Quick Reference

### When to Use Each ML Problem Type

**Classification**
- Spam detection
- Fraud identification
- Customer churn prediction
- Disease diagnosis
- Quality control (pass/fail)

**Regression**
- Sales forecasting
- Price prediction
- Demand estimation
- Risk scoring
- Performance prediction

**Clustering**
- Customer segmentation
- Document organization
- Anomaly detection
- Pattern discovery

**Recommendation**
- Product suggestions
- Content recommendations
- Next best action
- Personalization

---

*This worksheet is based on the [AWS Well-Architected Machine Learning Lens](https://docs.aws.amazon.com/wellarchitected/latest/machine-learning-lens/machine-learning-lens.html) framework for business understanding and ML problem framing. For more resources and guidance, visit [IncitesAI.com](https://incitesai.com)*