Title: Best Practices
Locale: en
URL: https://sensorswave.com/en/docs/analytics/best-practices/
Description: Master best practices and implementation recommendations for data analysis
This article summarizes best practices for using Insights, helping you conduct data analysis more effectively, avoid common pitfalls, and improve analysis quality.

## Analysis Framework

### Define Clear Analysis Goals

Before starting your analysis, clarify the question you want to answer:

**Good analysis goals**:
- "What is the Day-1 retention rate for new users over the past 7 days?"
- "What is the conversion rate from browsing products to completing payment?"
- "Which channel brings the most active users?"

**Vague analysis goals**:
- "Let's see how the data looks"
- "Analyze user behavior"
- "Create a report"

### Choose the Right Analysis Model

Select the most suitable analysis model based on your goal:

| Analysis Goal | Recommended Model | Description |
|---------|---------|------|
| Understand the count and trends of a specific event | Event Analysis | Supports multi-dimensional breakdown and comparison |
| Analyze users' continued usage | Retention Analysis | Measures product stickiness and long-term value |
| Analyze conversion efficiency of multi-step processes | Funnel Analysis | Identifies drop-off points and optimization opportunities |
| Perform complex custom analysis | SQL Query | Flexibly handles any complex requirement |

See [Choosing the Right Analysis Model](choosing-analysis-model.mdx) for details.

### Build Your Analysis Approach

**From macro to micro**:

1. **Start with overall trends**: Understand the general direction of metrics
2. **Break down by dimensions**: Use Group By to discover differences
3. **Drill down to users**: View specific user behaviors to understand the root cause

**From known to unknown**:

1. **Validate known hypotheses**: Use data to verify your assumptions
2. **Discover new insights**: Find unexpected patterns in the data
3. **Form new hypotheses**: Propose new questions based on discoveries

## Common Analysis Scenarios

### Scenario 1: Daily Data Monitoring

**Goal**: Track daily changes in key metrics and promptly identify anomalies.

**Recommended approach**:

1. **Create a core metrics Dashboard**:
   - DAU/WAU/MAU trends
   - Key feature usage rates
   - Core conversion rates

2. **Set up data comparisons**:
   - Compare with the same period last week
   - Compare with the same period last month

3. **Establish anomaly detection standards**:
   - Define normal fluctuation ranges (e.g., ±10%)
   - Investigate deeper when values exceed the range

**Example**:

```
Daily Monitoring Dashboard
├── Active Users (compared to same period last week)
├── New User Registrations
├── Core Feature Usage Rate
├── Purchase Conversion Rate
└── Revenue Amount
```

### Scenario 2: Feature Effectiveness Evaluation

**Goal**: Evaluate the usage and effectiveness of a newly launched feature.

**Recommended approach**:

1. **Define evaluation metrics**:
   - Feature penetration rate (proportion of users who used the feature)
   - Feature usage frequency (average usage per user)
   - Feature impact on core metrics

2. **Set comparison baselines**:
   - Before launch vs. after launch
   - Users who used the feature vs. those who didn't

3. **Monitor long-term effects**:
   - Don't just look at initial launch data
   - Continuously observe for 2-4 weeks

**Analysis steps**:

1. Use Event Analysis to view feature usage volume and trends
2. Use Retention Analysis to compare whether users who used the feature have higher retention
3. Use Funnel Analysis to check whether the feature improved conversion rates

### Scenario 3: Conversion Rate Optimization

**Goal**: Find bottlenecks in the conversion process and improve overall conversion rates.

**Recommended approach**:

1. **Build a complete funnel**:
   - Cover the full path from entry to final conversion
   - Break down into sufficiently granular steps

2. **Locate drop-off points**:
   - Find the step with the lowest conversion rate
   - This is typically the optimization focus

3. **Compare across dimensions**:
   - Break down by channel, device, user type, etc.
   - Discover conversion differences across groups

4. **Drill down to analyze root causes**:
   - View the Activity of lost users
   - Discover specific reasons for drop-off

**Example**:

```
Registration Conversion Funnel
├── Step 1: Visit Registration Page (baseline)
├── Step 2: Enter Phone Number (conversion rate 80%)
├── Step 3: Get Verification Code (conversion rate 90%) → Drop-off reason: Verification code send failure
├── Step 4: Enter Verification Code (conversion rate 70%) → Drop-off reason: Verification code expired
└── Step 5: Complete Registration (conversion rate 95%)
```

### Scenario 4: User Growth Analysis

**Goal**: Understand user growth and optimize acquisition strategies.

**Recommended approach**:

1. **Analyze new user sources**:
   - Use Event Analysis, grouped by channel, to view new user counts
   - Compare acquisition efficiency across channels

2. **Evaluate user quality**:
   - Use Retention Analysis to compare retention rates across channels
   - Higher retention indicates higher quality users

3. **Calculate acquisition costs**:
   - Combine with advertising data to calculate CAC per channel
   - ROI = LTV / CAC

### Scenario 5: User Value Analysis

**Goal**: Identify high-value user characteristics to guide refined operations.

**Recommended approach**:

1. **Define user value metrics**:
   - Payment amount
   - Usage frequency
   - Lifecycle length

2. **Analyze high-value user characteristics**:
   - Use SQL Query to calculate user LTV
   - Analyze common characteristics of high-LTV users

3. **Build Cohorts**:
   - Create Cohorts based on analysis results
   - Develop strategies for different value groups

## Analysis Tips

### Time Range Selection

**Choose an appropriate time range**:

| Analysis Type | Recommended Time Range | Description |
|---------|-------------|------|
| Daily monitoring | Last 7 Days | Observe recent trends |
| Periodic comparison | Last 30 Days | Include complete periods |
| Long-term trends | Last 90 Days | Discover long-term patterns |
| Campaign analysis | 7 days before and after the campaign | Evaluate campaign effectiveness |

**Time granularity selection**:

- **Hourly**: Analyze user behavior distribution within a day
- **Daily**: Daily monitoring and short-term trends
- **Weekly**: Eliminate daily fluctuations, view weekly trends
- **Monthly**: Long-term trends and seasonality analysis

### Group By Dimension Selection

**Common Group By dimensions**:

| Dimension Type | Common Dimensions | Analysis Value |
|---------|---------|---------|
| User Property | Gender, Age, Region | Understand user composition |
| Device Information | Operating System, Device Type | Optimize in-app experience |
| Channel Source | Registration Channel, Referral Page | Evaluate channel effectiveness |
| Business Property | Membership Level, User Type | Refined analysis |

**Group By recommendations**:

- Use no more than 2 Group By dimensions per analysis
- Start with a single dimension to find differences, then combine to validate
- When there are too many dimension values, use Top N or merge the long tail

### Segment By Tips

**Use filters to focus your analysis target**:

```
Analyze high-value user behavior
├── Filter condition: Cumulative payment amount > 1000
├── Or use Cohort: "High-Value Users" Cohort
└── Then perform behavior analysis
```

**Exclude interfering data**:

```
Exclude test users
├── Filter condition: User Type != "Test User"
├── Or exclude specific User IDs
└── Ensure data authenticity
```

### Data Interpretation

**Focus on relative changes rather than absolute values**:

- Don't just look at "100K active users" — look at "5% growth compared to last week"
- Establish baselines and monitor deviations

**Distinguish correlation from causation**:

- "Users who use Feature A have higher retention" is correlation
- You need A/B testing to verify whether "Feature A improves retention" is causal

**Consider confidence levels**:

- Conclusions may be unreliable with small sample sizes
- Typically need at least 100-1,000 samples

## Naming and Organization Standards

### Analysis Naming Standards

**Use clear, descriptive names**:

```
Recommended:
"Daily Active User Trend - By Channel"
"Purchase Conversion Funnel - Mobile"
"New User 7-Day Retention - January 2024"

Not recommended:
"Analysis 1"
"test"
"Quick look"
```

**Naming format suggestion**:

```
[Metric/Analysis Type] - [Dimension/Filter Condition] - [Time Range (optional)]
```

### Analysis Organization Standards

**Create folders by topic**:

```
My Analysis
├── Daily Monitoring/
│   ├── Daily Core Metrics
│   ├── Weekly Data Report
│   └── Monthly Data Report
├── Feature Analysis/
│   ├── Search Feature Analysis
│   └── Recommendation Feature Analysis
├── User Growth/
│   ├── Channel Effectiveness Analysis
│   └── Acquisition Cost Analysis
└── Ad-hoc Analysis/
    └── ...
```

**Regular cleanup**:

- Delete or archive ad-hoc analyses promptly after completion
- Review monthly and clean up analyses that are no longer needed

## Do's and Don'ts

### Recommended Practices

**Do define your analysis goal before starting**:
- Think clearly about what question you want to answer
- Choose the appropriate analysis model

**Do use comparisons to enhance insights**:
- Compare with historical data
- Compare across groups
- Discovering differences is more valuable than looking at absolute values

**Do drill down from data to users**:
- Don't just look at numbers — understand the user behavior behind them
- Use user drill-down to discover root causes

**Do save and reuse analyses**:
- Save frequently used analyses as templates
- Add important analyses to Dashboards
- Regularly review and update

**Do record analysis conclusions**:
- Document key findings in analysis descriptions
- Facilitates future review and sharing

### Practices to Avoid

**Don't start analysis without a hypothesis**:
- Aimlessly "looking at data" is very inefficient
- Form a hypothesis first, then validate with data

**Don't only look at surface numbers**:
- Numbers going up doesn't necessarily mean good (could be fraud)
- Numbers going down doesn't necessarily mean bad (could be seasonal)

**Don't ignore sample sizes**:
- Conclusions from small samples may be unreliable
- Pay attention to statistical significance

**Don't over-interpret data**:
- Changes within normal fluctuation ranges don't need excessive analysis
- Avoid explaining normal fluctuations with low-probability events

**Don't cram too much into one analysis**:
- One analysis should focus on one question
- Break complex questions into multiple analyses

## Checklists

### Pre-Analysis Checklist

- [ ] Analysis goal and question to answer are clearly defined
- [ ] Appropriate analysis model is selected
- [ ] Data source and time range are confirmed
- [ ] Relevant events are confirmed to be instrumented

### During Analysis Checklist

- [ ] Filter conditions are set correctly
- [ ] Group By dimensions are appropriate
- [ ] Sample size is sufficient
- [ ] Data looks reasonable (no obvious anomalies)

### Post-Analysis Checklist

- [ ] Analysis conclusions are clear and definitive
- [ ] Naming is standardized and easy to understand
- [ ] Important analyses are saved
- [ ] Key findings are documented or shared

### Regular Maintenance Checklist

- [ ] Clean up analyses that are no longer needed
- [ ] Update outdated analysis configurations
- [ ] Review the accuracy of previous analysis conclusions
- [ ] Optimize frequently used analysis configurations

## Advanced Tips

### Using SQL for Complex Analysis

When built-in analysis models cannot meet your needs, use SQL Query:

**Common scenarios**:

- Multi-table join analysis
- Complex aggregate calculations
- Custom metric calculations
- Data export and processing

See [SQL Query](sql-query.mdx) for details.

### Combining Cohorts with Analysis

**Create analysis-specific Cohorts**:

1. Create Cohorts based on behavioral characteristics (e.g., "Users who purchased in the last 7 days")
2. Use Cohorts as filter conditions in analysis
3. Compare behavioral differences across Cohorts

See [Cohort Overview](../user-operation/cohort-overview.mdx) for details.

### Building an Analysis Framework

**Construct a metric framework**:

```
North Star Metric
├── Primary Metrics (Core Business Metrics)
│   ├── Active Users
│   ├── Revenue
│   └── Retention Rate
├── Secondary Metrics (Process Metrics)
│   ├── Feature Usage Rate
│   ├── Conversion Rate
│   └── User Growth
└── Tertiary Metrics (Detailed Metrics)
    ├── Per-channel Data
    ├── Per-feature Data
    └── Per-segment Data
```

## Next Steps

Now that you understand data analysis best practices, you can:

1. **[FAQ](faq.mdx)**: View answers to common questions
2. **[Event Analysis](event-analysis.mdx)**: Learn Event Analysis in depth
3. **[Retention Analysis](retention-analysis.mdx)**: Master Retention Analysis techniques

---

**Last updated**: January 19, 2026