CleanSmartLabs Product Support

CleanSmart is an enterprise-grade AI-powered data cleaning platform that transforms messy, inconsistent data into clean, standardized datasets. Whether you're dealing with duplicate records, inconsistent formatting, missing values, or data from multiple sources, CleanSmart automates the tedious work of data cleaning so you can focus on what matters.

Key Features

• AI-Powered Cleaning: Advanced algorithms detect duplicates, standardize formats, fill missing values, and flag anomalies
• Multi-Source Merging: Combine data from multiple sources with intelligent conflict resolution
• Full Audit Trail: Track every change with the ability to approve or reject modifications
• Flexible Export: Download cleaned data in CSV or JSON formats with optional metadata

Two Main Workflows

• Single-Source Cleaning: Upload a single CSV file and run it through our 4-step automated cleaning pipeline
• Multi-Source Processing: Combine multiple data sources into a unified master dataset with enriched spoke files

Creating Your Account

1. Visit the CleanSmart registration page
2. Choose your plan (Starter, Pro, or Business)
3. Enter your details and create your account
4.  All plans include a 7-day free trial with full access to features

Your First Data Cleaning Project

Quick Start (Single-Source):

1. Navigate to Data Sources
2. Click Upload CSV and select your file
3. Review the field mapping and click Confirm
4. Go to Data Cleaning and click Start Cleaning
5. Review changes in the Change Log
6. View results in Analytics
7. Download your cleaned data from Export

The Data Sources page is your central hub for managing all uploaded data files and source groups.

Uploading Files

To upload a single file:

1. Click the Upload CSV button or drag and drop a file
2. CleanSmart analyzes your file and displays a preview
3. Review the detected columns and data types
4. Click Confirm to add the source to your workspace

Supported file formats:

• CSV (Comma-Separated Values)
• Files up to your plan's size limit

Field Mapping

After uploading, CleanSmart automatically detects field types (name, email, phone, address, etc.). You can customize this mapping:

1. Click the Configure button on any source
2. Review each column's assigned field type
3. Use the dropdown to change field types if needed
4. Click Save Mapping to apply changes

Accurate field mapping ensures better cleaning results, especially for:

• Phone number formatting
• Email validation
• Name capitalization
• Address standardization

Managing Sources

Each uploaded source shows:

• File name and upload date
• Record count (number of rows)
• Mapping status (configured or needs mapping)
• Last sync time (for integrations)

Actions available:

• Configure: Edit field mappings
• Delete: Remove the source (this cannot be undone)
• Clear All: Remove all sources from your workspace

The single-source workflow is a 4-step automated pipeline that cleans your data using AI-powered algorithms.

The 4-Step Pipeline

Step 1: SmartMatch (Duplicate Detection & Merging)

SmartMatch uses AI-powered semantic similarity to find duplicate records that traditional exact-matching would miss.

What it detects:

• Exact duplicates (identical records)
• Near-duplicates (e.g., "John Smith" vs "Jon Smith")
• Semantic duplicates (e.g., "IBM" vs "International Business Machines")

Configuration options:

• Composite Key Selection: Choose which fields to use for matching
• Sensitivity Threshold: Adjust how strict the matching is
• Data Type: Select entity, transactional, or line-item data

Results include:

• Total duplicates found
• Similarity scores (0-100%)
• Matching reasons (why records were flagged)

Step 2: AutoFormat (Format Standardization)

AutoFormat fixes inconsistent formatting across your data.

What it fixes:

• Name capitalization: "john smith" → "John Smith"
• Phone numbers: Removes letters, standardizes format
• Email addresses: Lowercase, typo detection
• Dates: Consistent date formatting
• Addresses: Standardized abbreviations (St., Ave., etc.)
• Common typos: Corrects frequent spelling errors

Step 3: SmartFill (Missing Value Imputation)

SmartFill uses machine learning to intelligently fill missing values based on patterns in your existing data.

How it works:

• Analyzes relationships between fields
• Predicts missing values using existing patterns
• Assigns confidence scores to each prediction
• Never fills fields where prediction confidence is too low

Example:

If most customers from "90210" zip code are in "Beverly Hills, CA", SmartFill can predict the city/state for records with only a zip code.

Step 4: LogicGuard (Anomaly Detection)

LogicGuard identifies outliers and impossible values that may indicate data quality issues.

What it detects:

• Numerical outliers: Ages over 150, negative prices
• Statistical anomalies: Values far from the mean
• Pattern violations: Phone numbers with wrong digit counts
• Impossible values: Future dates for birthdays

Detection methods:

• Z-score analysis
• Interquartile Range (IQR)
• Isolation Forest algorithm
• Pattern-based detection

Running the Pipeline

• Navigate to Data Cleaning
• Select your data source from the dropdown
• Click Start Cleaning to run all steps, or click individual steps
• Watch the real-time progress bar
• Review results for each step as they complete

Duplicate Resolution Options

When duplicates are found, you have several resolution options:

Automatic Resolution:

• Accept all AI-suggested merges with one click
• System chooses the most complete record as master

Manual Resolution:

• Review each duplicate cluster one by one
• Select which record should be the master
• Choose field-by-field which values to keep
• Reject false positives

Resolution Strategies:

• Master Record: One record becomes the source of truth
• Field-Level Merge: Combine best values from each duplicate
• Keep Both: Mark as not duplicates if incorrectly matched

Multi-Source Processing allows you to merge and enrich data from multiple sources using a hub-and-spoke architecture.

Key Concepts

Source Group: A container for multiple related data sources that should be merged together.

Relationships: Define how sources connect to each other (e.g., Customer ID in Source A matches CustomerID in Source B).

Hub (Master Dataset): The unified, merged dataset created by combining all sources.

Spokes: Individual source files enriched with data from other sources in the group.

Creating a Source Group

1. Go to Data Sources
2. Click Create Source Group
3. Enter a group name (e.g., "Customer 360")
4. Click Create

Adding Sources to a Group

Open your source group

1. Click Add Source
2. Either upload a new file or select an existing source
3. Configure field mapping for each source
4. Repeat for all sources you want to merge

Defining Relationships

Relationships tell CleanSmart how to connect records across sources.

1. Go to the Relationships tab in your source group
2. Review Suggested Relationships (AI-detected connections)
3. Click Accept to use a suggestion, or Customize to modify it

To create a manual relationship:

1. Click Add Relationship

2. Select the first source and field (e.g., customers.customer_id)

3. Select the second source and field (e.g., orders.customer_id)

4. Choose the relationship type:

    - 1-to-1: One record in Source A matches one in Source B

    - 1-to-many: One record in Source A matches multiple in Source B

    - Many-to-many: Multiple records can match in both directions

5. Configure matching options (exact match or fuzzy matching)

6. Click Save

Merge Strategies

Choose how conflicts between sources should be resolved:

Strategy  Description  Best For 

• MASTER_SLAVE  One source is always trusted over others  When you have a primary system of record 
• CONSENSUS  Values appearing in multiple sources win  When no single source is authoritative 
• WEIGHTED  Sources have trust scores that determine priority  When sources have varying reliability 
• RECENT  Most recently updated value wins  When newer data is more accurate 

The 6-Step Processing Pipeline

1. Schema Validation: Aligns column structures across all sources
2. Relationship Detection: Verifies and applies defined relationships
3. Data Merging: Combines records from multiple sources
4. Conflict Resolution: Handles conflicting values using your chosen strategy
5. Data Enrichment: Adds fields from other sources to each dataset
6. Quality Verification: Validates the merged data quality

Running Multi-Source Processing

Go to Multi-Source Processing

1. Select your source group
2. Review the configuration (merge strategy, relationships)
3. Click Start Processing
4. Monitor progress through each step
5. Review and resolve any conflicts that require manual attention

Conflict Resolution

When the same field has different values across sources, CleanSmart flags it as a conflict.

Automatic Resolution: Based on your merge strategy, most conflicts are resolved automatically.

Manual Resolution: Some conflicts may require your review:

1. Click Review Conflicts
2. See side-by-side comparison of conflicting values
3. Choose which value to keep
4. Optionally apply the same rule to similar conflicts
5. Click Apply Resolution

Cross-Source Duplicate Detection

CleanSmart can find duplicates that exist across your sources:

• Customer in Source A is also in Source B with slight variations
• Identifies and suggests merging these cross-source duplicates

The Change Log provides a complete audit trail of every modification made during cleaning.

Understanding the Change Log

Every change is categorized by type:

• Duplicate_Resolved (SmartMatch): Duplicate records merged
• Format_Standardized (AutoFormat): Format corrections applied
• Value_Imputed (SmartFill): Missing values filled
• Anomaly_Detected (LogicGuard): Outliers flagged
• Conflict_Resolved (Multi-source): Cross-source conflicts resolved

Change Details

Each change shows:

• Record Number: Which row was affected
• Field Name: Which column was modified
• Original Value: What the data was before
• New Value: What the data is now
• Confidence Score: How confident the AI is (color-coded)
• Change Type: Which step made this change

Reviewing Changes

Filtering Options:

• By status (All, Pending Review, Approved, Rejected)
• By field name
• By change type
• Search by value

Actions:

• Approve: Accept the change (keeps new value)
• Reject/Revert: Reject the change (restores original value)

Bulk Actions:

• Approve All: Accept all pending changes
• Reject All: Revert all pending changes

Confidence Scores

Changes are color-coded by confidence:

• Green (90-100%): High confidence likely correct
• Yellow (70-89%): Medium confidence review recommended
• Orange (50-69%): Lower confidence careful review
• Red (below 50%): Low confidence manual verification needed

Workflow Requirement

You must address all pending reviews before proceeding to Analytics. This ensures you've verified all changes before finalizing your cleaned data.

The Analytics page shows the impact of your data cleaning operations.

Summary Metrics

Four main cards show your cleaning results:

1. SmartMatch Merges: Number of duplicate records merged
2. AutoFormat Fixes: Number of format corrections made
3. SmartFill Predictions: Number of missing values filled
4. LogicGuard Flags: Number of anomalies detected

Data Quality Score

A before/after comparison shows your data quality improvement:

• Before: Original data quality percentage
• After: Cleaned data quality percentage
• Visual bar chart showing the improvement

Key Insights

CleanSmart provides actionable insights about your data:

Success Insights (Green):

"Merged 45 duplicate customer records"

"Standardized 230 phone numbers"

Warning Insights (Amber):

"12 anomalies detected in 'age' field"

"5 conflicts required manual resolution"

Recommendations:

Suggestions for improving data quality

Tips for better results on future uploads

Multi-Source Analytics

For multi-source processing, additional metrics show:

• Sources processed count
• Relationships applied
• Total records enhanced
• Total fields in merged dataset
• Cross-source duplicate statistics

The Export page allows you to download your cleaned data in various formats.

Single-Source Export

Step 1: Select Dataset

Choose which cleaned dataset you want to export.

Step 2: Choose Export Mode

• All AI Changes Applied: Exports data with all cleaning applied
• With Accept/Reject Choices: Respects your Change Log approvals/rejections

Step 3: Configure Options

• Include Change Tracking Metadata: Adds columns showing what changed

Step 4: Select Format

•  CSV  Comma-separated values  Spreadsheets (Excel, Google Sheets) 
•  JSON  JavaScript Object Notation  Web applications, APIs 
•  Change Log (CSV)  Detailed change history  Audit trail, compliance 
•  Summary Log (CSV)  Aggregated change summary  Reports, analysis 

Step 5: Download

• Click Export to generate and download your file.

Multi-Source Export

For multi-source groups, additional options are available:

Dataset Selection:

• Choose which source datasets to include
• Option to include the Customer Master Hub

Export Settings:

• Include Customer Master Hub: The merged master dataset
• Include Additional Fields: Fields enriched from other sources
• Include Audit Trail: Record of conflicts resolved and values chosen
• Include Change Tracking Metadata: Column-level change indicators

Spoke Export:

Export individual enriched source files that can be imported back into their original systems (e.g., CRM import files).

Export Preview

Before downloading, review:

• Total records to be exported
• Columns included
• File size estimate
• Sample of the data

The Settings page allows you to customize CleanSmart's anomaly detection behavior.

Anomaly Detection Parameters

Fine-tune how LogicGuard detects outliers and anomalies:

•  Statistical Threshold  2.5-5.0  Sensitivity for Modified Z-score detection. Lower = more sensitive 
•  IQR Multiplier  0.5-2.5  Width of acceptable range using quartiles. Lower = stricter 
•  Z-Score Threshold  0.5-5.0  Standard deviations from mean before flagging. Lower = more sensitive 
•  Isolation Contamination  0-50%  Expected percentage of outliers in data 
•  Categorical Rare Threshold  0-10%  Minimum occurrence percentage for categories 
•  Pattern Std Multiplier  0.5-5.0  Sensitivity for text pattern anomalies 

Quick Presets

Choose a preset configuration:

• Strict (High Sensitivity): Catches more potential issues. Good for critical data where false positives are acceptable.
• Balanced (Default): Recommended for most use cases. Balances detection with avoiding false positives.
• Relaxed (Low Sensitivity): Only flags obvious outliers. Good for data where some variation is expected.

Applying Settings

1. Adjust sliders or enter values directly
2. Preview how changes affect detection (if available)
3. Click Save Settings to apply
4. Use Reset to Defaults if needed

The Profile page lets you manage your account information and preferences.

Personal Information

Edit your account details:

• Full Name: Your display name
• Email Address: Your login email
• Account Role: Admin or Member (read-only)

For Organization Admins:

• Organization Name: Your company name (editable)

Email Preferences

Control which notifications you receive:

• Processing Notifications: Alerts when cleaning jobs complete
• Weekly Reports: Summary of your data cleaning activity
• Product Updates: News about new features and updates

Saving Changes

• Click Save Changes to update your profile. Changes take effect immediately.

The Team page (Pro and Business plans) lets organization admins manage team members.

Team Overview

View your team status:

• Seats Used: Number of active team members
• Seats Available: Remaining seats on your plan
• Progress Bar: Visual representation of seat usage

Team Members

Each member shows:

• Name and email
• Role (Admin or Member)
• Join date
• Actions (remove, for admins)

Inviting Team Members

1. Click Invite Member
2. Enter the email address
3. Choose a role (Admin or Member)
4. Click Send Invitation

The invitee will receive an email to join your organization.

Pending Invitations

View and manage outstanding invitations:

• See invited email addresses
• View expiration dates
• Cancel invitations if needed

Managing Members (Admin Only)

• Remove Member: Revoke access (frees up a seat)
• Change Role: Promote/demote between Admin and Member

Seat Limits

If you've used all seats:

• Click Purchase Additional Seats to add more
• Each additional seat has a monthly cost based on your plan

The Subscription page manages your plan, billing, and usage.

Current Usage

View your current billing period usage:

• Data uploaded (MB used / limit)
• Records processed
• Team seats (used / total)

Upgrading Your Plan

1. Click Upgrade on your desired plan
2. Review the new features and pricing
3. Complete payment via Stripe
4. New features activate immediately

Adding Seats

1. Go to Manage Billing
2. Increase the seat count
3. Review the price change
4. Confirm and pay for additional seats

Downgrading

1. Click Downgrade on a lower plan
2. Review what features you'll lose
3. Confirm the downgrade
4. Change takes effect at the end of your billing period

Note: If your team size exceeds the new plan's seats, additional seats will be charged.

Billing History

View all past invoices:

• Invoice date and number
• Amount charged
• Payment status
• Download PDF or view online

Cancelling Your Subscription

1. Click Cancel Subscription
2. Confirm your cancellation
3. Account remains active until period end
4. After cancellation date, you cannot log in
5. Data retained for 30 days

Reactivating

If you've scheduled a downgrade or cancellation:

1. Click Keep Current Plan or Cancel Scheduled Change
2. Confirm to continue with your current plan

CleanSmart integrates with popular marketing and CRM platforms (Pro and Business plans).

Mailchimp Integration

Import your email audiences directly from Mailchimp:

Setup:

1. Go to Data Sources
2. Click Connect on the Mailchimp card
3. Authorize CleanSmart to access your Mailchimp account
4. Select the audience to import

Importing:

1. Choose an audience from your Mailchimp account
2. Review the fields that will be imported
3. Click Import
4. Contacts are added as a new data source

Synced Fields:

• Email address
• First name, Last name
• Custom fields
• Tags and segments

Klaviyo Integration

Import customer lists from Klaviyo:

Setup:

1. Go to Data Sources
2. Click Connect on the Klaviyo card
3. Enter your Klaviyo API key
4. Authorize the connection

Importing:

1. Select a list from your Klaviyo account
2. Review the customer profile fields
3. Click Import
4. Profiles are added as a new data source

Synced Fields:

• Email, phone
• Name fields
• Custom properties
• Profile attributes

Q: How long does data cleaning take?

A: Processing time depends on your dataset size. Most files under 10,000 records process in under 2 minutes. Larger files may take longer.

Q: Can I undo changes after cleaning?

A: Yes! Use the Change Log to reject specific changes, which reverts them to original values. You can also re-export with "Accept/Reject Choices" to include only approved changes.

Q: Why weren't duplicates detected in my data?

A: Check your field mapping. Duplicate detection works best when fields are correctly identified (e.g., marking a column as "Name" helps SmartMatch compare names). Also, verify your composite key selection includes the right fields.

Q: How does SmartFill decide what values to predict?

A: SmartFill analyzes patterns in your existing data. It only fills values when it has high confidence based on relationships between fields. Low-confidence predictions are not applied automatically.

Q: Can I process the same file multiple times?

A: Yes. Upload the file again or re-run cleaning on the existing source. Previous changes in the Change Log will be preserved.

Q: What happens to my data after I cancel?

A: Your data is retained for 30 days after cancellation. After that, it is permanently deleted.

Common Issues

Upload fails:

• Ensure your file is in CSV format
• Check that the file isn't corrupted or empty
• Verify you haven't exceeded your plan's upload limit

Processing stuck or slow:

• Large files take longer to process
• Check your internet connection
• Try refreshing the page and checking job status

Field mapping incorrect:

• Re-configure field mapping on the source
• Manually set the correct field types
• Check that column headers are clear and descriptive

Duplicates not merging correctly:

• Review your composite key selection
• Adjust sensitivity thresholds
• Use manual resolution for complex cases

Export file is empty:

• Ensure cleaning has completed
• Check that you selected a dataset
• Verify changes weren't all rejected