Back StatQuestions Manual
ManualCore WorkflowUpload & Deduplication
Core Workflow

Upload & Deduplication

Turn raw email exports into clean, deduplicated case data.

Step-by-Step: Upload Your First CSV

Follow these exact steps to upload your first batch of customer emails:

1
Step 1: Go to Datasets page

Click "Manage Data" in the left sidebar (under Collect section). You will see the DashboardSelector page.

2
Step 2: Click "Upload New Dataset"

Top-right corner of the page. A blue button with Upload icon.

3
Step 3: Select your CSV file

File picker dialog opens. Choose your exported email CSV (from Salesforce, Outlook, or other CRM). Max 100MB.

4
Step 4: Map your columns

The Column Mapping screen appears. For each required field, select the matching column from your CSV dropdown:

5
• Email Body → select "Message Body" or "Description"

Required. This is the actual customer message text.

6
• Case ID → select "Case Number" or "Ticket ID"

Required. Unique identifier for each email/case.

7
• Sent Date → select "Date Opened" or "Created Date"

Required. When the customer sent this email.

8
• Business Unit → select "Division" or "Brand"

Optional. If you don't have this, leave as "Unassigned".

9
Step 5: Click "Continue to Classification"

Blue button at bottom-right of the mapping screen.

10
Step 6: Choose classification method

Select "AI Full Auto" for fully automated classification (recommended for first upload).

11
Step 7: Click "Start Processing"

Green button at bottom-right. Your upload begins processing immediately.

12
Step 8: Wait for completion

Processing takes 2-5 minutes for 500 emails. Watch the CXProcessingScreen for real-time progress bars.

13
Step 9: Click "View Dashboard"

Once complete, green button appears. Click to see your Unified CX Dashboard with all metrics.

CSV format tip: Export from Salesforce: Reports → New Report → Email Messages → Add columns → Export → CSV. Export from Outlook: File → Open & Export → Import/Export → Export to CSV → Select Mail folder.

Upload Screen Mockup

What you'll see when uploading your first dataset

Drag and drop your CSV file here

or click to browse (max 100MB)

Column Mapping
Email Body
Message Body
Case ID
Case Number
Sent Date
Date Opened
Business Unit
Select column...

AI Classification in Progress

Processing 500 emails • 2-5 minutes remaining

Deduplication ✓Classification (60%)Statistics (pending)
What Happens Behind the Scenes
1

Deduplication Engine

Collapses email threads into unique cases using subject + body fingerprinting

2

Keyword Classification

Scans for known taxonomy keywords (85% accuracy, instant)

3

LLM Fallback

Ambiguous cases sent to LLM with org context (94% accuracy, ~2 sec/email)

4

Statistical Analysis

Chi-square tests, DPU calculations, sentiment scoring

Expected Results (500 Email Batch)
Total rows uploaded500
Duplicates removed~75 (15%)
Unique cases classified425
Classification accuracy87-94%
Processing time2-5 min
Common Issues & Solutions

Error: "Missing required column: Email Body"

Cause: Your CSV doesn't have a column with the actual customer message.
Solution: Re-export from your CRM including the message/description/body field. In Salesforce: add "Message Body" to your report columns.

Issue: High duplicate rate (>30%)

Cause: Your CRM exports multiple replies per case as separate rows.
Solution: This is expected! StatQuestions deduplication collapses email threads into unique cases. A 30-50% duplicate rate is normal for support data. The system handles this automatically.

Warning: "Low classification confidence (<70%)"

Cause: Your emails are short, ambiguous, or lack taxonomy keywords.
Solution: This triggers LLM fallback automatically. Accuracy will be 92-95% but processing takes longer. Consider adding more context to your org taxonomy in Org Setup.

What Success Looks Like
487/500
Emails Classified
13
Duplicates Removed
87%
Classification Accuracy

Next Step: Click "View Dashboard" to see your Unified CX Dashboard with issue breakdowns, chi-square analysis, and sentiment trends.