AI Sales Analytics Across 3 Channels — One Dashboard to Rule Them All

The Multi-Channel Data Nightmare

My restaurant sells through 3 main channels:

Dine-in — data from POS
Delivery Platform A — sends PDF reports via email
Delivery Platform B — sends CSV reports via email

Every channel sends data daily, but the formats are completely different.

POS sends HTML, Platform A sends PDF, Platform B sends CSV — each summarizing sales in different units. Some include GP already deducted, some don't. Some include packaging costs, some don't.

The result: not a single month with reliable consolidated numbers.

Before AI: Manual Spreadsheet Wrangling Every Week

The old workflow looked like this — every week I had to:

Dig through email for report files
Copy numbers into Excel by hand
Calculate GP deductions manually
Merge everything into a grand total

It took about 1-2 hours per week and was error-prone.

The Tools I Actually Used

My main stack:

Claude Code — wrote Python scripts to parse each data format
Gmail IMAP — pulls emails automatically via Python imaplib
SQLite — stores data locally, lightweight, fast, no server needed
Python pandas — handles tables, pivot calculations by menu item

No-code alternatives:

Google Sheets + Apps Script — if your data volume is small, a simple Apps Script can pull CSVs from email into Sheets
n8n — use the "Gmail", "CSV", and "Google Sheets" nodes, drag-and-drop to connect them — minimal coding, great for visual automation

After AI: Fully Automated, Every Day

I had Claude Code build a system that:

Pulls data from email automatically

Every morning, the script:

Connects to Gmail via IMAP
Finds emails from Platform A, Platform B, and the POS system
Downloads attachments (PDF/CSV) or reads the email body (HTML)
Parses out the data — sales totals, order counts, GP amounts

The prompt I gave Claude Code to write the email script:

Write a Python script that automatically pulls sales data from Gmail.

Details:
- Gmail: use IMAP (imaplib)
- Emails from Platform A: sender "no-reply@platform-a.com"
  Has PDF attachment named "sales-report-*.pdf"
  Net sales figure is in the PDF on a line starting with "Net Sales:"
- Emails from Platform B: sender "report@platform-b.com"
  Has CSV attachment
  Sales amount is in column "order_amount", GP is in column "commission"
- Emails from POS: sender "report@pos-system.com"
  Data is in the email body as an HTML table
  Need the row containing "Total Sales"

Output: save to SQLite table daily_sales
columns: date, channel, gross_sales, gp_amount, net_sales

Parses each format differently

PDF (Platform A) → pdfplumber extracts text → regex finds "Net Sales: X,XXX.XX"
CSV (Platform B) → pandas read_csv → sum the relevant columns
HTML (POS)       → BeautifulSoup parses the table → extracts the Total row

Claude Code wrote all of this. I just explained where the data lives in each format and showed it sample files.

Calculates real profit automatically

What platforms report is "sales" — not "profit."

Restaurants selling through delivery platforms have to deduct GP (Gross Profit sharing), which is massive — some platforms take 40-50% of gross sales. And that's before packaging, bags, and labels.

The system calculates:

Real profit = Sales - GP - packaging - raw material cost

Then displays profit margin per order and per menu item.

Why GP Must Be Calculated Separately

Platforms typically report only the "amount received" after GP deduction, but don't break down the GP percentage per SKU. Without doing the math yourself, you'll never know which menu items are actually profitable.

What We Discovered in the Data

A few low-priced items were bestsellers on delivery, but when you factor in GP, packaging, and raw materials — the margin was negative.

The more we sold, the more we lost.

Without consolidated numbers like this, we'd never have known — the gross sales figures looked great.

Finding 2: Total Sales Were Higher Than Expected, But Profit Was Lower

Combined sales across 3 channels were higher than estimated, but after deducting delivery GP — net profit was significantly lower than expected.

This data drove the decision to push dine-in sales harder for certain menu items rather than delivery.

Finding 3: The "Best-Selling" Channel Wasn't the Most Profitable

The platform with the most orders wasn't the platform generating the most net profit, because GP rates differ between platforms.

This only becomes obvious when you have a dashboard that consolidates all channels in one place.

The Dashboard

The consolidated dashboard shows:

Data	Details
Total sales	Broken down by channel + combined
GP deducted	Per channel, per day
Net profit	After GP + packaging
Top menu items	Ranked by sales vs. by profit
Weekly trend	Week-over-week comparison

It updates automatically every morning. I just open it and look — no manual work needed.

Key Takeaway

The data sitting in your email every day is incredibly valuable — it's just not being parsed yet. AI handles the boring, error-prone part so you can focus on making decisions from the data instead.

How to Get Started

Step 1: Gather sample reports

Download reports from every channel for 1 month — PDF, CSV, whatever format. Figure out where the key numbers live in each one.

Step 2: Explain the data layout to AI

Open Claude Code and send a prompt like this:

I have sales reports from 2 delivery platforms in different formats.
I want to merge the data into a single table.

Platform A sends CSV with these columns:
[paste the CSV header row here]
Net sales after GP deduction is in column: "settlement_amount"

Platform B sends PDF
[attach a sample PDF or describe where the numbers are]

I need a Python script that:
1. Reads both file types automatically from a folder
2. Merges into one table: date, channel, gross_sales, gp_deducted, net_sales
3. Exports as a daily CSV

Step 3: Decide where to store your data

Google Sheets — fine for getting started; Python can push data via the Google Sheets API
SQLite — if you have a lot of data or want complex queries; free, no server needed
Airtable — if you want a UI to browse data without building a dashboard yourself

Step 4: Create the report template you want to see

I have daily sales data in Google Sheets called "Sales Data"
columns: date, channel, gross_sales, gp_deducted, net_sales

Create a Google Apps Script that:
1. Summarizes net_sales daily, broken down by channel
2. Calculates profit % relative to gross_sales
3. Displays a summary in a sheet called "Weekly Report"
   format: week, channel_a_net, channel_b_net, total, avg_margin%
4. Runs on a trigger every Monday at 8:00 AM automatically

Step 5: Set up a cron job to run automatically

If you're using a Python script on Mac:

# Open the crontab editor
crontab -e
 
# Add this line — runs every morning at 7:00 AM
0 7 * * * /usr/bin/python3 /path/to/your/fetch-sales.py >> /tmp/sales-log.txt 2>&1

The first setup might take half a day, but after that the system runs on its own every day — zero maintenance.