Module 03 · Marketing Data Science Made Easy
Import & export finance trusts
Correct inputs beat clever models. This module is the unglamorous core of the course.
Rule. Sales window = spend window. Currency consistent. No double-counted channels. Then — and only then — compute or ask an AI.
Decision this module trains
Is this export clean enough to compute Total ROAS — or do I fix dates/duplicates first?
Sales export (Shopify)
- Pick Total Sales or Net Sales and label it on the file.
- Export or API-sum for the exact start/end dates you will use for spend.
- Refunds/returns: know whether your Net basis already nets them.
Spend import (CSV)
Free Mcfly path is paste/import: Day + Meta + Google (+ Other), or long date,channel,amount. Pro adds named channels. Sales columns in a spend file are ignored — do not mix.
date,channel,amount
2026-07-01,meta,1200.50
2026-07-01,google,890.00
- One row per channel per day (or a wide daily row — either works if consistent).
- Amounts in store currency; no “$” symbols; no thousands commas if your parser hates them.
- Include zero-spend days only if your process needs a continuous series — otherwise omit.
QA checklist (print this)
- Start date and end date match on sales and spend.
- No channel appears twice for the same day.
- Totals roughly match the ad UIs for that window (± fees timing).
- Timezone: store TZ vs ad account TZ — pick one story.
- File name includes the window: spend_2026-07-01_2026-07-31.csv.
Retrieve
Name three QA failures that must stop Total ROAS before any AI paste.
Paid pack has the worksheet: csv-qa-checklist.md + sample-spend-dirty.csv.