We're sorry but this app doesn't work properly without JavaScript enabled. Please enable it to continue.

This lesson's interactive features are locked, please to keep using them

Data Analytics Workflow

Data work gets messy fast.

If you try to do everything at once, you'll miss bugs, skip edge cases, and build analysis on top of bad data. That's why I think about data work using these seven steps.

Click to play video

Step 1: Extract

Get the data from wherever it lives. Common sources:

  • APIs
  • Databases
  • CSV or Excel files
  • Event logs
  • Web scraping

The worst part of extraction is often not the code – it's getting access in the first place. Credentials, tokens, permissions... delightful.

Step 2: Inspect

Before you touch the data, look at it. You want quick answers to questions like:

  • What fields exist?
  • What types are they?
  • How many rows are there?
  • What do a few sample records look like?

Step 3: Clean

Fix the obvious problems. That usually means:

  • Filling or removing missing values
  • Removing duplicates
  • Standardizing formats
  • Fixing bad data types

People use the word normalize here a lot. That's fine in practice, just know it means something more specific in database design.

Step 4: Transform

Now make the data more useful by reshaping it for analysis or storage. Common transformations:

  • Rename columns
  • Create derived fields
  • Merge data from multiple sources
  • Reshape rows and columns

Step 5: Aggregate

Roll the raw data up into summaries. This is how you get from "a mountain of records" to "something the product team can use in Excel." Examples:

  • Average order value by region
  • Sign-ups by day
  • Cancelled subscriptions by plan
  • Streams per artist

Step 6: Analyze

Now dig in to answer the actual question, things like:

  • How many subscribers churned last week?
  • Which playlists keep listeners engaged the longest?
  • Where are checkout conversions dropping unexpectedly?
  • Are refund requests increasing after the latest app update?

Step 7: Deliver

Hand the results to "stakeholders" (corpo-speak for "people that care"). This might take the form of:

  • A dashboard
  • A report
  • An API
  • A scheduled email or Slack message

Example

As a data analyst at a music streaming service, your workflow end-to-end might look like this:

  1. Extract: Pull yesterday's listening events from an API
  2. Inspect: Check fields, types, row count, and sample records
  3. Clean: Fix missing track IDs and bad timestamp formats
  4. Transform: Add a minutes_played field and standardize artist names
  5. Aggregate: Calculate total plays per genre and listeners per day
  6. Analyze: Notice that lo-fi playlists spike every weekday at 9am
  7. Deliver: Update the dashboard and send a summary to the content team

In real-world data work, you'll spend most of your time in Extract, Clean, and Transform. That's why the industry is obsessed with ETL/ELT pipelines.