What you can and cannot learn on a phone during your commute

Sort tasks by what a phone allows

A phone has a small screen, no real keyboard, frequent interruptions, and often a patchy connection. That's fine for some kinds of learning and useless for others.

The predict-the-output drill

This is the best commute exercise for a data learner. At your desk the night before, write a few short queries on a small table and save them as a note or screenshot without the answers. On the train, read each one and write your predicted result in your notes app. Back at your laptop, run them and compare.

Here's a three-card set with synthetic data, and the script that checks your predictions:

import sqlite3

con = sqlite3.connect(":memory:")
con.executescript("""
CREATE TABLE riders (rider_id INTEGER, line TEXT, trips INTEGER);
INSERT INTO riders VALUES (1,'Blue',12),(2,'Blue',NULL),(3,'Aqua',7),(4,'Aqua',7),(5,'Aqua',3);
""")

# Your predictions, written on the commute
cards = {
    "Q1": ("SELECT COUNT(*), COUNT(trips) FROM riders;", [(5, 4)]),
    "Q2": ("SELECT line, ROUND(AVG(trips), 2) FROM riders GROUP BY line ORDER BY line;",
           [("Aqua", 5.67), ("Blue", 12.0)]),
    "Q3": ("SELECT COUNT(DISTINCT trips) FROM riders WHERE line = 'Aqua';", [(2,)]),
}

for name, (query, predicted) in cards.items():
    actual = con.execute(query).fetchall()
    print(name, actual, "OK" if actual == predicted else "CHECK")
    assert actual == predicted

Each card tests something that often trips people up. Q1: COUNT(*) counts rows, but COUNT(trips) skips the NULL. Q2: the Blue average is 12, not 6, because AVG ignores the NULL instead of treating it as zero. Q3: two Aqua riders have 7 trips, so there are only two distinct values. If you get one wrong, that's the concept to look up on the ride home. Write your own cards on whatever you studied that week, and make sure at least one card has a trap in it.

A plan for two 40-minute rides

Adjust the minutes to your route. The split of tasks matters more than the exact timing.

Morning ride (fresh attention):

1. 1First 5 minutes: reread yesterday's error-log entries.
2. 2Next 20 minutes: predict-the-output cards, five to eight of them.
3. 3Last 15 minutes: read one official documentation page on a function you'll use tonight, and note one question.

Evening ride (tired attention):

1. 1First 20 minutes: listen to one explanation-heavy segment of a class recording.
2. 2Last 20 minutes: write tonight's laptop plan in three lines: what you'll build, which cards to check, which question to ask your mentor.

The evening plan is the most important part. People who arrive home with a plan usually start on it. People who arrive home without one tend to open a video.

Make the commute reliable

• Assume you'll be offline. Signal drops in tunnels and underground stretches. Before you leave, save your cards, notes and any documentation pages you want to read. If your course allows downloads, download the recordings too. If it doesn't, keep the offline material for the underground parts.
• Standing and crowded is fine. Flashcards and predict-the-output work with one hand. Save anything that needs two hands for when you get a seat.
• Interchanges break focus. On a route with a change, for example from the Delhi Metro's Blue Line to the Noida Metro's Aqua Line, plan short tasks that fit between stations rather than one long block.
• Protect your eyes and battery. Use a larger font and a dark theme, and carry a power bank if your phone is also your ticket and payment method.

Don't count commute time as build time

Commute study is valuable, but it isn't practice in writing code. If you track your weekly hours, as in measure your learning progress weekly, log commute minutes separately. Two hours a day on trains can make a week feel full even when you haven't written a single query.

The Data Analyst programme runs over 3–4 months, live online with optional in-person labs at our Noida centre, and its sessions are recorded. Pair its recordings with a system for timestamped notes so the commute segments you listen to connect to what you build later.