Time series

1One value, lined up on dates

A time series is not 'any table with a date column'. It is one measured thing — here, till in rupees, the money in the drawer at closing — written in date order, one row per day. Two weeks is fourteen rows, not eight customers. The customer sheet can be sorted by spend. The till series cannot, or Saturday stops being after Friday.

Week 1, Monday through Sunday, is 400, 420, 380, 450, 900, 1100, 800. Add them and the week is Rs 4450. If you shuffle those seven numbers by size, the series is gone: you still have the same rupees, but you no longer have 'over time'.

Figure. Week-1 till in day order: Mon 400 through Sat 1100 and Sun 800. The seven heights are the series. Shuffle them by size and Saturday is no longer after Friday.

Coding lab. Sum week 1 in day order runs in the app, with checks on your output.

2The slow rise versus the weekend bump

Put the two week totals next to each other before you reach for a word. Week 1 is Rs 4450. Week 2 is 410 + 430 + 390 + 460 + 920 + 1080 + 820 = 4510. That is a rise of Rs 60 on a Rs 4450 week — a small lift, not a leap. The slow change from one week to the next is what people call a trend.

Now look at the same weekday across the two weeks. Saturday is 1100 then 1080. Sunday is 800 then 820. Monday is 400 then 410. The Saturday jump happens both weeks, then Monday falls back near 400. A bump that repeats every weekend, then drops, is what people call season here — not a forecast word, just 'the shape that comes back'. Do not call the Saturday spike a trend. A trend would keep climbing through next Monday.

Figure. Saturday stays the peak both weeks (1100 then 1080). Monday stays near 400 both weeks (400 then 410). The weekend bump repeats; the week-total lift of Rs 60 is the slow rise.

Coding lab. Week totals side by side runs in the app, with checks on your output.

3A small pandas path

Put the dates in a column, sort them, then mark which days are Saturday and Sunday. pandas numbers weekdays from Monday: Monday is 0, Tuesday is 1, Wednesday 2, Thursday 3, Friday 4, Saturday 5, Sunday 6. So 'the weekend' is the days whose number is 5 or 6 — written dayofweek >= 5. That flag is a yes/no on each day, not a model.

Add till inside that flag and you recover the weekend-versus-weekday story from the same seven (or fourteen) rows. Week 1 weekdays are 400 + 420 + 380 + 450 + 900 = 2550. Saturday plus Sunday is 1100 + 800 = 1900. Two weekend days hold 1900 of the 4450 week. You do not need a forecast package or a twenty-page elective.

Figure. Week-1 weekdays sum to 2550; Saturday plus Sunday is 1900. Two weekend days hold 1900 of the 4450 week. dayofweek >= 5 is that flag.

Weekend vs weekday till

import pandas as pd
days = pd.date_range('2026-01-05', periods=7, freq='D')
till = [400, 420, 380, 450, 900, 1100, 800]
s = pd.Series(till, index=days)
weekend = s.index.dayofweek >= 5
print(int(s[weekend].sum()), int(s[~weekend].sum()))

Coding lab. Flag the weekend runs in the app, with checks on your output.

4Lab: weekend till

Build the seven-day week-1 series and print the weekend sum. days is a Monday-to-Sunday date range starting 2026-01-05. till is the seven rupee amounts you already added by hand. dayofweek >= 5 flags Saturday (5) and Sunday (6). The check looks for 1900 — Saturday 1100 plus Sunday 800.

This course stops after that sum because the question was the repeating bump, not a twenty-page forecast. You have not held out a test fold and you have not fitted a model. You have added the two weekend days.

No diagram — the weekend sum is printed by the coding lab.

Coding lab. Sum the weekend till runs in the app, with checks on your output.