← Ugh, stats

Case 3 · outliers · mean v median

One class, 3,667 answers

1Ask

On 12 November I looked at the last four weeks: 13,664 answers in 28 days, a mean of 488 a day. I was pleased. That looked like growth. Was it?

2Collect

Then I looked at the days. On Tuesday 3 November a single class in Malaysia answered 3,667 questions. On 23 October a class in Wales answered 3,099. Take those two days out and the other 26 days average 265. Two classes, on two days, had nearly doubled my "typical" day.

The week of 2 November
Mon 2Tue 3Wed 4Thu 5Fri 6
1083,667402228578

3Picture it

Answers each day, 15 October to 11 November 2026

The two tall bars are single classes.

Show the figures as a table
Answers each day, 15 October to 11 November 2026
DayAnswersWhole-class day
Thu 15 Oct513no
Fri 16 Oct451no
Sat 17 Oct71no
Sun 18 Oct140no
Mon 19 Oct214no
Tue 20 Oct142no
Wed 21 Oct118no
Thu 22 Oct586no
Fri 23 Oct3,099yes
Sat 24 Oct85no
Sun 25 Oct124no
Mon 26 Oct732no
Tue 27 Oct172no
Wed 28 Oct735no
Thu 29 Oct80no
Fri 30 Oct27no
Sat 31 Oct51no
Sun 1 Nov93no
Mon 2 Nov108no
Tue 3 Nov3,667yes
Wed 4 Nov402no
Thu 5 Nov228no
Fri 6 Nov578no
Sat 7 Nov228no
Sun 8 Nov115no
Mon 9 Nov310no
Tue 10 Nov357no
Wed 11 Nov238no

4Decide

A value that sits far away from the rest is an outlier. The mean adds every value, so one outlier drags it a long way. The median only looks at the middle value, so it barely notices. This is why the Office for National Statistics reports the median wage as its headline figure for pay. A few very high earners pull the mean up, so the mean describes nobody's pay packet. The median is what the person in the middle earns. So when we talk about a typical day now, we use the median, and we report big class days on their own.

What this data can't tell us

Your turn

Download the data (CSV)