S5 ICT: Big Data and the Data Analysis Pipeline (three Vs, cleaning, MapReduce) Tutor Prep Guide with a 45-minute plan, teaching script and common errors

2026-11-15

Tutor lesson-prep guides · Secondary ICT

Students hear big data and collect new words, yet stop when asked how many rows survive cleaning or how much faster parallel work is. This lesson turns the pipeline into five steps, each one only addition, subtraction, multiplication or division.

Objectives — By the end of the lesson students can: (1) describe what each of the five pipeline steps does and how the data volume changes at each one; (2) multiply 5 million rows by 240 bytes to get 1.2 GB and name the three Vs as volume, velocity and variety; (3) subtract 4 per cent missing (200,000 rows) and 1.5 per cent duplicate (75,000 rows) to leave 4,725,000 rows, then split them across 6 shards at 787,500 rows each; (4) take 4 map tasks of 25 minutes plus a 10-minute reduce to get 110 minutes sequential, 35 minutes in parallel and a 3.14 times speedup, and explain that correlation is not causation.

01 Prerequisites (3-minute check)

If any of these fails, fix it before the lesson starts.

Skill	Quick check	Pass mark
Percentages	What is 4 per cent of 5 million rows	200,000 rows
Unit conversion	How many GB is 1,200,000,000 bytes	1.2 GB
Division	4,725,000 rows shared across 6 shards	787,500 rows
Rates	10 million events a day is how many a second	115.74 events

02 Lesson flow (45 minutes)

Time	Stage	Tutor does	Student does
0-6 min	Entry check	Ask 4 per cent of the rows, 1.2 GB, rows per shard and events per second; keep the answers on the board	Spoken answers, working on paper
6-16 min	Act one: five steps, each one shrinking the data	Draw five arrows and write the volume above each: 5,000,000 rows, 4,725,000 rows, 1.134 GB, then shards	Students name the step where the data starts shrinking
16-26 min	Act two: the three Vs and cleaning as subtraction	Volume 1.2 GB, velocity 115.74 events a second, variety as three formats; subtract missing and duplicate rows	Students compute 4,725,000 and 787,500
26-36 min	Act three: MapReduce and presentation	4 map tasks of 25 minutes: 110 sequential, 35 parallel, a 3.14 times speedup; always ask for a third variable	Students compute 3.14 and explain why it is not 4
36-42 min	Question types with changed numbers	Switch to 8 million rows, 5 per cent missing and 8 map tasks, then run the four question types	Students do one and write the unit (rows, GB, seconds)
42-45 min	Exit check	Three spoken questions; note which act each error comes from	All three correct

03 Teaching script (read it out)

Scene 1: Act one: five steps, and every step shrinks the data

Data analysis is not a single program but a pipeline. Collection brings in sensor readings, web logs and transactions, here 10 million events a day. Cleaning deals with missing, duplicate and badly formatted values, and this step only ever removes rows, never adds them. Storage puts the survivors into a database or a distributed file system, usually sharded by key, so six shards mean six pieces. Analysis computes averages, groups and trends, and at this scale splits into map and reduce. Presentation shows dashboards, charts and reports. In an exam, a question about rows is answered by subtracting step by step from the source, and a question about GB by multiplying rows by bytes per row. None of the five steps may be skipped.

Can the row count rise after cleaning above the original 5 million, and why not?

Scene 2: Act two: the three Vs point in three directions, and cleaning is subtraction

Big data has three Vs. Volume is size: 5 million rows times 240 bytes is 1,200,000,000 bytes, or 1.2 GB, about 1.12 GiB in binary units. Velocity is speed: 10 million events a day divided by 86,400 seconds is about 115.74 a second, which is why streaming is used instead of a nightly load. Variety is format: one system holds structured transaction tables, semi-structured logs and unstructured text. Cleaning is subtraction: 4 per cent missing is 5 million times 0.04, or 200,000 rows; 1.5 per cent duplicate is 75,000 rows; remove both and 5,000,000 minus 200,000 minus 75,000 leaves 4,725,000 rows, so the size drops from 1.2 GB to 1.134 GB and six shards hold 787,500 rows each.

If the missing and duplicate rows overlap, does plain subtraction overcount, and how do you fix it?

Scene 3: Act three: parallelism has a ceiling, and presentation must ask about cause

Take four map tasks at 25 minutes each. On one machine they run in sequence for 100 minutes, plus a 10-minute reduce, so 110 minutes. On four machines they run together in 25 minutes, but the reduce still takes 10, so 35 minutes in total. The speedup is 110 over 35, about 3.14 times rather than 4, because reduce cannot be parallelised and starting machines and moving data both cost time. Presentation deserves caution: a chart showing ice cream sales and drownings rising together does not mean ice cream causes drowning, since both follow the weather. The value of the pipeline is asking whether a third variable explains both.

If the map tasks rise to 8, does the speedup pass 6 times, and why not?

04 Question types and worked steps

Type	Example	Steps	Answer
Rows after cleaning	5,000,000 rows with 4 per cent missing and 1.5 per cent duplicate	Compute each group in rows, then subtract in order	4,725,000 rows
Capacity	5,000,000 rows at 240 bytes each	Multiply, then divide by 1,000,000,000 for GB or by 2 to the power 30 for GiB	1.2 GB, about 1.12 GiB
Throughput	10 million events a day	Divide by 86,400 seconds, and multiply back to verify	About 115.74 events a second
Parallel speedup	4 map tasks of 25 minutes and a 10-minute reduce	Sequential total divided by parallel total	110 over 35, about 3.14 times

Sanity check (teach this): Four checks: (1) the cleaned row count must be below the original, and a larger one means an addition slipped in; (2) GB uses a billion bytes and GiB uses two to the power thirty, so follow the unit the question states, both numbers being right but not interchangeable; (3) events per second times 86,400 must return the daily total, or a wrong divisor was used; (4) the speedup must be below the number of map tasks, since 4 from 4 maps means the reduce was treated as free.

05 Independent practice (3 questions)

Question	Answer	What to watch
5,000,000 rows with 4 per cent missing: how many rows are removed	200,000	Check 5,000,000 times 0.04, not times 4
4,725,000 rows across 6 shards: rows per shard	787,500	4,725,000 divided by 6, not the original 5 million
4 map tasks of 25 minutes and a 10-minute reduce: the speedup	About 3.14 times (110 over 35)	The denominator must include the 10-minute reduce

06 Common mistakes and what to say

Mistake	Student writes	Say this
Treating cleaning as addition	Answering 5,000,000 plus 200,000 plus 75,000	Missing and duplicate rows are removed, so the operation is subtraction. Adding them treats bad data as new data and the dataset grows.
Mixing GB with GiB	Writing 1.2 GiB for 1.2 GB	1.2 GB is about 1.12 GiB. Both figures are right but the units are not interchangeable, so follow whichever the question states.
Assuming parallelism is free everywhere	Answering 35 minutes while dropping the reduce, or claiming a 4 times speedup	Only the map phase is parallel, the reduce still takes 10 minutes, so the figure is 110 over 35, about 3.14 times.
Reading correlation as cause	Concluding ice cream causes drowning because both rise together	Both follow the weather, a common cause. Before presenting, ask whether a third variable explains both, or the report misleads its reader.

07 Homework (paper, section and question numbers)

Secondary ICT prep guide: big data and the data analysis pipeline chapter (question types and chapter exercises) — Chapter exercises, questions 1 to 4

08 Exit check (last 5 minutes)

Question	Answer	Decision
5 million rows with 4 per cent missing and 1.5 per cent duplicate: rows left	4,725,000	Wrong: reteach act two, both groups are subtracted
4 map tasks of 25 minutes and a 10-minute reduce: parallel total	35 minutes	Wrong: reteach act three, maps run together but reduce is added back
Two lines rise together on a chart: may we claim cause?	No, first ask about a third variable	Wrong: reteach act three, correlation and cause

09 Tutor FAQ

Must S5 students write MapReduce code?

No. The exam tests the concept and the arithmetic: what each step does, how the volume changes and how much faster parallel work is. Describing the split between map and reduce in words is enough.

How does this differ from S4 spreadsheet work?

S4 handles a few hundred rows on one machine, so formulas, filters and pivot tables suffice. This chapter handles millions of rows, which brings sharding, streaming and parallelism. The same sum, at a different scale, needs a different method.

Are 1.2 GB, 4,725,000 rows and 3.14 times real figures?

No, they are teaching numbers chosen for mental arithmetic. Real cleaning ratios and speedups vary widely, and the point is the algorithm and the unit at each step.

Find a tutor — free Tutor Registration →

A TutorZone lesson-prep guide. Written by the TutorZone team; questions and lesson flow are original and are not taken from any school or publisher material.