CS23Exam Technique: Papers 1 and 2

• Paper 1, Computational thinking and programming skills: algorithms and programming. 2 hours, 90 marks, 50% of the GCSE.
• Paper 2, Computing concepts: data representation, computer systems, networks, cyber security, databases and SQL, and ethical, legal and environmental impacts. 1 hour 45 minutes, 90 marks, 50%.
• Both are written exams on paper. Answer every question: there are no choices and no tiers. Both papers are in the same June.
• No calculator on either paper. The front cover of both says so.
• There is no coursework. Your school confirms you had the chance to do programming tasks, but that earns no marks.
• Time per mark: Paper 1 is 120 ÷ 90 ≈ 1.3 minutes a mark; Paper 2 is 105 ÷ 90 ≈ 1.2 minutes a mark.
• Paper 1 gets harder as it goes. The last two or three questions are code-writing tasks worth 8 to 11 marks each. Save time for them.
• Paper 2 usually starts with number questions. SQL comes near the end, and there are one or two Discuss questions of 6 or 9 marks.
• Questions can be synoptic: Paper 1 might need binary or a file size inside an algorithm, because both papers assume the whole course.

Same marks, different time: Paper 2 is 15 minutes shorter.

Synoptic question

A question that needs knowledge from more than one part of the course.

• AQA sets Paper 1 in three versions: C#, Python 3 or VB.NET. Your school chooses one, and you write all your code in it. This site uses Python, the most common choice; your school may use C# or VB.NET, and the technique on this page is the same.
• Answer in the language the question asks for. “Your answer must be written in Python” means pseudo-code can't get full marks. An answer in a different language is passed to a senior examiner and may not get all the marks.
• Questions that print code to read use AQA's pseudo-code (←, OUTPUT, USERINPUT) or your language. When you are asked to answer “using pseudo-code”, any clear, consistent pseudo-code is fine.
• “Develop an algorithm using either pseudo-code or a flowchart” means either is fine. A paragraph of English is neither.
• Past papers: AQA's June 2022 to June 2025 papers, mark schemes and examiner reports, plus sample papers. Paper 1's content didn't change for exams from 2027, so every Paper 1 question is still fair practice.
• Paper 2 lost some content for 2027. No longer examined: Von Neumann architecture, optical storage, star and bus topologies, Ethernet and Wi-Fi as named protocols, UDP and FTP. A few past-paper questions use them: skip those (table below).
• Read the examiner reports. They say what answers got the marks, and what lost them.

Pseudo-code

A way of writing an algorithm that looks like code but isn't a real programming language. AQA prints its own style in exam questions.

The command word says what kind of answer gets the marks. Underline it, the number asked for (“two”), the context, and any instruction about form (“in Python”).

• Multiple choice: completely fill in the lozenge. “Shade one” means one: shade two and you get nothing. “Shade two” means exactly two.
• To change a lozenge, cross it out and shade the new one. To go back to a crossed-out answer, draw a ring round it.
• Read every option before shading. Wrong options are usually near-misses: a wrong boundary, the wrong layer, the converse of the right statement.
• Definitions are where many easy marks go. Examiners say students need to learn them “parrot fashion”. Keep a list and test yourself.
• Marks tell you how much to write. 1 mark: one point. 2 marks: two points, or one point with its reason. Stop when you have done what the marks need.
• Write in black ink. Use pencil only for drawing. Cross through any rough work you don't want marked.

Command word

The word in a question that tells you how to answer it, such as State, Explain or Discuss.

Lozenge

The oval you fill in to choose an answer in a multiple-choice question.

• Each mark is a separate idea. Two marks need two different points, or a point and its development. Saying the same thing twice in other words earns one.
• “Describe two differences”: every difference names both things. The second difference must be a new idea, not the first one turned round.
• Qualify every comparison. “Faster”, “cheaper”, “easier”, “more efficient” and “more secure” earn nothing on their own. Say what is faster and why.
• Use precise words: “data” (not “information”) for the CPU, memory and databases; “number” (not “amount”); never “things” or “stuff”.
• Use the context. If the question is about a garage or a school, every point should be about the garage or the school.
• Don't offer alternatives. Examiners mark the first answer you give. If you change your mind, cross the first one out.

Qualified comparison

A comparison that says what is better and why, such as “takes less time to load because the file has fewer bits”.

• Most calculations are on Paper 2: binary and hexadecimal conversions, units, image and sound file sizes, and Huffman bit counts.
• Show your working. Write the method, then the numbers, then the answer with its unit. Examiners say most marks for wrong answers come from working.
• Units go up in 1000s: 1 kB = 1000 bytes, 1 MB = 1000 kB, 1 GB = 1000 MB, 1 TB = 1000 GB. AQA's mark schemes reject 1024.
• Bits to bytes: ÷ 8. Then ÷ 1000 for each step up (bytes → kB → MB).
• ASCII uses 7 bits per character when you compare it with Huffman coding.
• Make the sum easy before you do it. Cancel the ÷ 8 against a number that divides by 8, and split awkward numbers: 48 × 25 = 48 × 100 ÷ 4 = 1200.
• Answer in the unit asked for (bits, bytes, kB or MB), and check the size is sensible: a short sound clip isn't 3 bytes or 3 TB.

• Work line by line. Never guess what a loop does: follow it. Under half of students could state one algorithm's output in a 2023 question.
• One column per variable, plus OUTPUT. Write a new value only when it changes. “You may not need all the rows”: leave spare rows blank.
• Outputs inside a loop appear on every pass, so the OUTPUT column fills up as you go. An output after the loop appears once.
• Write outputs exactly: no quotation marks round text, and for an f-string the text with the values in, not the braces.
• DIV and // give whole numbers; / can give a decimal. MOD and % give the remainder.
• Indexes start at 0 unless the question says otherwise.
• “State the purpose” wants plain words about the result (“it counts the even numbers”), not a line-by-line description.

total ← 0
FOR n ← 2 TO 8 STEP 3
    total ← total + n * n
    OUTPUT total
ENDFOR

Trace table

A table that records the value of each variable, and each output, as an algorithm is followed by hand.

• Tick off the bullet list. AQA's own advice: each bullet in the question is usually worth marks. Tick each one as your code covers it.
• “Keep asking”, “repeat until”, “for each” all mean a loop. A missing loop is the commonest big loss on Paper 1.
• Convert input: input() gives a string, so use int() or float() for numbers.
• print, not OUTPUT. Examiners say OUTPUT in a Python answer can't get the mark, every year.
• Return or output? If the brief says a subroutine returns a value, use return; if it says output, use print. The wrong one gets no credit.
• Write each comparison in full: day == "Sat" or day == "Sun", never day == "Sat" or "Sun". Join limits with and (n >= 1 and n <= 9): examiners discourage 1 <= n <= 9 because it fails in C# and VB.NET.
• Use the indentation guides in the answer box. Keep identifiers meaningful, one word, and short enough to spell the same way each time.
• Don't use a built-in that the question bans, such as in, index() or max() when told to write your own search.
• Keep it simple. Examiners say over-complicated answers cause logic errors. Prompt text in input() isn't marked unless asked for, so keep it short.
• Completing code: read the lines around each gap, use the identifiers already in the code, and write only what fits the gap.

Brief

The description of what a program must do, often given as a bullet list in the question.

• Code questions on Paper 1 are points-marked. The mark scheme has lettered marks: Mark A, Mark B and so on.
• Design marks are for choosing the right techniques: meaningful names, a loop where one is needed, selection to make a choice. They are given even if the syntax is wrong or the program doesn't work, but there must be a genuine attempt: writing just while and if earns nothing.
• Logic marks are for each piece that works: an input stored and converted, a correct condition, a total worked out, an output in the right place.
• The cap: “Maximum n − 1 marks if any errors in code.” One real error anywhere means you can't get full marks.
• Any correct method gets full marks, even one the mark scheme doesn't show. For example, while True: with break is accepted for REPEAT … UNTIL, and Python's match for selection.
• Examiners mark how the logic flows, not perfect syntax. They are told to ignore minor slips that an editor would flag (table).
• Indentation counts as far as the logic: the first line after if or while is taken as inside it, but later lines that aren't indented are treated as outside.

Design marks for the right techniques, logic marks for working pieces, and a cap if anything is wrong.

total = 0
for night in range(3)
    hours = input("Hours: ")
    total = total + hours
Average = total / 3
if average < 8:
    OUTPUT("Get more sleep")
else:
    print("Well rested")

Design mark

A mark for choosing a suitable technique, such as a loop or selection, given even if the code has errors.

Logic mark

A mark for one part of the program working correctly.

• SQL is the only code on Paper 2: usually 2 to 4 questions, about 8 to 14 marks, near the end of the paper.
• Write SQL keywords, not Python or made-up words: SELECT … FROM … WHERE … ORDER BY; INSERT INTO … VALUES; UPDATE … SET … WHERE; DELETE FROM … WHERE. Answers using GET, SEARCH or PRINT score little.
• A 6-mark query is usually marked A to F: the fields, each table, the join, the condition and the order (diagram). Fields asked for “and no others”, so don't use * or add extra fields.
• Use = to compare, not ==. Put text and dates in quotes.
• Write Table.Field with the table first: Pupil.ClubID, never ClubID.Pupil.
• Copy table and field names exactly, with no spaces added.
• Only your first attempt is marked. Cross out a rough version you don't want marked.
• Gap-fill SQL: write exactly what goes in the gap. If the brackets are already printed, don't add more.

A two-table query, line by line, with the mark each line usually earns.

• Paper 2 has one or two Discuss questions, worth 6 or 9 marks. They are the only levels-marked questions on either paper.
• 9 marks: Level 1 is 1–3, Level 2 is 4–6, Level 3 is 7–9. 6 marks: Level 1 is 1–2, Level 2 is 3–4, Level 3 is 5–6.
• The examiner reads the whole answer, decides which level fits it best, then gives a mark within that level.
• Cover every bullet strand in the question. Answers that leave a strand out are usually held in Level 2.
• Develop each point: say what happens, why, and what it means for the people in the context. A list of statements stays in Level 1, however long it is.
• Stay on the context. If the question is about a company, write about the company; examiners report answers drifting to customers or hackers instead.
• Give both sides where there are two, and finish with a short conclusion that weighs them up.
• Plan for two minutes: write each strand as a heading with two points under it. Then write in paragraphs, one strand at a time. Allow about 11 minutes for a 9-marker.
• No law has to be named. “Data protection laws” is enough; laws from other countries, such as the USA, aren't relevant.

The levels for a 9-mark Discuss question. Development, not length, moves an answer up.

Levels of response

A way of marking where the examiner places the whole answer in a level that best fits it, then chooses a mark within it.

Strand

One of the bullet points in a Discuss question that the answer must cover.