IB Computer Science IA rubric: the five criteria explained

From May 2027 the IB Computer Science IA is a computational solution: you choose a problem and design and code a software product using computational thinking. It is marked on five criteria, A–E (problem specification, planning, system overview, development, evaluation), out of 30, with the same criteria at SL and HL. It counts for 30% at SL and 20% at HL.

Updated 2026-11-09

IB Computer Science IA at a glance

Total
30 marks (A 4, B 4, C 6, D 12, E 4)
Weighting
30% at SL, 20% at HL
Documentation
Up to 2,000 words
Video
Up to 5 minutes
Files
Documentation PDF, video, appendices PDF with full source code
Applies from
May 2027 onwards
CriterionMarksWhat it assesses
A. Problem specification4Whether you set out the problem scenario with measurable requirements and success criteria, and explain the choice of computational context.
B. Planning4How you decompose the problem into components and whether the plan addresses the success criteria.
C. System overview6Whether the system model, the algorithms for its components and the testing strategy are complete enough for a third party to recreate the product.
D. Development12How functional the product is, the techniques used to implement the algorithms, how you justify implementation choices, and how effective the testing was. The heaviest criterion.
E. Evaluation4How you evaluate the product against the success criteria and justify improvements.
Total30

Based on the IB Computer science guide (first assessment 2027), section “Task descriptors and criteria—SL and HL”. Band descriptions are summarised in our own words; follow the official guide your school provides.

A. Problem specification (4 marks)

Whether you set out the problem scenario with measurable requirements and success criteria, and explain the choice of computational context.

MarksWhat the work shows
0Does not reach the 1–2 descriptors.
1–2Outlines the scenario; states limited success criteria; outlines the computational context.
3–4Describes the scenario in terms of measurable requirements; states appropriate success criteria; explains the choice of computational context.

What the guide clarifies

  • Success criteria are measurable outcomes derived from the requirements and are reused in planning, development and evaluation.
  • The computational context is the specific area of computing used. About 300 words is recommended.

Where marks are lost

  • Success criteria such as user-friendly or fast that cannot be measured.
  • A vague scenario with no real requirement.

Reaching the top band: Write each success criterion as something testable, with a number or clear pass condition, and argue why this computational context fits better than the alternatives.

B. Planning (4 marks)

How you decompose the problem into components and whether the plan addresses the success criteria.

MarksWhat the work shows
0Does not reach the 1–2 descriptors.
1–2A partial decomposition; a plan addressing some success criteria.
3–4A reasonable decomposition; a plan addressing the success criteria.

What the guide clarifies

  • The plan includes a chronology for planning, designing, developing, testing and evaluating; Gantt or Agile charts work well.
  • A reasonable decomposition breaks the problem into the essential components that support the plan, often as a diagram. About 150 words is recommended.

Where marks are lost

  • A timetable with no link to the success criteria.
  • A decomposition no finer than front end and back end.

Reaching the top band: Map every component in the decomposition to specific success criteria, and mark which criteria each stage of the plan delivers.

C. System overview (6 marks)

Whether the system model, the algorithms for its components and the testing strategy are complete enough for a third party to recreate the product.

MarksWhat the work shows
0Does not reach the 1–2 descriptors.
1–2Outlines a limited system model; identifies algorithms; identifies a testing strategy for at least one success criterion.
3–4Constructs an incomplete model; algorithms giving partial functionality; outlines testing for at least three success criteria.
5–6Constructs a complete model; algorithms enabling the product to perform; describes testing aligned with the success criteria.

What the guide clarifies

  • The system model is a set of diagrams of components and connections, including the user interface; algorithms can be natural language, flowcharts or pseudocode.
  • The testing strategy checks correctness and handling of unexpected or incorrect input, often as a table of test data and expected outcomes. About 150 words is recommended.

Where marks are lost

  • Interface screenshots with no components or relationships.
  • Tests that cover only normal input.

Reaching the top band: Use three linked pieces: a component diagram, pseudocode per component and a test table in which each row traces to a success criterion and includes boundary and invalid input.

D. Development (12 marks)

How functional the product is, the techniques used to implement the algorithms, how you justify implementation choices, and how effective the testing was. The heaviest criterion.

MarksWhat the work shows
0Does not reach the 1–3 descriptors.
1–3Very limited functionality; no appropriate techniques; states the choices; states the testing strategy.
4–6Limited functionality; at least one appropriate technique; outlines the choices; states the effectiveness of testing.
7–9Partial functionality; some appropriate techniques; explains the choices; describes the effectiveness of testing.
10–12Fully functional; appropriate techniques; evaluates the choices; justifies the effectiveness of testing.

What the guide clarifies

  • Techniques are how the algorithms are coded, such as loops, data structures, existing libraries and integrated tools. Highlight the code key to efficiency, keep it commented, consistent and readable, and reference it to the full source in the appendix.
  • Testing covers correctness, reliability and efficiency, is described and justified in the documentation and shown in the video. Without the full source code in an appendix, the techniques cannot earn full marks. About 1,000 words is recommended.

Where marks are lost

  • Pasting long code without explaining it.
  • A video that shows no testing.
  • Templates or drag-and-drop builders with no visible algorithms of your own.

Reaching the top band: Pick three or four code excerpts that show your algorithmic thinking; for each, explain why it is written that way, what alternatives you considered and the trade-off, then use test results to show it is reliable and efficient.

E. Evaluation (4 marks)

How you evaluate the product against the success criteria and justify improvements.

MarksWhat the work shows
0Does not reach the 1–2 descriptors.
1–2States how far the success criteria were met; describes improvements.
3–4Evaluates how far the success criteria were met; justifies improvements.

What the guide clarifies

  • The evaluation must be consistent with the problem specification and success criteria in criterion A. About 400 words is recommended.

Where marks are lost

  • A general claim that the goals were mostly met.
  • Improvements unrelated to the success criteria.

Reaching the top band: Go through each success criterion with evidence (test results or user feedback) and a verdict, then propose specific, feasible improvements for the ones not fully met, with reasons.

How the IA is marked

Best fit

For each criterion the teacher chooses the band that best matches the work overall. Not every point in a band has to be met; when the work sits in different bands on different points, the mark reflects the balance.

Upper or lower mark in a band

The upper mark goes to work that largely meets the band and is close to the next; the lower mark to work that meets it to a lesser extent. Only whole marks are given.

Criteria are independent

A high mark on one criterion does not imply high marks on the others. The top band does not require faultless work.

Teacher-marked, IB-moderated

Your teacher marks the work against the published criteria, and the IB moderates a sample, which can move marks up or down.

Computer Science IA word count, video and files

  • Three types of file: the documentation (one PDF in five sections, A–E), a video, and appendices (one PDF with the full source code and any other resources referred to).
  • The documentation is limited to 2,000 words, excluding code excerpts, comments and diagrams, with the total on the first page. Recommended: A 300, B 150, C 150, D 1,000, E 400.
  • The video is up to 5 minutes in a common format such as mp4, avi or wmv, showing full functionality and examples of testing, not the code.
  • Appendices are not used to award marks and need not be read, but without the full source code criterion D's techniques cannot earn full marks.
  • Every solution must be coded; copied code, website templates, sample databases or drag-and-drop builders are not appropriate.

Old and new rubric compared

Up to the November 2026 session the Computer Science IA was marked out of 34: A planning 6, B solution overview 6, C development 12, D functionality and extensibility 4, E evaluation 6, and it required a real client. The new IA is out of 30, drops the client requirement, adds A problem specification, builds testing into C and D, and gives recommended word counts per criterion.

Current criteria (2027 on)MarksOld criteria (to 2026)Marks
A Problem specification4A Planning6
B Planning4B Solution overview6
C System overview6C Development12
D Development12D Functionality and extensibility4
E Evaluation4E Evaluation6

Checklist before you submit

  1. The scenario sets out measurable requirements and every success criterion is testable.
  2. The choice of computational context is explained.
  3. The decomposition and plan both map to the success criteria.
  4. The system model is complete with the interface; each component has an algorithm; the test table covers every success criterion and invalid input.
  5. Code excerpts are commented and referenced to the appendix; implementation choices are explained and evaluated.
  6. The video is within 5 minutes and shows full functionality and testing.
  7. The evaluation goes criterion by criterion and justifies improvements.
  8. Documentation within 2,000 words with the count on page one; the appendix has the full source code.

Get feedback on your IA

Want your Computer Science IA read against these criteria? Contact us: a tutor will go through your plan or draft criterion by criterion and tell you where the marks are going.

Frequently asked questions

How many marks is the IB Computer Science IA out of?

From 2027 it is out of 30: A 4, B 4, C 6, D 12, E 4, with the same criteria at SL and HL. It is 30% at SL and 20% at HL.

What is the word limit?

2,000 words, excluding code excerpts, comments and diagrams. Recommended: A 300, B 150, C 150, D 1,000, E 400.

Do I still need a client?

The new criteria do not require a client; you choose the problem, which can come from the world around you, another field or a current computing issue.

Should the video explain the code?

No. The video, up to 5 minutes, shows functionality and testing; code belongs in the documentation and appendix.