F202 Coursework: The Set Assignment Guide
How F202 is assessed
F202 is the non-examined assessment unit of the qualification, so there is no written paper. You complete an OCR-set assignment, your teacher marks it against OCR's marking criteria, and OCR moderates a sample of your centre's work to check that the marking matches the national standard.
| Feature | Detail |
|---|---|
| Unit | F202 Spreadsheet data modelling |
| Qualifications | Mandatory in both the Certificate (H019) and the Extended Certificate (H119) |
| Size | 75 guided learning hours, 15 of them for completing the set assignment |
| Assessment | OCR-set assignment, centre-assessed and OCR-moderated |
| The brief | A scenario with a client, a problem and tasks, published by OCR and given to you by your teacher |
Set assignments are replaced over time, and each is valid for a stated period. Your teacher will give you the version your centre is using, with its tasks, any templates OCR supplies, and your deadlines. Everything on these pages is general teaching: the examples use an invented brief, so they show the quality of each kind of evidence without answering yours.
What to produce at each stage
| Stage | The evidence you produce | What makes it strong | The common gap |
|---|---|---|---|
| Understand the brief | Notes on the client, the users, the problem and the purpose | In your own words, with the needs the brief only implies drawn out | Copying the brief back out |
| Plan (TA2) | Numbered requirements, sheet designs, a data dictionary, a processing plan, a test plan | Every design choice linked to a requirement; expected results worked out by hand | Designs with no annotation; tests written after the build |
| Create (TA3) | The working model, with evidence of how it was built | Named cells, validation, protection, lookups and what-if tools, each one justified | One technique used over and over; numbers typed into formulas |
| Deliver (TA4) | The completed test log with fixes and retests, the outputs presented to the client, the user guide | Honest failures with fixes; a recommendation that answers the question | Every test passing first time; no recommendation at all |
| Evaluate (TA5) | A verdict for every requirement, feedback, limitations, improvements and reflection | Evidence behind every verdict; specific improvements | "It went well", with nothing to support it |
Keep dated versions of your files as you go, model-v01.xlsx, model-v02.xlsx, and screenshots of work in progress. They show how the model developed, and they protect you if a file is lost or corrupted.
Your own work: the rules
Coursework rules are set by the JCQ, the body the exam boards act through, and by OCR. Your teacher will explain how they apply in your centre. The main points:
- Your work must be your own. You sign a declaration saying so, and your teacher has to be able to confirm it.
- Your teacher can teach you and guide you, explain what the marking criteria mean and give general feedback. They cannot tell you exactly what to change, correct your work for you, or give you a model answer to the assignment.
- Do not share your work with other students, or use anybody else's.
- Artificial intelligence. Copying or paraphrasing text, formulas or designs from an AI tool and submitting them as your own is malpractice. If your centre allows an AI tool for a particular purpose, you must acknowledge it: name the tool, say what you asked and when, and show how you used the response. The JCQ's AI guidance sets this out in full.
- Reference every source you use, whether data, images, ideas or quotations, both in the work and in a reference list.
Referencing in the Harvard style
| Source | The pattern |
|---|---|
| A web page | Organisation or author (year) Title of the page. Available at: URL (Accessed: date). |
| A data set | Publisher (year) Title of the data set. Available at: URL (Accessed: date). |
| A book | Author, A. (year) Title. Place: Publisher. |
Work found not to be your own can be given no marks, and in serious cases a student can be disqualified from the qualification. If you are unsure whether something is allowed, ask your teacher before you do it.
Worked extracts from an invented brief
These extracts show each kind of evidence at its best. They come from the Harbour Lights festival model used throughout this unit. Do not copy their wording: your brief, your client and your decisions are different, and it is yours that are assessed.
R3. The model must compare pessimistic, expected and optimistic scenarios side by side, so the committee can see the risk before setting the price.
Every assumption sits on the Inputs sheet in a shaded, unlocked cell with its own name (TicketPrice, DemandAtZero, Capacity). Volunteers can then only type in the right places (R5), and every formula reads as a sentence, which makes the formulas easier to test (R1) and the technical documentation shorter.
=MIN(Capacity, MAX(0, DemandAtZero-LostPerPound*TicketPrice)). MIN stops the model selling more tickets than the licensed capacity. MAX stops a high price producing a negative number of tickets, which test 5 found in the first version.
Test 5, boundary data, a price of £60.00. Expected: 0 tickets and a loss of £4,000. Actual: −120 tickets and a loss of £10,600. Cause: the formula had no lower limit. Fix: added MAX(0, …). Retest: 0 tickets and a loss of £4,000. Passed.
R6 is partly met. Two of three volunteers could not find Goal Seek, which is three menus deep. A dashboard button running a Goal Seek macro would take the menus out of the task, and I would retest with the same volunteers to check it worked.
Before you hand in
- Every requirement is numbered, testable, and traced through the design, the tests and the evaluation
- Every sheet has an annotated design, and every input is in the data dictionary
- No formula contains a typed-in assumption
- Every input is validated, and every formula cell is protected
- The test log includes boundary, invalid and erroneous data, with failures, fixes and retests
- The outputs answer the client's question, with a recommendation and its uncertainty
- Somebody who had never seen the model has followed the user guide
- The evaluation gives a verdict and evidence for every requirement, and specific improvements
- Every source is referenced, and any permitted use of AI is acknowledged
- Files are saved as numbered versions, with the final version clearly named
Sources and further reading
- OCR: IT Data Analytics (H019, H119): the qualification page, with the specification and assessment information your teacher works from.
- The H019 and H119 specification (PDF): F202's content and how it is assessed.
- JCQ: instructions for conducting non-examination assessments: the rules every centre follows for coursework.
- JCQ: artificial intelligence: what counts as misuse of AI, and how to acknowledge permitted use.
- Office for National Statistics and data.gov.uk: reliable data that is free to reuse, if your brief allows you to bring in data of your own.
Key terms you must be able to define
- Non-examined assessment
- Assessment completed as coursework rather than as a timed written paper.
- OCR-set assignment
- A brief with a scenario and tasks written by OCR and completed by students in centres.
- Centre-assessed
- Marked first by the student's own teachers, against the board's criteria.
- Moderation
- The board checking a sample of a centre's marking against the national standard.
- Malpractice
- Breaking the assessment rules, such as submitting work that is not your own.
- Harvard referencing
- A system that cites sources by author and year in the text, with full details in a reference list.
Flashcards
Reach the top grade on this topic
- Read the brief closely, and write the client's needs in your own words
- Keep every stage's evidence, including plans that later changed
- Trace every requirement from the plan, through the tests, to the evaluation
- Reference every source, and acknowledge any permitted use of AI
- Save numbered versions, so your process is visible and your work is safe
Mistakes that cost marks here
- Starting to build before the planning is done.
- Throwing away the first plan when it changes, instead of recording why it changed.
- Copying wording from examples, including the ones on this site.
- Using an AI tool to write or build any part of the work you submit.
- Forgetting to reference data sources and images.
- Keeping one file that is overwritten all the way through.
The phrase 'garbage in, garbage out' dates from the early days of computing, and it applies in full to coursework: a beautifully built model resting on an unchecked assumption is still wrong.
Knowledge check
Check you could explain this
- LOCR writes the assignment and its tasks
- LThe centre's teachers mark the work against OCR's criteria
- LOCR checks a sample of the marking against the national standard
An OCR-set assignment is a brief written by OCR, with a scenario, a client and tasks, which every centre uses rather than writing its own. Centre-assessed means the work is marked first by the student's own teachers, using OCR's marking criteria. OCR-moderated means OCR then checks a sample of each centre's marked work, and adjusts the marks if the centre has been more generous or more severe than the national standard, so that a mark means the same thing in every school.
- LIt shows which work is yours and which is from others
- LNot doing so can be plagiarism or malpractice
- LIt lets the reader check the source, or shows the data is reliable
Referencing makes clear which parts of the work are your own and which come from other people, so you get credit for your own thinking and others get credit for theirs. Using someone else's words, images or data without saying so is plagiarism, which is malpractice and can lose you the marks or even the qualification. A reference also lets the reader find and check the source, which matters in a model whose assumptions rest on data: a demand figure taken from a named, reliable publisher is far more convincing than one with no source at all.
- LTests planned in advance from the requirements, with hand-calculated expected results
- LA range of test data: normal, boundary, invalid and erroneous
- LScreenshots or formula views showing the actual results, annotated
- LFailures recorded with the cause, the fix and the retest
Convincing test evidence starts with a plan written before the model was built, from the requirements, with every expected result worked out by hand so there is an independent answer to compare with. It uses a range of test data, normal, boundary, invalid and erroneous, so it shows the model rejecting what it should reject as well as calculating what it should calculate. Each actual result is backed by a screenshot or the formula view, annotated to say what it proves. And it records the failures honestly, with the cause, the fix and the retest, because a log where everything passed first time looks like a log written to pass.