How to Build a CBT Exam Platform for a School

Planning CBT software development Nigeria schools can use starts with the exam, not the software. Decide how many students sit at once, whether the result ranks or only practises, and what happens when a computer dies at minute 40. A basic build then needs a clean question bank, a trustworthy timer, simple scoring and an offline plan.

Key points

• A CBT exam moves cost from printing and marking to computers, power and planning.
• The question bank is the real asset. Clean it before you build screens.
• The timer and the offline plan decide whether students and teachers trust the system.
• Software helps with exam integrity. It does not replace people in the room.
• My exam platforms start from $8,000, quoted in writing.

Who needs a CBT platform?

Schools that set many objective questions, have a computer room or enough devices, and want fast results. It suits a JAMB practice centre or a termly exam. It suits long written answers badly, so those parts usually stay on paper.

In June 2016, Vanguard reported that the University of Benin's ICT Unit had developed its own CBT software, used for the university's exams including Post UTME and PUDE. The vice chancellor called UNIBEN the first Nigerian university to develop and use such software. I found no independent source for "first", so read it as the university's claim. What I take from it is simpler: a team inside an institution can build this, and most schools have no such team.

What should I decide before writing any code?

Answer five questions first. They change the build more than any feature list does.

1. How many students sit at once, and on how many devices?
2. Is the exam for ranking, passing or practice?
3. Who writes the questions, and who may see them before the exam?
4. What happens if a student's computer fails at minute 40?
5. Who explains the result to a parent?

A practice platform can allow retries and show explanations. A termly exam needs one attempt, a log of what happened and a way to recover a student after a crash. The screens may look the same. The products are different.

What features does a school CBT platform need?

The core is a question bank, randomisation, a timer, navigation and scoring.

Question bank. A record holds the question, options, correct answer, subject, topic, an editable difficulty tag and an explanation. Etest JAMB CBT has 5,553 questions across 11 subjects, and reaching that number meant throwing questions away. The build rejected duplicate options, stub questions, empty options, bad answer keys, misfiled items and questions that depended on a diagram I couldn't show. A big bank full of broken items is worse than a small clean one, because students doubt everything after the first broken question. For a school, build an import check before the exam screen: no empty options, one correct answer, a subject assigned.

Randomisation. Shuffle question and option order to make copying harder, and store the order each student saw. Without it, marking disputes can't be settled. Let staff pin questions such as "all of the above", which break when shuffled.

Timer. Students complain about it first. Etest JAMB CBT offers timed sessions of 60, 90, 120, 150 and 180 minutes, plus a study mode with no clock. For an exam that counts, my advice is to check the clock against the server's time, so a page refresh can't pause it. At zero, submit whatever the student has answered. Never discard it.

Navigation and flags. A grid showing answered, unanswered and flagged questions is what makes a CBT feel like the real thing to a nervous candidate. Build it early.

Scoring. Write the rules down: marks per question, whether blanks count against a student, how a total converts. Etest JAMB CBT shows a score, accuracy, a subject breakdown and explanations. A school also wants results by class, subject and question. The by question view shows items most of the class got wrong, which points to a teaching gap or a bad question.

How do you run CBT exams with unreliable power and internet?

Plan for failure, because power cuts, dropped routers and finished data plans are normal in many schools. There are two honest options.

A local server inside the school lets computers connect over the school's own network, so the exam doesn't need the internet. Results sync to the cloud afterward. It takes more setup and someone has to look after the server.

A hosted platform is easier to run and update, and it fails when the connection does. I'd only recommend it with a backup, such as a mobile hotspot and a way to resume an attempt.

Here is the limit of my own example. Etest JAMB CBT is a practice app for individual students, and its 209 study guides work offline. It is not a school exam system, and it has no class lists or teacher accounts. What carries over is the work on question quality, timers and navigation. The school administration side is separate work that I scope per client, after seeing your computer room.

Can a CBT platform stop cheating?

No, and be wary of anyone who says it can. Software helps in specific ways: shuffled questions and options, one attempt per student, a log of start, submit and disconnects, questions released only when the exam begins, and a record of who logged in on which machine.

It can't replace an invigilator. Use the log to settle disputes, not as a guarantee. Keep student data to name, class and results.

What does a CBT exam platform cost?

My exam and CBT platforms start from $8,000, and every project gets a written scope and quote. What moves the price:

Cost driver	Why it matters
Exam types and question formats	Single answer is simple. Multiple answer and fill in take more work.
Admin tools	Class lists, bulk student import, teacher accounts, result exports
Offline requirement	A local server setup adds the most work
Branding and reports	Certificates, parent facing results
Questions to clean and import	Often longer than the exam screen itself
Support and training	Needed after launch

After launch there are running costs: hosting, backups, someone to load questions each term and a person at the school who owns the system. They are small next to the build, but they never stop.

How should a school roll it out?

1. Pilot with one class and one subject, using low stakes questions.
2. Run the same exam twice in the lab, once with a planned failure such as unplugging a computer, so you see what recovery looks like.
3. Train invigilators for an hour on what to do when a screen freezes.
4. Keep paper copies ready for the first live use.
5. Go live for one term, collect complaints, fix them, then expand.

Give students who have never sat an exam on a computer a practice session first. Handling the mouse alone costs some students marks.

Questions to ask any CBT vendor

• Can I see an exam recovered after a computer fails mid test?
• Does the timer follow server time?
• How do you store which questions each student saw?
• What works with no internet, and what doesn't?
• Who owns the code and the question bank when I've paid?
• What are the running costs after launch, in writing?

Frequently asked questions

How much does it cost to build CBT software for a school?

My exam and CBT platforms start from $8,000, scoped and quoted in writing. The final figure depends on exam types, admin features and offline needs. I don't quote averages, because they hide the features that matter to your school, such as a local server or bulk student import.

Can a CBT platform work without internet?

Yes, if it runs on a local server inside the school, or keeps content on the device and syncs results later. A fully hosted platform fails when the connection does. I'd choose the setup after seeing your computer room and your power situation.

Can we use our own past questions?

Yes, and you should. The work is cleaning and tagging them: one correct answer, subject and topic filled in, no broken images. Expect the import step to take longer than building the exam screen, since every bad item shows up in front of students.

Do I own the platform after it's built?

Yes. You own the code and assets once the project is paid for. Hosting and loading new questions stay with you, or with me under a separate support arrangement, written down before we start. Ask any vendor the same question first.

Can you guarantee students can't cheat?

No. Software can shuffle questions, limit attempts and keep logs of who did what and when. It can't replace people watching the room, and anyone who promises a cheat proof platform is selling you something. Plan for invigilators, and use the logs to settle disputes afterward.

If you're planning one

Tell me about the exam: how many students, what devices, how often, and what went wrong last time. I'll reply myself with how I'd build it, what I'd leave for later, and the cost in writing. You can read more on my CBT exam platform development page, see the practice app I built in Etest JAMB CBT, or read about starting an online school in Nigeria.

Start a project

Written by John Osakwe, founder of Etest Tech Hub in Lagos. Last reviewed 15 October 2026.

How I wrote this. I wrote this from my own work on Etest JAMB CBT, and checked the app figures against the app's data files and the UNIBEN story against the Vanguard article on 15 October 2026. The design advice is my opinion from building exam software. I used an AI assistant to tidy the draft; I chose what goes in and I stand behind it.

How I wrote this

I wrote this from my own work on Etest JAMB CBT, checked every app figure against the app's data files on 15 October 2026, and checked the UNIBEN story against the Vanguard article. The design advice is my opinion from building exam software, not a studied benchmark. I used an AI assistant to tidy a draft; I chose what goes in, and I stand behind it.