Koshish runs the whole assessment loop for a coaching institute: build a mock from the verified question bank, assign it to one batch or a set of students in a single action, and have every paper scored automatically on submission. Results show per-student scores, a class distribution and batch-on-batch comparison, with CSV export.
Last reviewed 2026-09-26
Explore
The full loop, with the real implementation behind each step.
Create a mock: pick a title, exam, duration, total marks and passing score
Choose questions from the verified bank, in the order students will see them
Assign to a batch, to named students, or both
Students sit it on their phones; scoring happens on submission
See per-student results, a score distribution and batch comparison
Explore
Stated precisely, because a mark scheme you cannot predict is worse than none.
Each question carries its own marks, so total marks follow the paper
Negative marking applies where a question carries it
A paper never scores below zero
Accuracy is correct answers over questions on the paper
Submission is idempotent — a retried submit cannot double-count a score
Explore
Worth knowing before you build a process around it.
No institution-authored questions — the bank is Koshish's
No multiple sections with per-section timing or marks
No descriptive or typed-answer papers
No server-authoritative timing or webcam proctoring
FAQ
Yes, with a defined scope. You can build a mock from Koshish's verified question bank, assign it to one batch or a set of students, and each student's paper is scored automatically on submission — including negative marking and an accuracy percentage. What we do not yet support is institution-authored questions, multiple sections with per-section timing, or descriptive papers.
No, and we would rather say so up front. Koshish covers learning analytics and the assessment loop for competitive-exam batches: who is weak in which topic, how batches compare, and running a test end to end. It does not do fees, attendance, admissions or payroll, so most institutes run it alongside their existing software rather than instead of it.
From your students' real attempts on their phones. Koshish's BKT and IRT models estimate each student's mastery per topic; the cohort view rolls those attempts up into an accuracy rate you can read at a glance. Nothing is typed in by hand and nothing is estimated from a headcount.
Yes. Performance, progress and comparison reports all export to CSV at any time, and you can request erasure of your workspace's data through the grievance officer. We are not relying on data lock-in to keep you.
Once your roster is imported and students have been practising, the first cohort report reflects real attempts. We agree a realistic timeline with you at the outset rather than quoting a number we cannot stand behind.
Tell us your batch sizes and exam mix and we will scope the setup honestly, including what it will not cover.
Talk to our team