Every lost mark
has a reason.
CogniRank AI reads every test your students take — online or on paper — and separates what they don't know from what they got wrong anyway. Then it builds the lessons, the practice and the next paper around the difference.
- Online + OMR
- Four error types
- Topic mastery
- Adaptive papers
- Five portals
One paper, read the way CogniRank reads it. The score was never the finding — the reasons beside it are.
Four ways marks leave a paper. Only one of them is a syllabus problem.
A percentage tells a student to work harder. A reason tells them what to work on. Every response is sorted into one of four types before anything else happens.
Conceptual
everything else that is wrong
The model in the student's head is off. These are batched to the AI reader with the question, the key and the option the student chose, and come back with the misconception named in plain language.
Careless
wrong · easy · under half the student's own median time
They knew it and lost it anyway. Nothing about the syllabus will fix this one — the fix is pace and checking, and it needs to be counted separately to be trainable.
Time pressure
wrong or blank · final fifth of the paper, or over 2.5× median time
The clock beat the concept. Read next to the same topic's earlier questions, it tells you whether the topic is weak or the strategy is.
Unattempted
no answer recorded
Avoidance is a signal, not an absence. A blank counts against mastery, because skipping a topic is not the same as knowing it.
The rules run first because they are free, instant and reproducible — a student can be shown exactly why a mistake was labelled careless. The AI is spent only on the questions the rules cannot explain.
One loop, running for every student, every week.
Nothing here is a report that gets filed. Each step is the input to the next, and the last step starts the first again.
- 01
Test
Students sit the paper in a locked browser window that records time per question — or the institute scans the OMR sheets from an offline test. Both land in the same pipeline.
- 02
Diagnose
Every response is graded on the key, then classified by reason. Errors roll up per topic, and the paper comes back with the misconceptions written out.
- 03
Profile
Behaviour across recent papers places each student in a working style — concept builder, careless sprinter, slow perfectionist or balanced performer. A label only changes after two evaluations agree, so it never flaps.
- 04
Teach
The weakest topics get notes, slides and narrated lessons, written for that working style and opened with the student's own mistakes on that topic.
- 05
Adapt
The next paper over-samples weak topics, leaves out recently seen questions and ramps difficulty to the profile. A topic leaves the weak list only after two strong showings in a row.
→ back to 01
What an institute actually gets.
Assessment, analysis, content and oversight in one system, so the diagnosis and the fix never live in different tools.
Online test player
Fullscreen-gated, autosaving, with a question palette and per-question timing. Tab switches and window blurs are recorded, and the server — not the browser — submits the paper at the strike limit.
OMR sheet ingestion
Upload a stack of scanned answer sheets. Roll numbers, paper numbers and bubbles are decoded automatically; anything the decoder is unsure of is flagged for a human before a single mark is committed.
Topic mastery map
A subject-to-topic heatmap per student, updated after every paper on a weighted average, so recent work moves the number without erasing the history behind it.
Cognitive profile
Accuracy, careless share, unattempted share and pace against the cohort, rolled over recent attempts into a working style with advice attached.
Adaptive papers
Generated from a stored blueprint — weak topics weighted heaviest, difficulty ramped to the profile, recently seen questions excluded — so the next test measures learning rather than memory of the key.
AI doubt tutor
A per-question chat that already knows the question, the worked solution and what the student actually answered, so the student never has to explain their own mistake first.
Mistake book
Every wrong answer the student has ever given, kept with its reason, its topic and its solution, filterable by subject and error type.
Generated lessons
Notes, slide decks and narrated audio produced per topic and per working style, then reused across every student who needs them — so the cost lands once, not once per learner.
Question bank tools
Bulk import, image-based questions, multi-correct support, and AI-assisted extraction that turns a photographed question paper into bank entries for review.
Institution analytics
Enrolment, attempt volume, accuracy and subject strength rolled up by branch and by faculty member, with AI usage and cost visible to whoever signs the invoice.
One platform, five points of view.
Everyone sees the same truth, cut to what they are responsible for. Access is enforced on the server, not hidden in the interface.
- Students
- Their dashboard, papers to sit, mistake book, assigned lessons and a doubt tutor on every question.
- Teachers
- Only the students assigned to them: per-student reports, test authoring, topic material to curate, and a public profile in the faculty directory.
- Branch admins
- One branch end to end — students, faculty, question bank, papers, OMR batches, parent access, and the AI prompt and cost controls behind it all.
- Directors
- Every branch side by side: enrolment, performance, faculty strength and subject-level results across the chain.
- Parents
- A read-only view of their own child's progress and nothing else, printable as a report to keep.
SUMZ Academy & IT Solutions
An EdTech and software company building intelligent learning platforms, industry-ready training programmes and digital transformation work for institutions. CogniRank AI is our flagship product.
Vision
To be a global leader in AI-powered education technology, building systems that change how learning is measured and improved.
Mission
To put smarter learning, honest insight and measurable progress in the hands of every student, teacher and institution we work with.
How we build
- Education first, then the technology that serves it.
- Deterministic maths for every number a student is judged on.
- AI reserved for the language work it is genuinely better at.
- Built for a chain of branches from the first table onwards.
- Training and consulting alongside the platform.
- Support from the people who wrote it.
CogniRank is the flagship product of SUMZ Academy & IT Solutions — an EdTech and software company building learning platforms, industry-ready training and digital transformation work for institutions.
Bring one paper your students have already sat.
We will run it through CogniRank and show you the report your faculty would have had the next morning — the topics that broke, the marks lost to pace rather than knowledge, and the paper each student should sit next.