Back to Home
PROPRIETARY GRADING TECHNOLOGY

PrepFocus's Intelligent
Hybrid Grading Engine

How we grade calculations, bookkeeping entries, and essays with up to **98% accuracy** compared to senior human examiners—all in absolute compliance with official ICAN marking guides.

The Technology Edge: Overcoming the Limitations of Traditional AI

Traditional e-learning platforms face a massive technical wall when preparing students for chartered professional examinations. Multiple-choice tests are too rigid to simulate the actual exam, while standard Large Language Models (LLMs) suffer from calculation errors, drift, and halluncinated feedback on essay answers.

PrepFocus solves this with a **proprietary, multi-stage Hybrid Grading Engine**. By separating pure numerical logic, structural bookkeeping checks, and semantic essay rubrics, we provide instant, extremely fair, and highly defensible scoring aligned with real ICAN grading guides.

99%+
Math Accuracy
95%+
Ledger Matching
90%+
Theory Alignment

How Our Hybrid Engine Operates

Three custom modules dedicated to matching the three components of professional exams.

1. Zero-AI Mathematical Precision

LLMs are notoriously unreliable with multi-step arithmetic. PrepFocus does not allow generative AI to handle calculations.

Our Solution: A custom, deterministic rules engine parses your final answers and sub-working values, comparing them directly to the database marking scheme with a strict ±1% tolerance bracket. This dynamically absorbs rounding variances just like human examiners.

99% Numeric Reliability

2. Fuzzy Semantic Ledger Matching

Candidates rarely write ledger account titles identically—writing "Bank" instead of "Cash at Bank", or "Capital" instead of "Share Capital".

Our Solution: A proprietary NLP string-similarity algorithm calculates conceptual matches. It checks that accounts are posted to the correct side (Debit/Credit), credits appropriate fractional marks, and handles ledger discrepancies in real-time.

Fuzzy String Matching

3. Grounded 3-Stage Theory Assessment

Unconstrained AI models hallucinate or grade inconsistently. PrepFocus uses a strict 3-Stage Pipeline to secure essay integrity.

Our Solution: We use local Jaccard keyword overlaps, conceptual cosine similarity, and feed the **official ICAN Marking Guidelines** directly to a rubric-constrained LLM set to a rigid, zero-temperature context. The AI is limited to acting solely as an auditor of the guide.

Hallucination-free scoring
OFFICIAL SYLLABUS COMPLIANCE

Grounded in ICAN Marking Guides

When you practice subjective essays or case scenarios in the PrepFocus Arena, your final scorecard provides a line-by-line comparison showing exactly where you scored marks under official examiner rubrics. Real tutor-reviewed standards are built directly into the system.

Technology & Grading FAQ

Answers to common candidate and instructor queries regarding our automated grading.

How does PrepFocus handle spelling mistakes or synonyms in written answers?

Our bookkeeping and essay modules do not rely on exact letter matching. Using our localized Semantic Similarity indexes, the grading engine automatically recognizes accounting synonyms (e.g. recognizing that 'Cost of Sales' is conceptually identical to 'Cost of Goods Sold'). Minor spelling mistakes are handled intelligently without penalizing your core concepts.

Does the AI make up its own grading standards?

No. The system is structurally locked down. We load the official ICAN Marking Guidelines for each particular question directly into its processing environment. The algorithm is restricted to acting only as a matching auditor, preventing the AI from creating arbitrary scores, drifting, or hallucinating results.

How are calculation rounding differences scored?

For all calculations (such as Capital Allowances, tax schedules, or inventory costings), our zero-AI mathematical checker utilizes a ±1% tolerance bracket. If you rounded a figure slightly differently in an early step, the engine will still accept your final answer as correct if it sits within the examiner tolerance range.

Can I receive partial marks for my workings?

Yes. In the PrepFocus Practice Arena, ledgers and multi-step calculations are graded step-by-step. If your final balance sheet doesn't balance, but your individual ledger entries are correct, the Hybrid Grading Engine automatically breaks down the question and awards correct fractional marks for all compliant steps.

Ready to try the simulation?

Experience the Hybrid Grading Engine in action with real diagnostic practice questions today.