Sign in
Each analyst belongs to one bank. The bank is never chosen by the client — it comes from the session, which is why one bank cannot ask about another's data.
Obligor
Search by any combination of the four. Name matches on the normalised form, so “northwind trading pte ltd” finds “Northwind Trading”; the other three are identifiers and match exactly.
| CIF No. | Obligor Name | RM Code | RM Name | Team Code | Date Created | Access |
|---|
Create Obligor
Search before booking a customer. If the firm already banks them, open that record rather than creating a second one — which is the whole reason this screen searches before it offers New Obligor.
| CIF No. | Obligor Name | RM Code | RM Name | Team Code | Date Created | Access |
|---|
New Obligor
The search that reached this screen found no record for this customer. Book them here.
Re-assign Obligors
Choose whose book to move, show it, select the obligors, then choose who takes them. All of the selection moves, or none of it does.
Obligor Details
| CIF Number | Obligor Name | Team Code | Select All |
|---|
New RM Details
Obligor information
Ownership and audit
Financial spreadsheet
Spreading puts a customer’s accounts into a standard template, so the ratios a rating model reads are computed the same way for every borrower. Find the obligor to spread.
| Obligor name | Customer reference | Access |
|---|
| Financial year | Start | End | Status | Reconciliation | Template version | Report type | Report level | Created by | Last saved | Submitted |
|---|
| Scenario Name | Info | Status | Template Version | Report Level | Created By | Date Created | Last Modified By | Date Last Modified | Last Saved By | Date Last Saved | Submitted By | Date Submitted |
|---|
Scenario Management
Base Year: tick the historical years to average, oldest first. The drivers become "Avg of N Periods".
Spreadsheet Management
| Line | This period | Prior period | Source |
|---|
The line belongs to this customer and this template, so it appears on every statement they spread on it.
| This bank's lines | Under |
|---|
Saved with the sheet, by Save and recalculate.
| Display Name | Short Name | Section |
|---|
| Display Name | Short Name | Formula | Value |
|---|
Portfolio Projection
| Portfolio Name | Status | Filter Type | Date Scheduled | Scheduled By | Queue No. | Batch Run Start Date | Batch Run End Date | Date Last Modified | Last Modified By | Date Created | Created By | Date Cancelled | Cancelled By |
|---|
Portfolio Selection Details
Select a criteria on the left and press > to include it. Several picks in one list select obligors matching any of them; picks across lists select obligors matching all lists.
Caps and Floors Setting
| Rate | Min | Max |
|---|
Financial Projection (Assumption)
One value per assumption line per projected year. Percentages as decimals: 0.05 is 5%.
OG Projection (Risk Level Adjustment)
Tick an item to set its risk level: an absolute level, or upgrade/downgrade by a number of levels. A measured item is decided by the projected figures and cannot be set. "Default, if Item is Null" answers an item the approved scorecard never did.
| Factor | Item | Absolute Risk Level | Relative Risk Level | Default, if Item is Null |
|---|
Batch run
Grade migration
How many obligors moved, and by how much. The scale runs 1 to 26 with 26 the weakest, so a higher grade after the run is a downgrade.
| Grade before | Grade after | Movement | Obligors |
|---|
Every obligor in the run
| Obligor | Before | After | Score | Reason |
|---|
Run the whole book without a scenario
Re-grade every obligor against the active rating model as it stands, with no projection and no notching. The count moves while it runs and it can be stopped without losing what it finished.
Earlier runs
| Started | Status | Obligors | Finished |
|---|
Administration
The six entries the legacy registers, in its order. Each is gated on a functional access type in the original; the ones not built here say so rather than being left out, because a shorter list would misrepresent how much of the system this is.
Rating reference files
The external, implied and sovereign rating files. The version in force is what Retrieve Ratings reads; the trend walks the implied versions back by month. Upload is not built here: the delivered files are seeded.
| Kind | File | Date code | Rows | In force | Previously activated | Activated at |
|---|
Batch Review
Select a model to review its scorecards.
| Regrading Required? | Review Status | Obligor Name | Entity Type | BOG | FOG | Reviewed Date | Reviewed By | Approved Date | Approved By |
|---|
Rating details
| Field | Current | Prior |
|---|
Batch Review Configuration
Rating Details Configuration
Full Regrading Requirement Configuration
Retail early warning
Test every retail customer against the rule set once a month, and give each one an IFRS 9 stage and a single reason. A customer whose salary has fallen appears here before anything is overdue, which is the whole argument for it.
Every retail customer in the run
Including the quiet ones. A list of alerts on its own cannot be told apart from a list of the customers somebody happened to look at.
| Customer | Stage | Reason | Could not be tested |
|---|
On the watch list
Who is on it now, and why. A spell is never deleted: the question a supervisor asks is whether a customer was on the list in the three months before a restructure, because that makes the restructure forbearance. Show the closed spells to answer it.
| Customer | Since | Until | Why |
|---|
Earlier runs
| Started | Status | Customers | Finished |
|---|
Case
What fired
| Rule | What it says | Authority | Action |
|---|
What could not be tested
These rules did not answer "no". Nobody could ask them, because a figure they read is missing. That is a different thing from a clean customer.
| Rule | Missing figure |
|---|
The figures it was measured on
| Figure | Value |
|---|
Rating engine
Scorecards in Credit Predix are not hard-coded forms. The arithmetic that turns
an answer into a score lives in 199,253 formulas stored as XML
case lists in the model database. The delivered RuleEngine.dll is
only an interpreter, so there was no engine to port — a parser and an
evaluator had to be written from the language itself.
The 88 that do not parse are four defects in the stored data — a comma written for a decimal point, a missing third argument — not gaps in the grammar. Each is a correction for Pilot to make in the model.
Scorecard
A scorecard grades one obligor against one version of a rating model. Find the obligor to grade, or open one from the Obligor tab and choose Scorecards.
| Obligor name | Customer reference | Access |
|---|
Scorecards
Obligor Grade Scorecards
| Created | Rating model | Grade | Status | Financial Type | Financial Year | Scenario |
|---|
This customer has never been graded. Create scorecard starts the first one.
Facility Grade Scorecards
| Created | Rating model | Grade | Status | Purpose | Currency | Limit |
|---|
Facility limits
The limits that make up this facility. A trade line with an overdraft sub-limit is one facility and two limits.
| # | Facility type | Currency | Limit | Expected utilisation |
|---|
Collateral
The security behind the facility. Which columns a row may carry is decided by its collateral group, so a row with no market value has a group that does not record one.
| # | Group | Type | % commit | Fair market value | Forced sale value |
|---|
BAR Derivation
BAR Adjustment
| Factor | Current Date | Current EDF (%) | Current Implied Rating | Historical Date | Historical EDF (%) | Historical Implied Rating | Factor Value |
|---|
Financial spreadsheet search
Submitted statements for this customer. Choose one to grade against.
| Period | Financial year | Report type | Level | Currency | Type | Scenario |
|---|
Rating models
A model defines the questions, the weightings, and the arithmetic that turns answers into a grade. Every scorecard records which model version graded it, so a grade stays reproducible after the model changes.
Properties
Formula
Cases are tried in order and the first whose condition is true wins. A case with no condition is the default and belongs last.
Insert
Validation results
Score to grade
The score the formulas produce falls in a band, and the band is the grade. Lower bound inclusive, upper exclusive.
| Grade | Score from | Score to | Probability of default |
|---|
All screens
Every screen the legacy application declares, with its own fields and captions. These are rendered from the delivered source, so a screen here shows what a user of the old system sees. Most have no server behind them yet — a screen that saves is one that has been wired up.
Choose a screen from the list.