Journal · Product · 7 October 2026 · 14:24 SAST · by Jeremy Rich

Point Power BI at the book

The VB6 version of CreditPlus had hundreds of tables, which suited the engine well enough and suited nobody who had to write a report. At Brait, with the volumes they ran, we worked hard just to make it perform, and a report writer needed to know which of those hundreds of tables held what, and which joins would quietly double a number. The usual cure for that, then and now, is a data warehouse fed overnight. It works, but the figures are always yesterday’s, and a new product doesn’t show up in a report until someone extends the extract, which is never the week you launch it.

We didn’t want either problem again, so LedgerTM now has reporting tables.

What you get

Your platform operator issues a read-only login, and you point Power BI, Tableau, Excel or any SQL tool at it. The connection is encrypted or it doesn’t connect, and the login sees your organisation’s tables and nobody else’s. Every query reads the book as it stands at that moment. Nothing is copied overnight, so nothing is out of date.

The tables follow the Kimball dimensional model, which is the shape every BI tool expects. Fact tables hold what you count and sum: accounts, balances, a balance history after every day-end, exposure, posted transactions and journal lines. Dimension tables describe them: account types, counterparties, operating entities, currencies and rates, lookup values with their hierarchy laid out in levels. A column means the same thing in every table it appears in, so you join on columns with the same name, and every column carries a description that most BI tools show when you hover over it.

The part that follows your configuration

This is the bit I like. In LedgerTM a product is configuration, an account type with its own properties, and now the reporting is configuration too. Every account type gets its own table, named after the type, so a personal loan type becomes account_personal_loan. It carries every column of the general account table plus one column per property declared on the type, so purposeOfLoan becomes purpose_of_loan, with a purpose_of_loan_label beside it for the readable value.

Publish a change to the account type and the table changes with it, straight away. A new property is a new column, and its description becomes the column’s description. So the day a new product goes live is the day it’s in Power BI, and nobody touched an extract. A property marked Sensitive never reaches a reporting table at all, though the property list still shows that it exists.

Yesterday’s book, today

By default every table answers for now. To see the book as it stood at a month-end, run one statement when you connect:

set reporting.as_at = '2026-09-30';

Balances, exposure, rates and each account’s property values then answer for 30 September, until you change it or disconnect. Take a loan advanced 100,000.00 on 30 September, then 5,000.00 on 1 October and 2,500.00 on 2 November, with day-end run on 30 September and 1 October. Set the date to 30 September and the principal is 100,000.00. Set it to 15 October and it’s 105,000.00, from the 1 October day-end. Leave it unset and it’s 107,500.00, the book now. Set it to a date before any day-end had run and you get an empty value, not today’s balance pretending to be history.

The join that doubles a number

Anyone who has built a credit report has met this one. An account has an owner and maybe other parties, a guarantor or a co-borrower, and one person can hold two roles on the same account. Join them without filtering on the role and that person appears twice, and so does the exposure: a guarantor on a 100,000.00 loan who also co-signs it shows 200,000.00. The guide article on reporting tables has the safe join, and the owner sits on the account table and never in the party table, so it can’t be counted twice either.

Each transaction also carries its ISO 20022 bank transaction code, the same code we put on the client’s statement, so a report and a statement classify a fee or a repayment the same way.

And for your agent

The same tables sit behind the SQL tool on the MCP door, arriving in the next release, so an analyst can ask their own agent for “openings per month by branch” and let it write the query, reading the column descriptions as it goes. The BI login keeps the as-at setting; the agent reads the live book.

Reporting tables are licensed per organisation, and your operator issues each login. The connection string is shown once and never stored, so copy it straight into your BI tool.

Written by Jeremy Rich.