Banks
Risk data you reproduce, not reconstruct.
BCBS 239 asks for risk data aggregated on a largely automated basis, to keep errors down. Rebuilding a past figure by hand works against exactly that.
Reconstructed
- Rebuilt by hand from archives and extracts
- Approximate, with an error nobody can size
- Weeks of a program team's time, per question
- Can't show which version of the rules ran, because nobody kept it
- The reported figure and its inputs held in different systems, if at all
Reproduced with Masseef
- Read from what was recorded on the day
- Repeatable: the same inputs and config give the same figure, within a tolerance recorded with each check
- On demand, run by the people who need the answer
- Each mastered value names the rule version that produced it
- Each risk figure stored with the inputs and code version behind it
The test to run on any system, ours included. Value a portfolio as of a past date. Restate one input. Value it again as of the same date. If the number moves, the past figure can't be reproduced.
What Masseef brings
Three things, in order of strength.
Mastering with versioned rules
Where sources disagree, a versioned rule picks the value, and a change to a rule needs a second person's approval. Rules are written in a mainstream programming language, not a vendor rule language.
As-of reconstruction
Reference data, time series, configuration, trades and positions as they were known on any date, including the inputs to pricing.
Coverage, stated plainly
Completeness across the group depends on how much of your estate lands in Masseef. We'll tell you how much, and by when.
Masseef works beside your lineage tools. They map what can flow through your estate, which is what impact analysis needs. Masseef shows what did flow, which is what reproduction needs.
Bring one figure you had to reconstruct.
We'll show you what reproducing it looks like on a running system, and what it would take for your estate.