MASTIX
Back to News & Resources

Article

Next-Generation ALM

How the computational foundation of ALM is changing, and what an architecture built for interactive, auditable analysis looks like in practice.

Published

March 21, 2024

Read Time

12 min read

Author

MASTIX

Industry InsightsALMTreasury
Next-Generation ALM

Next-Generation ALM

Next-generation ALM is not just a newer interface on top of the same architecture. It is a change in how balance-sheet analytics are calculated, explained, and used in day-to-day decisions.

The case for a different ALM architecture

Traditional ALM platforms were built for reporting cycles, not for interactive analysis. They can produce the required metrics, but they often do so through separate models, overnight batches, and limited drill-down.

Where legacy ALM still breaks down

Legacy ALM systems face several recurring challenges:

  • Siloed Calculations: Risk and accounting handled separately, leading to inconsistencies
  • Overnight Processing: Critical metrics available only after lengthy batch processes
  • Limited Flexibility: Rigid interfaces that constrain analytical capabilities
  • Poor Transparency: Black-box calculations with limited insight into methodologies

What next-generation ALM means

Next-generation ALM starts from a different set of requirements:

1. One Analytical Foundation

  • Consistent Calculations: EVE, NII, sensitivities, and attribution come from the same analytical foundation.
  • Shared Cash Flows: Metrics are derived from the same projected cash flows instead of separate model chains.
  • Less Reconciliation: Teams spend less time explaining differences between systems.

2. Speed That Changes Workflow

  • Interactive Analysis: Lighter scenarios can be explored in seconds while teams are still discussing alternatives.
  • Full Balance-Sheet Runs: Broader balance-sheet scenarios can complete in minutes instead of overnight batches.
  • Faster Iteration: The point is not just runtime. It is the ability to test, compare, and explain decisions in the same working session.

3. Sensitivities In One Run

  • Single-Pass Sensitivities: Compute sensitivities as part of the original calculation instead of rerunning once per factor.
  • Full Coverage Across Modelled Instruments: Keep the same method across the portfolio instead of switching to approximations when scale increases.
  • Direct Link To Outcomes: Fewer reruns means faster analysis, fewer delays, and fewer inconsistencies.

4. Explainability Built In

  • Full Traceability: Follow a reported result back to contracts, market data, and assumptions.
  • Built-In Attribution: Explain changes through rates, new business, runoff, and model effects without reconstructing the answer afterward.
  • Governance Fit: Results are easier to defend in ALCO, audit, and model governance discussions.

What makes it possible

Adjoint Algorithmic Differentiation (AAD)

Adjoint Algorithmic Differentiation (AAD) is one of the key technologies behind this shift:

  • One Analytical Pass: Compute sensitivities in the valuation itself instead of through bump-and-revalue loops.
  • Exact Derivatives: Avoid numerical approximation errors from finite differences.
  • Foundation For Attribution: When sensitivities come from the same calculation, explainability is connected to the engine rather than added later.

Shared Projection Framework

  • Contract-Level Modelling: Start from individual instruments and aggregate upward.
  • Shared Projections: Use one projection foundation across valuation, scenarios, and reporting.
  • Consistent Metrics: Keep EVE, NII, and related outputs aligned because they come from the same engine.

Workflow Integration

  • Excel And Python Access: Meet teams where they already work.
  • API Support: Fit into existing reporting and governance workflows.
  • Operational Use: Make the analytics usable during analysis, not only after a reporting cycle completes.

What changes in practice

The practical benefits of next-generation ALM are operational, not just technical.

Shorter Decision Cycles

  • Faster Iteration: Compare scenarios during the meeting instead of requesting another overnight run.
  • Less Waiting: Move from batch-dependent analysis to working-session analysis.
  • More Time For Judgment: Spend less time collecting numbers and more time interpreting them.

Less reconciliation

  • One Number Across Teams: Treasury, ALM, and governance functions work from the same foundation.
  • Fewer Methodology Disputes: Differences between systems no longer dominate the conversation.
  • Cleaner Reporting Flow: Metrics and explanations arrive together.

Stronger Governance

  • Traceable Results: Reported outputs can be followed back to their source data and assumptions.
  • Repeatable Analysis: The same inputs reproduce the same result.
  • Better Questions, Better Answers: Model validation, audit, and management reviews can focus on the content of the result instead of the mechanics of reconciliation.

What to evaluate

Define The Workload, Not Just The Feature List

  • Scenario Scope: Be clear about what needs to run interactively and what can remain a broader scheduled run.
  • Metric Scope: Confirm which metrics need to come from the same engine.
  • Data Inputs: Decide how contract data, market data, and assumptions enter the workflow.

Design For Operating Model Change

  • Committee Use: The biggest gain often comes when teams can answer questions during the discussion.
  • Governance Use: Traceability matters as much as runtime.
  • Team Adoption: Treasury, ALM, and reporting teams need a shared understanding of what the new architecture changes.

Check whether the analytical foundation is real

  • One Engine Or Many: Check whether the platform truly uses one foundation across metrics.
  • Sensitivities And Attribution: Verify that these are structural properties, not post-processing add-ons.
  • Workflow Fit: Make sure the outputs can be used in the actual tools and review processes your teams rely on.

Where MASTIX fits

MASTIX is built around this model:

  • One Engine Across Outputs: Valuation, projection, sensitivities, and attribution come from the same analytical foundation.
  • Adjoint Algorithmic Differentiation (AAD): Exact sensitivities are computed in the valuation itself.
  • Flexible Interfaces: Teams can work through Excel, Python, and APIs.
  • Traceable Results: Outputs remain connected to the same calculation engine and the drivers behind them.

Conclusion

Next-generation ALM is not mainly about adding more dashboards or more compute. It is about replacing fragmented, batch-oriented workflows with an analytical foundation that can calculate, explain, and support decisions in the same cycle.

That is the shift MASTIX is built for.