Basel III in SAC: modeling CET1, LCR and NSFR with the right aggregation

Basel III is the global rulebook that decides how much capital and liquidity a bank must hold. Three ratios sit at its core — the CET1 capital ratio, the LCR and the NSFR — and in a SAP Analytics Cloud model they fail in exactly the same way Solvency II ratios do: the moment a denominator is summed across time or a ratio is averaged across entities, every number above it is wrong. This guide covers what Basel III requires, the thresholds that matter in 2026, and how to structure the ratios in SAC so they always hold.

What Basel III requires

Basel III is set by the Basel Committee on Banking Supervision (BCBS), hosted at the Bank for International Settlements, and transposed into law by each jurisdiction (CRR/CRD in the EU, the PRA rulebook in the UK, the federal banking agencies in the US). Like Solvency II it rests on three pillars: Pillar 1 sets the minimum capital, liquidity and leverage requirements; Pillar 2 covers supervisory review and a bank's own ICAAP; Pillar 3 governs disclosure. A SAC model is a Pillar 1 and Pillar 3 tool — it computes the ratios and feeds the regulatory reporting.

The ratios you must model

Capital ratios. Capital is expressed as a percentage of risk-weighted assets (RWA). The minimums are CET1 ≥ 4.5%, Tier 1 ≥ 6% and Total capital ≥ 8% of RWA. On top sits the capital conservation buffer of 2.5% (so CET1 is effectively ≥ 7%), a countercyclical buffer of 0–2.5% set by national regulators, and a surcharge for global systemically important banks (G-SIBs). Common Equity Tier 1 — chiefly common shares and retained earnings — is the highest-quality capital and the number markets watch most.

Leverage ratio. A non-risk-based backstop: Tier 1 capital ÷ total exposure ≥ 3%. It catches the case where risk weights flatter a balance sheet that is simply too large.

Liquidity Coverage Ratio (LCR). High-quality liquid assets ÷ net cash outflows over a 30-day stress ≥ 100%. It asks: could the bank survive a month-long run on its own liquid buffer?

Net Stable Funding Ratio (NSFR). Available stable funding ÷ required stable funding ≥ 100%, over a one-year horizon. It pushes banks to fund long-term assets with durable liabilities rather than flighty short-term money.

RWA and the output floor (Basel III finalisation)

Everything on the capital side hangs on RWA, and RWA is where Basel keeps evolving. The Basel III finalisation package — agreed by the BCBS in 2017 and known as CRR3 in the EU, Basel 3.1 in the UK and the Basel III Endgame in the US — tightens how RWA is measured. Its signature element is the 72.5% output floor: a bank using internal models cannot report total RWA below 72.5% of what the standardised approaches would produce. It also revises the standardised approaches for credit, market (FRTB) and operational risk.

As of 2026 the timelines have diverged. In the EU, CRR3 has applied since 1 January 2025, with the output floor phasing in from 50% toward 72.5% by 2030, while the FRTB market-risk start was pushed to 1 January 2027. The UK confirmed a 1 January 2027 start (with the market-risk internal model approach deferred to 2028). The US issued a re-proposal in 2026 on a broadly capital-neutral basis, with phase-in from 2027. The practical message for a model builder: the structure is stable, but the parameters and dates move — so build the model to let them change without a rebuild.

Structuring it in SAP Analytics Cloud

A clean Basel III model rests on a few dimensions and the same aggregation discipline that Solvency II demands.

Dimensions

Four dimensions carry most of the reporting: Risk type / exposure class (credit, market, operational; or the CRR exposure classes for the standardised approach), Capital tier (CET1, Additional Tier 1, Tier 2), Entity (solo vs consolidated group), and Version (Actual, plus your stress and ICAAP scenarios). A Risk-type dimension is what lets you show the RWA build-up as a waterfall and apply the output floor as a visible, explainable step.

Measures and their aggregation

This is where models silently break. RWA and exposures are additive across risk types and entities — SUM is correct there. But capital (CET1, Tier 1, Total) is a balance: across a time hierarchy it must aggregate with LAST (closing value), never SUM, or a yearly figure becomes the sum of four quarter-end balances and is meaningless. The same is true of HQLA and the funding measures behind the LCR and NSFR. We unpack this distinction in how to choose the right aggregation in SAC; for Basel III it is the single most important decision in the model.

The ratios must be calculated measures. The CET1 ratio is CET1 ÷ RWA computed at the level being displayed; the LCR is HQLA ÷ net outflows; the NSFR is ASF ÷ RSF. None of them may be stored as a number that SAC then sums or averages. Averaging a CET1 ratio across entities, or summing an LCR across business lines, produces figures that look reasonable and are completely false. Always compute numerator and denominator first, divide last.

The output floor as a calculation

Model both RWA figures — internal-model RWA and standardised RWA — as separate measures, then make reported RWA a calculated measure: max(internal-model RWA, 72.5% × standardised RWA). That keeps the floor transparent and lets you flag, on a dashboard, exactly when and where it binds. Hard-coding a single RWA number hides the very thing supervisors and your own risk committee want to see.

Versions: stress testing and ICAAP

Capital adequacy is a forward-looking exercise. The Version dimension is how the model answers "what if a downturn hits". Build an Actual version from the closing figures, then derive scenario versions — a recession, a rate shock, a funding squeeze — and let each ratio recompute. This is the Actual / Budget / Forecast / scenario pattern from using the Version dimension in SAC Planning, applied to capital and liquidity. Done well, the ICAAP stops being an annual document and becomes a live dashboard.

A worked mini-example

Suppose a bank reports CET1 capital of 90 and RWA of 1,000 under its internal models, while the standardised approach would give RWA of 1,300. The CET1 ratio on internal models is 90 ÷ 1,000 = 9.0% — comfortably above the 7% effective minimum. But the output floor sets a lower bound of 72.5% × 1,300 = 942.5, which is above the 1,000 internal figure only if the standardised number were higher; here 942.5 < 1,000, so the floor does not yet bind and reported RWA stays 1,000. Push the standardised figure to 1,500 and the floor becomes 1,087.5; reported RWA rises to 1,087.5 and the CET1 ratio falls to 90 ÷ 1,087.5 ≈ 8.3%. Same capital, lower ratio — the floor did its job, and a model that hard-coded RWA would never have shown it.

Common mistakes to avoid

• SUM on capital balances. CET1, Tier 1, Total capital and HQLA aggregate with LAST across time, not SUM.
• Storing ratios as data. CET1 ratio, LCR and NSFR are calculated measures — numerator and denominator at the display level, divide last.
• Ignoring the output floor. Carry both RWA figures and compute the floor explicitly so you can see when it binds.
• One version only. Without scenario versions the model cannot support stress testing or the ICAAP.
• Mixing solo and consolidated. Keep an Entity dimension so solo and group views never contaminate each other.

Where this fits

Basel III is the banking twin of Solvency II for insurers — same "calculate, don't sum" discipline, different ratios — and banking groups increasingly report ESG risk alongside capital. If you want a running start, our banking regulatory reporting template and credit-risk template ship with the dimensions and ratio measures already structured, and the verified & maintained regulatory kit adds pre-filled thresholds and an import guide.

Sources

BCBS Basel III framework and the 2017 finalisation standards (Bank for International Settlements); EU CRR3 / CRD6; UK PRA Basel 3.1 policy. Implementation dates and the output-floor phase-in differ by jurisdiction and have been revised repeatedly — always check the current BCBS and national-regulator texts for the parameters in force.