Blog · · EN
Your Savings Account Is a Financial Contract — You Just Don't Know Its Type
What ACTUS taught us about 745 Swiss retail banking products.
We parsed the fee schedules and interest rate sheets of 26 Swiss cantonal banks. That gave us 745 products, 1,560 fees, and 410 interest rates. Then we tried to make sense of it all — and discovered that an obscure financial standard from Zürich had already solved the hard part, twenty years ago.
The problem: 745 products that are secretly 6 things
Every Swiss cantonal bank has its own product names. BEKB calls it a “Sparkonto,” GLKB calls it “Glarix-Sparkonto,” BCJ calls it “Compte épargne,” BSKB has an “IV-Rentenkonto” (for disability pension recipients), and GLKB even offers a “Grabfondssparkonto” (a savings account for grave maintenance — yes, really).
These are all the same thing. They are interest-bearing cash accounts with no fixed maturity that the customer can withdraw from at any time, possibly subject to notice periods or withdrawal limits.
We discovered this pattern everywhere. Across 26 banks and 745 products, the entire retail product universe collapses into a handful of fundamental contract types. The question was: has someone already formalized this?
Someone has.
Enter ACTUS
ACTUS — Algorithmic Contract Types Unified Standards — is an open standard created by Dr. Willi Brammertz and a team at ZHAW Zürich. The core insight is disarmingly simple:
Every financial contract in existence — every loan, every deposit, every bond, every swap — is really just an agreement about future cash flows. And there are only about 30 distinct patterns of how cash flows can be exchanged.
ACTUS defines these patterns as contract types, each with a standardized set of attributes and a deterministic algorithm that computes all future cash flows given those attributes. The data dictionary is open, the reference implementation is in Java, and the standard is now an ISO TC68/SC9 liaison. The US FDIC references it for regulatory purposes.
Brammertz laid out the theory in his 2009 book Unified Financial Analysis: The Missing Links of Finance, which argues that a single, well-organized set of contract types and algorithms can derive all types of financial analysis. Risk, valuation, accounting, regulatory reporting — all from the same building blocks.
The idea influenced a generation of Swiss financial software. The instrument taxonomy in core banking systems — the way Avaloq, Finnova, and others classify financial products internally — echoes the same principle: don't model products, model the underlying cash flow patterns.
Mapping 745 products to ACTUS
Here is what happens when you take our 745 Swiss retail banking products and map them to ACTUS contract types:
| Retail product | ACTUS type | Why |
|---|---|---|
| Privatkonto, Sparkonto, Jugendkonto, IV-Rentenkonto, Grabfondssparkonto | UMP | Interest-paying, no fixed maturity, callable by client |
| Säule 3a, Freizügigkeit | UMP (constrained) | Same cash flow pattern, but with regulatory withdrawal restrictions |
| Kassenobligation, Festgeld, Fiduziareinlage | PAM | Fixed term, periodic interest, full principal at maturity |
| Hypothek (Festhypothek) | ANN | Periodic payments of principal + interest over fixed term |
| Hypothek (SARON) | ANN (variable) | Same structure, floating reference rate |
| Überziehungskredit, Lombardkredit | CLM | Credit facility, callable, no fixed maturity |
| Debitkarte, Kreditkarte, TWINT | CSH / payment instrument | Cash flow access layer, not a contract per se |
| Staatsgarantie | CEG | Third-party guarantee on payment obligations |
That's it. 745 products. 6 contract types. Everything else — the product names, the branding, the age segmentation, the package bundling — is marketing wrapped around a handful of cash flow patterns.
What this means, concretely
1. The BEKB Kinderkonto and the GLKB Glarix are the same contract
Both are UMP contracts. They differ in exactly three ACTUS attributes:
nominalInterestRate(IPNR): 1.50% vs. 0.40%notionalPrincipalcap: CHF 1,000 vs. CHF 250,000contractPerformanceCondition: age < 18 vs. none
In our database, these are two separate products with different names, different banks, different source documents. In ACTUS terms, they are the same algorithm with different parameters. Every future cash flow — every interest payment, every withdrawal — is deterministically computable from those three values.
2. The BCJ Consignation-Konto vs. a normal savings account
The Banque Cantonale du Jura has a “Compte de consignation” that costs CHF 250 to close. A normal savings account costs CHF 12 to close. Both are UMP contracts. The difference? A single ACTUS attribute: penaltyRate or penaltyType on the termination event. The 20x price difference in closure fees is a parameter, not a different product.
3. A Festhypothek and a Kassenobligation are cousins
A fixed-rate mortgage (ANN) and a medium-term note / Kassenobligation (PAM) look completely different in a bank's product catalog. One is under “Lending,” the other under “Savings.” But in ACTUS terms, they are close relatives — both are fixed-income contracts with known maturity dates. The difference is in contractRole (are you the borrower or the lender?) and in the amortization schedule (periodic principal repayment vs. bullet at maturity).
An AI agent that understands ACTUS contract types can instantly see that a bank offering 0.95% on a 10-year Kassenobligation while charging 2.1% on a 10-year Festhypothek is earning a 115 basis point spread on essentially the same risk profile in opposite directions.
4. Banking packages are portfolios of contracts
A “ZKB inklusiv Silber” package, a “JUKB e-PACK VIP,” or a “BCVs Conto Pack Premium” are not products. They are portfolios — bundles of UMP + CSH + payment instruments, with a meta-fee layered on top and conditional discounts based on the portfolio composition.
ACTUS doesn't have a native “package” contract type, because packages aren't contracts — they're commercial wrappers. This is a useful distinction. The cash flow obligations come from the individual contracts inside the package. The package fee is just a recurring payment for the privilege of having them bundled.
What ACTUS doesn't solve
ACTUS was designed for institutional risk management and regulatory reporting. It models the financial mechanics of contracts. It deliberately ignores several things that matter enormously in retail banking:
- Pricing. ACTUS tells you the structure of cash flows, not the price. It knows that a savings account pays interest; it doesn't know that GLKB pays 0.40% while BCJ pays 0.00%. Pricing is a parameter, not part of the standard.
- Fees. Our dataset has 194 distinct fee types — dormancy fees, counter transfer fees, paper statement surcharges, FX markups, card replacement fees. ACTUS has
feeRateandfeeBasisbut not the granularity of real-world retail fee schedules. - Conditional pricing. UBS charges CHF 0, 3, 5, or 7 per month for the same Privatkonto depending on your asset level, document preference, and product line. This if/then pricing logic is richer than what ACTUS attributes can express.
- Marketing and brand. ACTUS doesn't care that BSKB has an “IV-Rentenkonto” for disability recipients or that GLKB has a grave maintenance fund. Both are
UMP. The social purpose of a product is invisible to the contract type.
This is not a criticism. It's the correct separation of concerns. ACTUS models the financial contract. The retail product layer — names, fees, conditions, eligibility, branding — sits on top.
The architecture that makes sense
After parsing 26 banks and staring at 1,560 fee line items, here's what we think the right data model looks like:
Layer 1: ACTUS contract types. The foundation. Every financial product maps to one of ~6 contract types (UMP, PAM, ANN, CLM, CSH, CEG). Standardized attributes. Deterministic cash flow algorithms. This is the layer where a computer can reason about financial products without knowing their names.
Layer 2: Product catalog. The human layer. Product names, categories, bank-specific identifiers, eligibility criteria, age segments, regulatory constraints. This is where “Grabfondssparkonto” lives. It maps 1:1 to an ACTUS contract type but adds the commercial and social context.
Layer 3: Pricing. The comparison layer. Fees, interest rates, FX markups, package discounts, conditional waivers. 194 fee types, 410 interest rates, the full mess of real-world pricing. This is what changes quarterly and what customers actually care about.
Layer 4: Source documents. The trust layer. Every data point links to a specific page in a specific PDF or HTML document published by the bank. Verifiable. Auditable. This is what makes the data credible.
We currently have layers 2, 3, and 4. Adding layer 1 — ACTUS contract type annotations — would make the data dramatically more powerful. An AI agent could reason: “Show me all UMP contracts with IPNR > 0.003 and feeRate = 0” instead of guessing which product names correspond to savings accounts.
Why this matters now
Three trends are converging:
AI agents need structured financial data. When a user asks Claude or GPT “which bank should I use?,” the agent needs to query a database of financial products. Product names are useless without a taxonomy. ACTUS provides that taxonomy — for free, as an open standard.
Regulators are moving toward granular data. The US FDIC already references ACTUS. The EU's BIRD (Banks' Integrated Reporting Dictionary) and the ECB's AnaCredit push in the same direction: away from aggregate reporting, toward contract-level granularity. ACTUS is the obvious bridge.
Open banking needs a contract language. APIs can move money, but they can't describe products. A machine-readable format for bank products — standardized, versioned, discoverable — would do for banking what
schema.orgdid for web content. ACTUS contract types are the natural vocabulary for such a format.
The irony is that ACTUS was designed for institutional finance — for banks managing trillion-dollar portfolios and regulators monitoring systemic risk. But its most intuitive application might be the simplest one: helping a family in Glarus understand that their “Glarix-Sparkonto” and a “Compte épargne” in Jura are the same financial contract with different parameters.
One pays 0.40%. The other pays 0.00%. ACTUS can't tell you which is better. But it can tell you they're the same question.
ACTUS (Algorithmic Contract Types Unified Standards) is an open standard maintained by the ACTUS Financial Research Foundation. The standard was created by Dr. Willi Brammertz and collaborators at ZHAW Zürich. The data dictionary and reference implementation are available on GitHub. The theoretical framework is described in Unified Financial Analysis: The Missing Links of Finance (Brammertz et al., Wiley, 2009). The retail banking data referenced in this article covers 26 Swiss banks, 745 products, 1,560 fees, and 410 interest rates as of Q2 2026.