bankproductsBlog

Blog · · EN

What Swiss banks do not publish

240 products, 64 banks, seven pull requests — and a map of what's missing from a bank price schedule.

TL;DR 240 products at 64 Swiss banks had no published price or interest rate in our database — a blank cell in the ACTUS Explorer. Unclear whether we had loaded sloppily or the bank was deliberately silent. In seven days we closed all 240 gaps: about 75 with real numbers, about 165 with an explicit "on request" tag. Cleaning up produced a map: cantonal banks publish almost everything, private banks almost nothing. The most surprising gap: at 31 banks, the Pillar 3a retirement account has no published interest rate.

Clicking on a product in our ACTUS Explorer, the price-or-rate cell was blank for 240 of roughly 2 700 published retail products.

For the reader, indistinguishable between:

  • "The bank published the price and we failed to load it", or
  • "The bank doesn't publish this price at all — quoted on request only".

In seven days we closed all 240 gaps. Cleaning up produced a small map — what Swiss banks actually publish and what they don't.

Three outcomes per product

Working through the list systematically, three outcomes emerged per gap:

1. Bank publishes, we hadn't loaded (~75 products). The classic load-gap. Aargau Cantonal Bank student account → 0.00 % interest, free maintenance, clear on the product page. Migros Bank 2-year fixed mortgage → 1.40 %, clear in the Q3 refresh PDF. These got loaded with real numbers and a source reference.

2. Bank does not publish (~165 products). Construction-loan accounts, wealth-management mandates, individual corporate loans, pension-fund brokerage products. For these, clicking on the bank's website returns not a price but "schedule a consultation" or "conditions on request". Loaded with an explicit NULL-anchor row — an honest non-number with a reason.

3. Product should not exist. In some cases a product row in our DB was speculation from an earlier load — the bank no longer lists it on the website. In this series we didn't hit a clear case of this, but the outcome is documented in the 3-outcome triage workflow so it stays visible.

What the widget shows now

For outcome 1, real numbers. For outcome 2, a small italic tag: "Zinssatz auf Anfrage" or "fee auf Anfrage" (rate/fee on request). For outcome 3, the product isn't there.

Cells are no longer blank. No more guessing whether we were sloppy or whether the bank is deliberately silent.

Example — Alpha Rheintal Bank has six mortgage products in our DB:

ProductBeforeAfter
Alpha FEST Hypothek 2y(blank)1.45 %
Alpha BAUKREDIT(blank)conditions on request
Alpha SARON FIX Hypothek(blank)SARON + individual margin — on request
Alpha SARON VARIA Hypothek(blank)SARON + individual margin — on request
Alpha VARIA Hypothek(blank)variable rate — on request
Alpha ERNEUERUNGS Hypothek(blank)rate-discounted renovation financing — on request

Only the fixed mortgage is published as a concrete curve. The other five are negotiated bilaterally. Both are legitimate business practice — but the difference should be visible.

What banks don't publish — by category

Across the entire database (573 NULL-rate rows with an explanatory note):

Product categoryNULL rowsWhat that typically means
Savings account128Sub-variants (investment savings, deposit-guarantee, seniors' savings) — base savings has a rate, variants inherit "same as base"
Mortgage121SARON / variable / construction — negotiable; only fixed-mortgage has a curve
Current account116Mostly sub-variants (foreign-currency, numbered, client-deposit accounts)
Term deposit54On request, tenor- and market-dependent
Pillar 3a36Retirement account rate mostly NOT published (interesting — unlike many international peers)
Metal account30Almost never published — custody fee maybe, interest income no
Vested benefits24On request / via a separate foundation

The most surprising point: Pillar 3a retirement accounts have no published interest rate at 31 banks (36 NULL rows across 31 banks, because some banks offer multiple 3a variants). On an account whose sole purpose is to accumulate interest over the long term.

For customers who buy into Pillar 3a for exactly that reason. It's one of the notable transparency gaps in the Swiss banking sector — not illegal, but hard to defend.

What banks don't publish — by bank type

The 64 banks with gaps sorted cleanly into three groups:

Cantonal banks (24 banks loaded, 108 products fixed in the series). Publish almost everything — price schedules are thick and detailed. Gaps at AGKB, AKB, LUKB & co. were almost always load-gaps (outcome 1), not real "on request" cases. The large cantonal banks are the most transparent players in the dataset.

Regional banks and Sparkassen (~20 banks). Publish standard accounts with rates; for sub-variants (sustainability mortgage, family mortgage) mostly only a "discount on the base mortgage rate" — no absolute rate. Here the NULL-anchor ("modifier on base product X") was the honest answer.

Private banks and niche players (~20 banks). Publish little. LGT publishes a single base rate covering 5 savings-account variants. Cornèr Banca lists 4 credit cards whose prices live on a separate domain (cornercard.ch). Bank du Léman offers 10 "Point *" packages, 5 of which had no published price in our crawl before the series. At private banks, "not published" is the rule, not the exception.

Why we didn't make up a number

Filling in 240 gaps, the temptation wasn't small — for the 5 Bank du Léman packages we could have plausibly assumed CHF 8 per month; for the 5 LGT savings variants we could have carried over the base rate. Both would be dangerous.

Our three anti-hallucination rules (from CLAUDE.md, deep in the project docs):

  1. Never invent company names. If we haven't verified a bank's legal name in a source visited in the same session, no value.
  2. Never invent product names. If the bank's pricing material doesn't name a product, no product name in our DB.
  3. Never invent values to avoid NULL. NULL is strictly better than a plausible placeholder, because a placeholder looks legitimate to downstream validators and to buyers.

These rules fired 132 times in this series. Every anchor row has NULL for the amount and a note explaining why. The landing cells now showing "on request" instead of a fabricated number — that's the core benefit, not the number of filled rows.

Two real near-fabrications from earlier sessions made these rules explicit: once an LLM-generated UID for Zurich Cantonal Bank (CHE-105.939.310, which Zefix returns as non-existent; correct: CHE-108.954.607), once a plausible-sounding but non-existent company name (Frankly Technologie AG). Both nearly shipped to production. The three rules exist so it doesn't happen again.

What this means for buyers

When you open a bank in our Explorer and see a product, you now know:

  • Row with a number → bank publishes, we loaded it (with source reference in the DB)
  • Row with an "on request" tag → bank offers the product, but without a public price
  • No product in the list → bank doesn't offer this (or we haven't captured it as a product)

The uncertainty "did the bank not publish it, or is bankproducts.info sloppy?" is gone. What remains is the honest signal: where a bank stays silent, we say "it stays silent" — not "we don't know".

What's still open

Six crawl gaps queued for the next quarterly cycle:

  • Acrevis "Übersicht Dienstleistungspreise" PDF (for 3 package classes)
  • BCN Grille tarifaire (for Ecohabitat / Hypoflex details)
  • LGT Sparkonto variants factsheets (5 variants)
  • Cornèrcard tariffs (4 credit cards via cornercard.ch)
  • BCJ Q3 crawl (only Q2 available so far)
  • Banque du Léman Fiche produit (5 "Point *" packages)

All are documented as comments in the respective PRs (#333–#339). Once they appear in the next crawl cycle, the corresponding anchor rows can be replaced with real numbers — no widget rework needed.