Autocallable Structured Notes: A Practitioner's Guide to the Observation Calendar

What is an autocallable?

An autocallable structured note is a debt instrument with an embedded equity option, issued by a bank or its finance subsidiary, that can redeem itself early if its underlying assets perform in a specific way on a specific set of dates.

Three things distinguish it from a vanilla corporate bond. First, the return is linked to the performance of one or more underlying assets — single stocks, indices, ETFs, or baskets — rather than purely to the issuer's credit. Second, the note can be called automatically (hence the name) before its scheduled maturity if the underliers hit a trigger level on an observation date. Third, the coupon, where one exists, is often contingent — paid only if the underliers are at or above a coupon barrier on the relevant observation date.

The autocallable family is broad. A standard autocallable has a single trigger level evaluated on each observation date. A step-down structure has a trigger that declines over time, making early redemption progressively more likely. A step-up structure has a call premium that rises with each observation. A phoenix structure adds memory coupons — missed contingent coupons can be paid on a later date if the barrier is met then. These variants stack: step-down phoenix autocallables are common, as are step-down structures with step-up call premiums. The combinations are what make reference data hard.

Living with summary-level reference data

A reference data record on a structured note, from most vendors, looks fine on a fact sheet. Identifiers, issuer, headline coupon, maturity date, payoff archetype, a few key levels. For a casual look, it works. For anyone whose job actually touches the note, the gap between what the record captures and what the term sheet says shows up immediately. A few of the operational moments where that gap bites:

The ops team reconciling next observations. Monday morning: which notes in the book have observations this week? Without a materialized calendar, the answer requires reconstructing each note's schedule from frequency codes and endpoint dates, cross-referencing the prospectus to verify business-day conventions, and reconciling the result against an internal model. Across thousands of positions, this is not a query. It is a week-long research project, repeated every week.

The compliance team writing the client statement. The disclosure has to be right. The next observation date displayed to the client has to be the date the issuer named, not a date derived from a frequency. The "current call trigger level" has to be the level that actually applies this period, not the headline initial trigger from the cover page. If the reference data flattens the step schedule, the statement is wrong before it goes out.

The risk team running a scenario. What is the call probability on the next observation date? Easy if you know the trigger level on that date. Hard if the only data you have is the initial trigger, the final trigger, and a flag that says "step-down." Scenarios get run against approximations; the approximations propagate into hedge ratios and risk reports; the reports get used to make decisions.

The advisor explaining a missed coupon. A phoenix autocallable did not pay this period. The client wants to know why, and whether the missed coupon will accrue. The advisor's CRM shows that the note has memory: true and a 70% barrier. It does not show the specific memory rule — whether unlimited prior coupons accrue, whether there is a cap, whether the catch-up requires a higher level than the current coupon. The advisor calls the dealer.

The data team integrating across vendors. Three reference data sources, three different classifications for the same note. One calls it "autocallable contingent yield," another calls it "phoenix," a third gives it a proprietary code. None resolve the worst-of basket to underlier ISINs. The team writes glue code, maintains a translation table, fields complaints when something breaks downstream.

Each of these is fixable with a reference data feed that captures what the issuer actually wrote. None of them is fixable with a feed that captures the headline and stops.

Why this gap has persisted

Structured note reference data is hard. Issuers do not publish it in machine-readable form. Prospectuses are written for lawyers, not for parsers. Every dealer has its own document conventions, its own field labels, its own preferred phrasing for the same mechanic. And every note is a one-off — even within the same payoff archetype, no two deals are exactly alike.

The result is that the vendors who could plausibly cover this asset class each cover it in a partial way, shaped by the business they are really in.

The terminal-resident pricing engine. Some platforms can represent a structured note as a fully formed pricing object — payoff, schedule, underliers, the whole apparatus. The fidelity is real. The problem is portability: the data lives behind seat licenses inside the terminal. You can value the note at your desk; you cannot extract a normalized, joinable feed into your back-office systems, your data warehouse, or your client-disclosure stack. Model fidelity without data portability is a research tool, not a reference data feed.

The distribution platform's data store. Several structured-note distribution platforms maintain rich, accurate reference data on every note they offer — as long as it sits on their shelf. There is no public reference-data API, no coverage of notes a client bought before they joined the platform, and no coverage of notes bought through a different distributor. An island of clean data inside a sea of gaps.

The market-intelligence provider. Some firms publish excellent macro-level intelligence on the structured-note market — dealer rankings, volume by payoff archetype, what is selling this quarter, year-over-year issuance trends. Useful for product strategy and competitive positioning. Silent on the operational question: what are the exact terms of the specific notes I hold or distribute?

The bond-first reference data vendor. Several established reference-data vendors handle structured notes inside a broader fixed-income product. The schemas are normalized; the delivery is sensible. But the heritage is corporate and government bonds. Structured notes get folded in at the level of identifiers and headline terms — autocall mechanics flattened, payoff structures summarized as prose, observation calendars implied rather than materialized. A credible vendor in a category that isn't really their category.

The European-universe vendor. Vendors with strong European structured-product coverage typically build that coverage on issuer submissions — dealers and platforms in the European market that send their data into the vendor's pipeline. That works in Europe. In the US market, where almost every structured note flows through a 424B2 prospectus filing on EDGAR rather than through a submission pipeline, US coverage is close to empty unless an issuer chooses to participate.

No one has been wrong about what they built. They have built for adjacent problems. Reference data on US structured notes — at term-sheet depth, in a normalized schema, across the full universe — has been an asset class waiting for purpose-built coverage.

What SQX built

SQX has been pricing structured notes for years — more than 100,000 instruments under coverage on the broker-quoted pricing side, intraday updates from a network of domestic and foreign dealer desks, the largest US structured-note pricing universe in the market. The reference data feed is the natural extension: same notes, same dealers, same operational discipline, now with every field on the term sheet captured as data.

The coverage universe spans more than 350,000 instruments and $1.5 trillion in cumulative issuance, across 40-plus issuers and 3,000-plus unique underliers. Roughly 100 distinct payoff archetypes are represented — autocallables, reverse convertibles, participation notes, principal-protected, leveraged and inverse, digital, dual-directional, range accrual, credit-linked, and the long tail of variants on each. The historical record is complete. New notes flow in continuously from SEC EDGAR.

The data itself arrives as three files joined by a single instrument identifier. The first is the reference data file — one row per note, carrying the 60-plus single-valued fields: identification, classification, mechanics, autocall terms, coupon terms, issuer call rights, issuance economics, and tax treatment. The second is the underliers file, with one row per note per underlying asset, every underlier resolved to its own ISIN and cross-validated against an authoritative source, with initial fixing levels, fixing dates, basket weights, and credit-reference flags. The third is the payoff structure file, which gives the piecewise-linear payoff profile at maturity as a table of breakpoints you can plot directly — without reverse-engineering a paragraph of prose.

Every record is parsed directly from the issuer's source documents: the 424B2 prospectus filed with the SEC, the final terms document, the pricing supplement. Each filing goes through a multi-stage extraction pipeline with an independent review pass on every field. Amendments are layered onto the parent record without losing the original terms. When a filing is re-processed, prior analysis is wiped cleanly and rebuilt from source — no orphaned rows, no stale leftovers, no half-updated records. What the feed says today is what the source documents say today.

The observation calendar gets special treatment. For every autocall, contingent coupon, or issuer-call observation date the issuer has named, there is a row in the calendar carrying the date and every parameter that applies on it: the autocall trigger level, the call premium, the coupon barrier, the contingent coupon amount, the observation type, the observation scope. A five-year quarterly note has twenty rows. A note with sixty observations has sixty. Nothing to derive, nothing to interpolate, nothing to guess.

Delivery is pipe-delimited text files via SFTP, daily — each row carrying an action flag (insert, update, or delete) so consumers apply only what changed, and the files are idempotent: replaying yesterday's file produces yesterday's state. For interactive use, the same data is available through the HTTP API: query by ISIN, upload a PDF and receive the parsed reference record back, get programmatic access to the same nested structure that ships in the file feed.

SQX has been working with custodians, broker-dealers, asset managers, and fund administrators since 2001. The team is small and responsive — coverage requests, custom delivery formats, integration help, all handled by a person within a business day. This is what reference data on a structured note looks like when it is built from the term sheet up.

The observation calendar, up close

The observation calendar is the cleanest example of the gap between summary-level and term-sheet-depth reference data, and the cleanest demonstration of what SQX captures that other feeds do not. It is also the part of an autocallable where most of the economics actually live.

Two illustrations make the point.

Step schedules - where summaries fall apart: A common step-down trigger schedule on a five-year quarterly autocallable: 100% on observations 3 and 4 (after a six-month non-call period), then declining 5% per observation to a 70% floor, then holding at 70% for the rest of the note's life. Twenty observations in total.

A typical reference data record stores this as two numbers: an initial trigger of 100%, a final trigger of 70%, and a flag that says "step-down." The downstream consumer is expected to derive the actual schedule, usually by linear interpolation between the endpoints. The chart below shows what that derivation costs.

The actual schedule front-loads the trigger at 100% through observation four, descends in equal steps from observation five through observation ten, and then sits at 70% for the rest of the note's life. The naive two-endpoint interpolation crosses through the middle of the chart and never matches the issuer's schedule at any single point. Every call-probability calculation, every scenario-weighted valuation, every "what is the trigger level on the next observation" question gets a different answer depending on which line you read.

The materialized calendar in reference data: The right way to deliver this is one row per observation date, with every parameter that applies on that date carried as a field on the row. For the note above, with the addition of a 70% coupon barrier, a 2.00% contingent coupon with memory, a step-up call premium starting at 6% and rising 6% per quarter, and three underliers evaluated on a worst-of basis, the calendar looks like this:

Every downstream consumer sees the schedule the issuer actually wrote. Pricing models read the schedule directly. Risk systems compute call probability at each date with the right trigger. Operations teams pull the next observation across the entire book with a query. Compliance answers the question "what is the next observation date for this client's position" without phoning anyone. The fields the issuer named are the fields the data carries. Nothing more, nothing less.

The same principle applies across every field family on the note. Coupon memory rules, issuance economics, tax classification, underlier identification — all captured at the depth the term sheet states them, all queryable, all joinable.

What term-sheet depth enables

When the calendar is materialized and the surrounding fields are captured at term-sheet depth, the workflows that depend on this data become tractable rather than aspirational.

Valuation and pricing. Discounted-cashflow and Monte Carlo models for autocallables ingest the full forward schedule directly, producing comparable, repeatable valuations across the book.

Risk and scenario analysis. Call probability at each future observation date computes against the trigger that actually applies on that date. Scenario-weighted call probabilities and dollar-impact estimates run from data, not from derived approximations.

Operations and lifecycle event tracking. "What are the next observation dates across every note we hold" becomes a single query. Across thousands of positions, the difference between a daily operational report and a week-long research project.

Compliance and client disclosure. Statements show the next observation date and current trigger level the issuer actually wrote. Suitability reviews surface the gap between the public offering price and the estimated value at pricing. Periodic client communications represent contingent-coupon mechanics accurately, because the data behind them is accurate.

Screening and portfolio construction. "Show me every phoenix autocallable with memory coupons on worst-of indices, non-call period of six months or more, issued in the last twelve months, with an estimated value at issuance below 98% of par." Trivial when the fields exist as structured data. Impossible when they live in free-text prospectus descriptions.

Audit and reconciliation. Every reference-data field maps back to the issuer's source filing. When a record is challenged, the question can be answered: what filing said what, when, and where in the document does it appear.

This is the difference between reference data as a documentation artifact and reference data as a workflow input.

One note, two views

Consider an Auto-Callable Contingent Interest Note issued by a major US dealer: USD-denominated, five-year tenor, quarterly observations, six-month non-call period, three underliers on a worst-of basis, step-down trigger from 100% to 70%, step-up call premium starting at 6%, 70% coupon barrier, 2.00% contingent coupon with memory, 70% final barrier at maturity.

View one — the prospectus prose. "The Notes are subject to automatic redemption, in whole, on each Auto-Call Observation Date occurring on or after the First Call Date if the closing level of the Least Performing Underlying on such date is at or above the Auto-Call Trigger Level applicable to that date, as set forth in the schedule below. The Contingent Interest Amount for any Observation Date will be paid only if the closing level of the Least Performing Underlying on such date is at or above the Coupon Barrier (70% of the Initial Underlying Value). Missed Contingent Interest Amounts shall accrue and become payable on the first subsequent Observation Date on which the Coupon Barrier is met (the 'Memory Feature')."

View two — the SQX reference data record. An instrument row with every classification, mechanics, autocall, coupon, and economics field populated. Three underlier rows, each with its own ISIN, initial fixing level, and basket weight. A payoff structure table with the maturity breakpoints. An observation calendar with twenty rows — one per quarter — each carrying the autocall trigger, call premium, coupon barrier, and coupon amount for that date.

The first view is what the issuer wrote. The second is what the data should be. The difference is pretty significant.

Structured notes have always been operationally complex. The data infrastructure around them has not had to be. That's now changed.

SQX's structured note reference data is live now. Coverage spans more than 350,000 instruments, $1.5 trillion of cumulative issuance, 40-plus issuers, and 100-plus payoff archetypes. The feed is delivered as three clean, ISIN-joined files via SFTP, with the same data available through our HTTP API. Custodians, broker-dealers, asset managers, insurance companies, compliance teams, and fund administrators — anyone whose work touches a structured note — can move from reading prospectuses to running queries.

To talk through coverage, methodology, or specific instruments, reach out — we'd love to chat.

Want more SQX updates? Follow SQX on LinkedIn and subscribe to our
LinkedIn newsletter to stay ahead of the curve.