Structured Notes: Who Is on the Other Side of Your 15% Coupon?

On May 6, 2026, Citigroup priced a note that pays 15.25% a year as long as NVIDIA does not fall more than 35%. Two weeks later the filing landed on EDGAR, and on the cover, in the issuer’s own words, was this: the estimated value of the securities is $987.60 per $1,000.¹ Citi was telling buyers, in a document most of them would never open, that it valued the thing at twelve dollars and forty cents less than they were paying for it.

That gap is the least interesting part of the trade. The interesting part is what the buyer was selling, and to whom. Bloomberg’s Money Stuff: The Podcast put it plainly in a July 31, 2026 episode, describing a discussion of “sourcing weird volatility risk from retail customers to sell to hedge funds.”² The US structured note market issued roughly $180 billion across 53,700 notes in 2025.³ A large share of that is retail and private-bank money writing options without quite realizing it.

Below, three real notes are rebuilt from their prospectuses. Two of the issuers’ own published valuations come back out of the model within a dollar, and the outcome distribution that falls out of the same terms is worth more than the coupon is. The tool does the same for any terms you type in.

Where the risk goes

Start with a hedge fund that wants to buy crash protection on an index or a single stock. It buys a put from a dealer. The dealer is now short that put and needs the opposite exposure.

The dealer has several ways to get it: buy a listed put, hedge dynamically by shorting the underlying as the position’s delta moves, or trade with another institution. There is also a fourth route. The structuring desk can manufacture a note whose buyer is economically writing a similar option, and sell that note through the private bank.

An options market maker or a pension fund will take almost any lawful, hedgeable exposure at the right price. What they will also do is value each embedded option independently, put it out to several dealers, compare the implied volatility and skew against screen prices, negotiate collateral and termination terms, and walk away if the premium is thin.

A private-bank client comparing a 15.25% coupon against a 4% Treasury yield does none of that. The dealer’s advantage is that one counterparty can observe the price of what it is selling and the other cannot. That is a narrower claim than “banks bet against their customers,” and it is the one the evidence supports.

One more correction to the popular version of this story: the dealer is usually not matching your note against one identifiable hedge fund trade. It runs a book. Retail structured-product flow goes into that book, gets netted against institutional flow, and whatever is left over is hedged in listed markets. The dealer is a market maker managing inventory. The pricing question survives that; it just means nobody is sitting across a table from you personally.

What is inside the note

A structured note is an unsecured debt obligation of a bank with one or more derivatives embedded in it, and a payoff computed from a formula. Unlike an ETF it holds no portfolio corresponding to the advertised exposure. You have a contractual claim against the issuer and nothing else. The SEC puts the decomposition this way in its investor bulletin on structured notes:

“The estimated value of the notes is likely lower than the issuance price of the note to investors because issuers include the costs for selling, structuring or hedging the exposure on the note in the initial price of their notes.”⁴

So the note is issuer debt, plus embedded options, minus the costs of putting it together. In a high-coupon note the options run in one direction: you are the seller. The coupon is not yield in the sense a bond coupon is yield. Part of it is the issuer’s borrowing cost, and the rest is the premium for an option you wrote.

Which reframes the question a buyer should ask. Instead of “why is this paying 15.25% when bonds pay 4%?” the question is: what am I selling for the extra 11.25 points, what would that risk fetch in a competitive derivatives market, and how much of that value is being passed through to me?

Opening a real term sheet

The Citigroup note is a good specimen because nothing about it is exotic. It is an ordinary autocallable contingent-income note, the single most common shape in the market. From the pricing supplement:¹

Three features deserve attention before any math. First, the 65% level is a barrier, not a buffer. If NVIDIA finishes at 64% of where it started, you do not lose 1%. You lose 36%, because principal is reduced by the full decline from the initial value once the barrier breaks. The word “buffer” does not appear anywhere in the document, and that is deliberate.

Second, the upside is the coupon and nothing beyond it. NVIDIA could triple; you would still receive 15.25% a year and then have your money returned.

Third, the issuer chooses when to end the trade. You do not. That asymmetry has consequences we can measure.

Morgan Stanley, in a January 2026 note of its own, describes the pricing machinery with unusual candour:

“we use an internal funding rate, which is likely to be lower than our secondary market credit spreads and therefore advantageous to us. If the issuing, selling, structuring and hedging costs borne by you were lower or if the internal funding rate were higher, one or more of the economic terms of the securities would be more favorable to you.”⁵

That is the issuer stating, in its own filing, that the note is priced off a borrowing rate chosen because it is advantageous to the issuer, and that the buyer’s terms would be better otherwise. None of the critique that follows depends on anything hidden. It depends on a document that runs to dozens of pages and arrives after the sales conversation.

Rebuilding the dealer’s number

Here is the test that makes this concrete. The prospectus gives you every term. If you simulate the note under standard assumptions, can you recover the estimated value Citi printed on the cover?

You can. Running the exact terms through a quarterly Monte Carlo with a 41% volatility assumption produces a value of about $989 against Citi’s stated $987.60. The JPMorgan note on the same stock, priced one day later off the identical $207.83 strike, has a filed estimated value of $978.50; the same model reproduces it at $978 with 43% volatility.⁶ Two banks, one underlying, two independent filings, and a consistent implied volatility in the low forties, which is a reasonable place for multi-year NVIDIA options to trade.

The Citi S&P 500 note from July 2026 reconciles the same way at about 23% volatility against its filed $996.70.⁷ That is well above the roughly 18% you would see for at-the-money three-year index options, and the gap is informative: the option embedded in the note is struck 30% below spot, and downside implied volatility is persistently higher than at-the-money volatility. Which points at the main limitation of everything below.

A constant-volatility model ignores that skew. Real dealers price with a volatility surface, and the skew makes the downside you sold more valuable than a flat model says. The error runs in the note’s favour. Every give-up figure here is therefore a floor, not a worst case.

Load the Citi preset and the decomposition appears in the second panel. Of the $989 of model value, roughly $183 is the discounted value of the contingent coupons and roughly $806 is the principal you expect back. Set against that, the downside you wrote is worth about $125 in today’s money. That last number is the insurance you sold. It is what the coupon is paying for, and it is the number the phrase “65% downside protection” is describing.

Where the return comes from

Valuation tells you what the package is worth. It does not tell you what is likely to happen to you, and those are different questions. Pricing models work under risk-neutral probabilities, which are prices rather than forecasts. To ask about outcomes you have to supply a real-world expected return, and the answer is only as good as that input.

So here is the same note under a range of assumptions about NVIDIA’s performance, comparing the note’s average profit against Treasury bills held for the same length of time as the note happened to last on each path:

Read the first row carefully, because it is the whole argument. If NVIDIA earns only the risk-free rate, the note returns $49 per $1,000 while Treasury bills return $76 over the same holding period. The note loses to cash.

Every dollar of expected return above bills is therefore equity risk premium: the ordinary compensation for owning equity, delivered through a wrapper that hands you roughly a quarter of the gain while leaving your exposure to the full decline below 65% intact.

A caveat on that last column, in fairness to the note. The stock comparison runs the full three years while the note usually returns your cash long before that, so it is not like for like. You would reinvest. But that is the point rather than an objection: you would have to find somewhere to put the money, probably into a market that has just risen, and probably into another note at worse terms. Reinvestment risk is a real cost of the structure and it shows up nowhere in the coupon.

The shape of the payoff

Assume a 12% expected return on NVIDIA, generous by most standards. The distribution looks like this:

• Median outcome: +$153. Mean outcome: +$103. When the mean sits below the median you are looking at a left-skewed distribution, which is the signature of a short-volatility position. Most of the time you make a little. Occasionally you lose a lot.
• Worst 1 in 20: minus $492. Worst 1 in 100: minus $688. The best 1 in 20 is +$343, and it cannot be better, because the coupon caps it.
• 15% chance of losing money; 9% chance of losing more than 35%.
• 77% of the time the note is called early, and 53% of the time it is called at the first opportunity. Median holding period: one year, on a three-year note.

The last statistic has a consequence that is easy to miss. Of the outcomes that do run to maturity, 66% finish below the barrier, and the average profit among them is minus $154. The note gets called away when things go well and stays on your books when they do not. The 15.25% runs for as long as the issuer finds it convenient, and that window correlates almost perfectly with the stretch you would rather have been holding something else.

The same product, eight times the cost

Across four filings from three banks in 2026, the day-one give-up between what you pay and what the issuer says the note is worth ranges from $3.30 to $27.70 per $1,000:

The cheapest of these is the fee-based advisory version, which carries no selling commission because the adviser is paid elsewhere. If you are going to buy a note anyway, ask whether an advisory share class exists. The economics of the payoff do not change, but roughly two percent of your principal does.

The two NVIDIA notes make the case that terms matter more than the headline. They are on the same stock, struck at the same $207.83, priced a day apart, with the same 65% barrier and coupons within a quarter of a point of each other. Run both at an 8% expected return and the Citi note earns $78 against $72 from Treasury bills, while the JPMorgan note earns $52 against $57. One clears the cash hurdle and the other does not, because the JPMorgan version pays a larger commission, runs two years instead of three, and observes monthly, which gives the issuer twelve extra chances a year to hand the money back. None of that is in the coupon.

For scale on distribution economics generally, J.P. Morgan’s own brokerage disclosure states that it “is typically paid a distribution fee of up to 5% of the notional amount of the security.”⁸ That is the ceiling across its structured product shelf, not the fee on any particular note, but it indicates the range the category operates in.

What the research finds

Start with the most awkward result for the product, because it does not depend on fees at all. Perusset and Rockinger simulated fairly priced convertible bonds, reverse convertibles and barrier reverse convertibles inside a 60/40 portfolio and found that including them “generally lowers returns and risk-adjusted performance across all product types,” with an opportunity cost that is “overwhelmingly negative.” They expect “further performance deterioration for real-world structured products that are highly illiquid, exhibit high transaction costs, and are often more complex.”⁹ Strip out every dollar of markup and the payoff shape still does not earn its place in a portfolio.

On pricing, Henderson and Pearson studied 64 issues of a popular retail structured equity product and found offering prices “on average, almost 8% greater than estimates of the products’ fair market values,” with a mean expected return “slightly below zero.” Their conclusion was that it is “difficult to rationalize their purchase by informed rational investors.”¹⁰ The 2026 notes examined here are priced far better than that, which is worth saying plainly: disclosure of estimated values, which the SEC pushed issuers toward after that era, appears to have narrowed the gap.

On why complexity exists, Célérier and Vallée text-analyzed 55,000 European retail structured products issued from 2002 to 2010. As interest rates fell, headline rates detached from prevailing rates, complexity rose, and riskier structures became more common. Higher-headline-rate, more complex products “appear more profitable to the banks distributing them.” Their conclusion: “financial complexity is a by-product of banks catering to yield-seeking investors.”¹¹

On why people buy them, Hens and Rieger found that for rational investors with correct beliefs, the utility gain from structured products over a simple mix of the risk-free asset and the market portfolio is “typically much smaller than their fees.” The gain becomes sizeable only under behavioral preferences or incorrect beliefs, such as probability weighting and misestimation.¹² Which is a technical way of saying these products are designed for how people actually evaluate probabilities rather than how they should.

When a note is a reasonable buy

There are real cases, and a guide that skipped them would be less useful.

• Matching a specific liability. If you have a known obligation on a known date and a note produces the cash flow that meets it, the customisation has genuine value.
• An exposure you cannot otherwise access. Some accounts, mandates or jurisdictions prohibit direct options trading but permit notes. A worse instrument you can hold beats a better one you cannot.
• Competitive pricing. Independent multi-dealer auction platforms exist. Putting a structure out to five banks changes the economics materially compared with accepting the first term sheet.
• Turning a behavioral preference into a binding rule. If a defined payoff is what keeps you invested through a drawdown, that has value even though it does not show up in a pricing model.

The test is whether the value of the customisation exceeds the markup, plus the illiquidity cost, plus the issuer credit cost, plus the tax cost, plus the cost of holding something you cannot fully evaluate. For an institution with a genuine liability to hedge, that inequality can hold. For a DIY investor adding a note to a diversified portfolio because the coupon looks attractive, it generally does not.

Reasons that do not survive contact with the math

These are the ones worth being suspicious of:

• “The stock has never fallen 35% in three years.” NVIDIA fell more than 60% in 2022. Barriers get tested precisely in the environments nobody modelled.
• “There is 35% of downside protection.” A barrier is not a buffer. Breach it and you absorb the entire decline from the starting level, not the part beyond the barrier.
• “It has paid every coupon so far.” Of course it has. In this model at least one coupon is missed only about 29% of the time, and most of the danger is concentrated in the paths that survive to maturity.
• “I would not mind owning the stock anyway.” Then own the stock. It expects roughly four times the profit over the same window, with uncapped upside and no issuer credit risk.
• “The bank is investment grade.” True and insufficient. The SEC notes these are unsecured obligations and any protection “is only as good as the financial health of the structured note issuer.”⁴
• “I can sell it if I need to.” The notes here are not exchange-listed. The SEC observes that “the only potential buyer for your structured note may be the issuing financial institution’s broker-dealer affiliate,” and that issuers “often specifically disclaim their intention to repurchase or make markets in the notes they issue.”⁴

FINRA reached a similar place from the regulatory side in Notice 12-03, treating knock-in and knock-out features and worst-of structures as markers of complexity, and directing registered representatives to consider “whether less complex or costly products could achieve the same objectives for their customers.”¹³

Building the payoff yourself

The replication is not difficult, which is the point of understanding the decomposition. The Citi note is approximately a Treasury bill plus a written put on NVIDIA struck 35% below the money. Both legs are available to any brokerage account with options approval, priced on a screen, in a cleared market, with a bid and an offer you can see and a position you can exit on any trading day.

Doing it yourself means you observe the option premium instead of inferring it, you keep the collateral yield rather than lending it to a bank at its internal funding rate, you take no issuer credit risk, and you can close the position when your view changes. It also means you have to manage it, and that a listed put will not match a note’s exact barrier and autocall mechanics. That is a fair trade for most people.

The listed version of the same idea

A fund category exists to deliver defined-outcome payoffs on an exchange: buffer ETFs, run by Innovator (acquired by Goldman Sachs Asset Management in April 2026 and still operating under its own brand), First Trust, AllianzIM and Calamos among others. FINRA drew the line between them and notes explicitly in Regulatory Notice 22-08:

“Unlike structured notes offering similar exposure but without an underlying portfolio, these defined outcome ETFs typically invest in exchange-listed options to provide the exposure.”¹⁴

Three things genuinely improve. The fund holds FLEX options cleared through the Options Clearing Corporation, so as one prospectus puts it, “the Fund’s counterparty is a clearing house, such as the OCC, rather than a bank or broker.”¹⁵ Every issuer still discloses residual OCC risk, so the credit exposure moves from one bank’s unsecured balance sheet to a mutualized clearinghouse rather than disappearing. The shares trade daily at a visible price. And the cap and buffer sit in a prospectus you can read before buying.

Two things do not improve, and the prospectuses say so plainly. The stated outcome applies only to someone who holds from the first day of the outcome period to the last; First Trust puts it in bold, that “if you purchase shares after the Target Outcome Period starts or sell your shares before the Target Outcome Period ends, you may receive a very different return.”¹⁶ And the cap resets every period, which Innovator warns can compound against you: “over multiple Outcome Periods, the Fund may have losses that exceed those of the Underlying ETF.”¹⁵

Being more transparent is a different question from being a good investment, and on the second the evidence is unkind. AQR reviewed the category in the Journal of Portfolio Management and found that buffer funds “often underperform their reference assets in both returns and risk-adjusted terms,” with a median excess Sharpe ratio of minus 0.07 against the asset they track, and that “simple alternatives like mixing equities with cash generally outperform buffer funds on average and even in drawdowns.”¹⁷ AQR sells competing products, which is worth weighing. The counterargument, from Morningstar’s Jeffrey Ptak, is that the verdict depends on the benchmark: measured against a 60/40 portfolio rather than the index they track, buffer ETFs have delivered similar returns at lower volatility and better Sharpe ratios.¹⁸ Category median expense ratio is 0.79%.

Worth noting where regulators place them. FINRA lists defined-outcome ETFs among complex products requiring heightened sales-practice care, and the SEC has published no investor bulletin devoted to them at all. The listed wrapper solves the credit and liquidity problems. It leaves the question of whether you wanted a capped, short-volatility payoff in the first place exactly where it was.

The broader test applies to anything with an attractive headline rate. Work out which options you are buying and which you are selling, whether your upside is capped, and what the same exposure costs in a transparent market. If the person selling it cannot express the payoff as an equation and name the embedded options, that is information about the product. If you cannot follow the explanation, that is information about whether you should own it.

Frequently asked questions

What is the difference between a buffer and a barrier?

A buffer absorbs the first N% of losses: with a 20% buffer, a 30% decline costs you 10%. A barrier is a trigger: with a 35% barrier, a 36% decline costs you 36%, because principal is reduced by the full decline from the starting value once the barrier breaks. Most high-coupon notes use barriers. Sales conversations often use the word “protection” for both.

Is the estimated value on the cover the fee?

Not exactly. The gap between the offering price and the issuer’s estimated value covers selling commissions, structuring, hedging costs and the issuer’s use of its internal funding rate. Some of that is genuine cost. It is best read as the minimum you give up on day one, not as a profit figure.

Why would the bank call the note away early?

Because the terms say it can when the underlying is at or above its starting level, and by then the coupon it is paying you has usually become expensive relative to what the same protection now costs. Early redemption is the issuer exercising an option you granted it at no charge.

Are structured notes FDIC insured?

No. Notes of this type carry explicit cover-page language that they are not FDIC insured, not bank guaranteed, and may lose value. They are senior unsecured debt of the issuer.

How are structured notes taxed?

It varies by structure and is often uncertain. The SEC states directly that “the tax treatment of structured notes is complicated and in some cases uncertain” and advises consulting a tax adviser before purchase.⁴ Contingent coupons are frequently taxed as ordinary income, which makes these a poor fit for a taxable account relative to the equity exposure they partly replace.

Does a higher coupon mean a better note?

It usually means a riskier underlying or a nearer barrier. The coupon is set so the embedded option premium clears; a 15.25% coupon on a single volatile stock and an 8.50% coupon on an index can represent similar value to the buyer. The coupon on its own tells you almost nothing about expected return.

Key takeaways

• The coupon is option premium. Ask what you are selling for the excess over the issuer’s ordinary bond yield, and what that risk is worth in a market where you can see prices.
• A prospectus contains enough to check the price. The filed estimated value on the cover of the Citi and JPMorgan notes is reproducible from public terms at a plausible volatility assumption.
• Expected return above Treasury bills is equity risk premium. Assume the underlying earns only the risk-free rate and the note loses to cash.
• The distribution is left-skewed and capped. Median better than mean, a 9% chance of losing more than a third, and no upside beyond the coupon.
• The issuer controls the exit. Called early most of the time; the paths that reach maturity are mostly the ones where the barrier broke.
• Share class changes the cost by several multiples. $3.30 to $27.70 per $1,000 across four 2026 filings for economically similar packages.

Related guides

• Covered Calls Are Not Free Income: the same short-volatility trade in fund form, where the cap is explicit and the wrapper is listed.
• Cash-Secured Puts: the do-it-yourself version of the option embedded in a reverse convertible.
• The Black-Scholes Equation: where the risk-neutral valuation used above comes from, and why its probabilities are prices rather than forecasts.
• Dividends Are Not Free Money: the same “cash flow is not income” mental model applied to equities.
• How Financial Sales Pitches Hide the Real Cost of Investing: the general pattern this guide is one instance of.
• A 1,000% Upside and a 100% Downside Does Not Make It a Good Bet: why a payoff shape says nothing about whether it is worth its price.
• Don’t Mix Insurance With Investing: the same packaging problem in a different product category.

Sources and method

1. Citigroup Global Markets Holdings Inc., 424(b)(2) pricing supplement no. 2026-USNCH31841, priced May 6, 2026. Source of all Citi/NVIDIA terms, the $987.60 estimated value, the $2.50 underwriting fee and the quoted payoff formula. CUSIP 17332VX58.
2. Bloomberg, “No Kiddie Pool,” Money Stuff: The Podcast, published July 31, 2026. The quoted phrase is from the episode’s own published description. The characterization of the mechanism in this guide is ours.
3. Structured Retail Products, “Top underlyings US 2025”, January 5, 2026. Commercial vendor data; there is no regulatory equivalent for this market.
4. US Securities and Exchange Commission, Office of Investor Education and Advocacy, Investor Bulletin: Structured Notes, January 12, 2015.
5. Morgan Stanley Finance LLC, 424(b)(2) pricing supplement no. 13,646, priced January 28, 2026. Source of the internal funding rate passage, the $972.30 estimated value, the $17.50 commission and the $5 structuring fee.
6. JPMorgan Chase Financial Company LLC, 424(b)(2) pricing supplement, priced May 7, 2026. Source of the $978.50 estimated value and the $4.00 selling commission. CUSIP 46660TW22.
7. Citigroup Global Markets Holdings Inc., 424(b)(2) pricing supplement no. 2026-USNCH33327, priced July 27, 2026. Source of the $996.70 estimated value and the advisory share class terms. CUSIP 17334CC77.
8. J.P. Morgan, Guide to Investment Services and Brokerage Products, 2026, page 17.
9. Florian Perusset and Michael Rockinger, “Do structured products improve portfolio performance? A backtesting exercise,” Journal of International Money and Finance, vol. 157 (2025). Publisher page.
10. Brian J. Henderson and Neil D. Pearson, “The dark side of financial innovation: A case study of the pricing of a retail financial product,” Journal of Financial Economics, vol. 100, no. 2 (2011), pp. 227–247. The product studied was Stock Participation Accreting Redemption Quarterly-pay Securities.
11. Claire Célérier and Boris Vallée, “Catering to Investors Through Security Design: Headline Rate and Complexity,” Quarterly Journal of Economics, vol. 132, no. 3 (2017), pp. 1469–1508.
12. Thorsten Hens and Marc Oliver Rieger, “Can utility optimization explain the demand for structured investment products?” Quantitative Finance, vol. 14, no. 4 (2014), pp. 673–681.
13. FINRA, Regulatory Notice 12-03, Complex Products, January 17, 2012.
14. FINRA, Regulatory Notice 22-08, Complex Products and Options, March 8, 2022. Source of the defined-outcome ETF definition and of their classification as complex products.
15. Innovator ETFs Trust, summary prospectus for the Innovator U.S. Equity Power Buffer ETF — August (PAUG), filed August 3, 2026. Source of the OCC clearing language and the multi-period cap reset warning.
16. First Trust Exchange-Traded Fund VIII, summary prospectus for the FT Vest U.S. Equity Buffer ETF — August (FAUG), filed August 18, 2025. Emphasis in the original.
17. Clifford Asness, Jordan Cao, Antti Ilmanen and Thomas Villalon, “Rebuffed: An Empirical Review of Buffer Funds”, The Journal of Portfolio Management, vol. 51, no. 10 (September 2025). The authors’ firm manages competing strategies. Category median expense ratio of 0.79% across 401 funds is their figure, from Morningstar data as of May 15, 2025.
18. Jeffrey Ptak, “Buffer Beef”, April 21, 2025.

Method. All simulation figures come from 20,000 antithetic geometric Brownian motion paths stepped to each observation date, run in your browser by the tool above. Valuation uses the risk-neutral drift; outcome distributions use the expected return stated in the text. The model assumes constant volatility, a barrier observed at maturity, and no issuer default. It does not model volatility skew, which would raise the value of the downside the buyer writes and therefore widen every give-up figure shown. Monte Carlo standard error on the valuation figures is under one dollar per $1,000.

Editor’s note

Educational content, not investment advice, and not a recommendation to buy or avoid any security. Specific notes are discussed because their terms are public; nothing here is a claim about any issuer’s conduct. Citations verified against primary sources on August 6, 2026.