"concept_id": "convexity"
You run a consulting business. Base case Profit: $600K. You model customer Demand 20% above and 20% below your forecast, average the two scenarios, and get $575K - twenty-five thousand less than your base case. Your P&L loses value from uncertainty itself. Across the hall, a SaaS Operator models the same Demand swing on her P&L. Her base case Profit is also $600K. Her average across scenarios comes out to $650K - fifty thousand more than her base case. Same market, same Volatility, opposite effect. The difference is the shape of each Operator's Profit as Demand changes. That shape is called convexity, and it determines whether Volatility works for you or against you. Most Operators never check which side they are on.
Convexity measures whether the shape of your Profit bends toward you or away from you as an uncertain input (like customer Demand) changes. A convex position gains more from upside surprises than it loses from downside ones - meaning Volatility actually helps your Expected Payoff. A position with negative convexity is the reverse: Volatility destroys value. Two bets can share an Expected Value yet behave completely differently under uncertainty because of their convexity.
Convexity describes the curvature of the relationship between an input you do not control (customer Demand, market conditions, project complexity) and the output you care about (Profit).
You already know Expected Value (the center of your outcomes) and Volatility (how spread out they are). Convexity is the third lens: it tells you whether that spread helps or hurts your actual Expected Payoff.
The core principle: when your Profit accelerates as conditions improve - gaining more per unit of good news than it loses per unit of bad news - you hold a convex position. The average Profit across uncertain scenarios is higher than the Profit at your base case forecast. Uncertainty in the input literally creates value.
When your Profit decelerates - losing more per unit of bad news than it gains per unit of good news - you hold a position with negative convexity. The average across scenarios is lower than the base case. Uncertainty destroys value even if your forecast is correct.
This is why two positions with identical Expected Value at base case produce very different real-world outcomes. The curvature of the relationship between Demand and Profit - not just the center and spread - determines how much you actually keep.
Convexity is why single-point Expected Value analysis systematically misprices the bets you actually face as an Operator. Two places it shows up on your P&L:
1. Capital Investment selection. Digital products and software investments with near-zero Cost Per Unit are naturally convex - your downside is bounded at the Implementation Cost, but the upside has no ceiling if the thing works at scale. A spreadsheet that models three scenarios (bad, base case, good) with symmetric ranges understates the true Expected Payoff of these bets. Every time you cap the upside at a 'reasonable' number in your model, you are hiding convexity. But not all Capital Investment is convex. A capacity-heavy buildout has bounded upside (physical ceiling on output) and substantial downside (the Implementation Cost is sunk plus Operating losses if Demand does not materialize). Know which type you hold before applying the convexity lens.
2. Cost Structure shape. Fixed Revenue plus Costs that accelerate under stress carries negative convexity - you cannot earn more than the contract price, but Labor and material cost can run away from you when Demand spikes (rush Hiring premiums, overtime, rising defect rate). Variable Revenue plus Fixed Costs is convex - you might earn nothing, but your downside is bounded while upside floats. The qualifier matters: if your Costs scale truly linearly with Demand, the position is linear, not curved. The curvature comes from non-linear cost behavior - Costs that accelerate faster than Revenue under pressure or resist downward adjustment when Demand falls. The ratio of Fixed vs Variable Costs on your Operating Statement is the starting point, but how those Costs behave under stress determines the actual curvature.
The mechanics are best shown with two Operators running businesses in the same market.
Operator A: negative convexity
The downside loss ($250K) exceeds the upside gain ($200K) because Costs accelerate upward (rush Hiring) but resist downward adjustment (salaried Labor stays on payroll). That asymmetry is negative convexity.
Operator B: convex position
The upside gain ($320K) exceeds the downside loss ($220K) because Revenue accelerates on the upside (Expansion Revenue compounds on a growing customer base) and decelerates on the downside (subscriptions retain value longer than hourly billing). That is convexity.
The general rule: Look at every Revenue Line on your Operating Statement and ask - does Profit improve more than proportionally when conditions get better, or does it cap out? Does Profit deteriorate more than proportionally when conditions worsen, or is the downside bounded? Capped upside plus accelerating downside is negative convexity. Bounded downside plus accelerating upside is convexity.
Run a convexity check whenever you face these situations:
The concrete procedure: Take the single input variable with the most uncertainty (customer Demand, project complexity, market growth). Vary it +20% and -20% from your forecast. Calculate Profit in each scenario. If the gain in the upside scenario exceeds the loss in the downside scenario, the position is convex. If the loss exceeds the gain, it carries negative convexity. The bigger the gap between gain and loss, the more curvature you carry.
1. You are choosing between investments with similar Expected Value. If your Sensitivity Analysis shows that the Expected Payoff changes asymmetrically as you vary assumptions, one option has more convexity. In uncertain environments, prefer the convex one - Volatility is your ally.
2. You are structuring a contract or Pricing model. Fixed-price contracts with variable delivery Costs carry negative convexity for the seller. Subscription Pricing or usage-based Pricing is more convex. When you see a Pricing decision, ask which side of the table absorbs the curvature.
3. You are building your annual Budget. If every Revenue Line on your P&L has a hard ceiling but your Cost lines have no floor, your entire Budget carries negative convexity. You will underperform your base case projections on average. Deliberately add convex positions - Expansion Revenue, Upsell motions, products with near-zero Cost Per Unit - so that good years overperform by more than bad years underperform.
4. You are sizing a bet. Convex bets justify larger Bet Sizing relative to their Expected Value because the curvature compensates for Tail Risk. Bets with negative convexity justify smaller Bet Sizing, even when the Expected Value looks attractive, because bad surprises compound harder than you model.
When convexity analysis does not add much: When Volatility is genuinely low (stable market, predictable Demand) or when you are operating well within capacity with no path to non-linear scale. The value of convexity scales with the amount of uncertainty in the environment. In a perfectly stable world, a linear position and a convex position with the same Expected Value really are equivalent.
You are deciding whether to build a product feature. Implementation Cost: $150K (three engineers for one quarter). Your forecast: 200 customers adopt it at $50/month = $120K ARR. You assign 80% probability to this base case and 20% probability to low adoption (20 customers, $12K ARR).
Expected ARR = 0.80 x $120K + 0.20 x $12K = $96K + $2.4K = $98.4K.
Year-one Profit = $98.4K - $150K = -$51.6K. The spreadsheet says do not build it.
The first model is not wrong - it is incomplete. It assumed 200 customers was a hard ceiling. The question you should ask: is it? This feature has near-zero Cost Per Unit after the build. Nothing structural stops it at 200.
Assign 5% probability to 2,000 customers ($1.2M ARR) and re-weight accordingly: 75% base ($120K) + 20% low ($12K) + 5% breakout ($1.2M). This is a different assumption set - you are modeling a scenario the first version excluded, not correcting an error.
New Expected ARR = 0.75 x $120K + 0.20 x $12K + 0.05 x $1.2M = $90K + $2.4K + $60K = $152.4K.
Year-one Expected Profit flips to +$2.4K, and you now own a compounding Capital Asset rather than a linear service engagement. The $60K contribution from a 5% scenario dominates because the position is convex - the tail scenario is 10x the base, not 1.3x. A small probability of a large win shifts the Expected Payoff because the downside is bounded at $150K while the upside is open.
Insight: Capping upside at a 'reasonable' number is the most common way Operators hide convexity from themselves. The defining feature of convex bets is that tail outcomes contribute disproportionately to Expected Value. Always ask: what is the realistic maximum outcome, and what probability do I assign to it?
You sign a contract to deliver a data migration for $200K. Your estimated Cost Structure: $140K (seven engineers at $20K each for the engagement). Expected Profit: $60K. You model three scenarios based on project complexity.
Revenue is fixed at $200K regardless of Execution speed. There is no upside if you finish early - the client will not pay more.
Best case (20% probability): project finishes at 70% of estimate. Cost = $98K. Profit = $102K.
Base case (50% probability): on Budget. Cost = $140K. Profit = $60K.
Worst case (30% probability): 40% over Budget due to undocumented data formats. Cost = $196K. Profit = $4K.
Expected Profit = 0.20 x $102K + 0.50 x $60K + 0.30 x $4K = $20.4K + $30K + $1.2K = $51.6K.
That is $8.4K less than the $60K base case. The asymmetry: your best scenario gains $42K over base, but your worst scenario loses $56K from base. More Volatility in project complexity pushes your Expected Profit further below $60K every time.
Insight: Fixed Revenue plus Costs that accelerate under stress equals negative convexity. Volatility in Execution literally destroys value in this structure. This is why experienced Operators either pad cost estimates heavily, include contract terms that reprice when complexity grows, or avoid fixed-price structures entirely. You are selling a position with negative convexity - price it accordingly or restructure it.
Convexity is the curvature of your Profit as inputs change. It determines whether Volatility helps or hurts your Expected Payoff. Convex means Volatility helps: the upside gain exceeds the downside loss. Negative convexity means Volatility hurts: the downside loss exceeds the upside gain.
The convexity test: vary your key uncertain input +/- 20%, calculate Profit in each scenario, and compare the absolute gain to the absolute loss. If gains exceed losses, you hold a convex position. If losses exceed gains, you carry negative convexity. The gap between the two is the magnitude of your curvature.
Audit your P&L for curvature: Fixed Revenue with Costs that accelerate under stress carries negative convexity (fragile under uncertainty). Variable Revenue with Fixed Costs is convex (benefits from uncertainty). Deliberately tilt your mix toward convex structures, especially when the environment is volatile.
Confusing convexity with simply 'high Volatility' or 'high risk.' A convex position can have high Volatility, but its shape means that Volatility works in your favor. A position with negative convexity and the same Volatility is far more dangerous. Shape and spread are independent - you need both to evaluate a bet.
Building financial models that cap upside scenarios at 'reasonable' numbers while leaving downside scenarios uncapped. This systematically hides convexity in Capital Investment bets and exaggerates the Expected Value of positions with negative convexity like fixed-price contracts. If your model's best case is 30% above base but your worst case is 50% below, check whether the real world actually has that shape - or if you just truncated the upside out of conservatism.
Classify each position as convex, negative convexity, or roughly linear for the Operator: (A) A monthly SaaS product with fixed infrastructure Costs and per-seat Pricing. (B) A staffing agency placing contractors at a fixed percentage of their hourly rate. (C) A Capital Investment to automate a manual process currently costing $300K/year in Labor.
Hint: For each one, ask: is the upside capped or open? Is the downside bounded or unbounded? Does the gain from a 20% Demand increase exceed the loss from a 20% Demand decrease?
(A) Convex. Infrastructure Cost is fixed (bounded downside), Revenue scales per seat (open upside), and each additional customer adds Revenue at near-zero incremental Cost - so Profit accelerates as Demand grows. (B) Roughly linear. Revenue and Costs both scale proportionally with contractor volume, producing a fixed marginal contribution per placement regardless of Demand level. (C) Convex. Downside is bounded at the Implementation Cost; upside is $300K/year in Cost Reduction that compounds every year the automation runs, and potentially more if it extends to adjacent processes.
Your P&L has two Revenue Lines: $3M from annual enterprise contracts (fixed price) and $1M from a usage-based product. Volatility in your market is increasing due to a competitive shift. Which Revenue Line becomes relatively more valuable, and how would you adjust your Capital Allocation between them?
Hint: Run the convexity check on each Revenue Line: vary customer Demand +/- 20%. Which line's Profit gains more from the upside than it loses from the downside?
The enterprise contracts carry negative convexity - Revenue is fixed but delivery Costs fluctuate with the competitive shift (clients may demand more support, talent Costs rise to retain engineers). Increasing Volatility erodes their Expected Profit below the base case. The usage-based product is convex - build Costs are sunk, and if the market shift sends new customers searching for alternatives, usage and Revenue can spike without proportional Cost increases. A 20% Demand increase might lift usage Revenue by 25% (Expansion Revenue and Upsell) while a 20% decrease might only cut it by 15% (existing usage patterns are sticky). Shift Capital Allocation toward the usage-based product: invest in features, Expansion Revenue pathways, and sharper customer segmentation to capture upside from the turbulence. Consider restructuring the next round of enterprise contracts to include usage-based Pricing components, converting some negative convexity into linear or convex exposure.
You are choosing between two Pricing structures for the same product in an uncertain market. Customer Demand could be anywhere from 500 to 1,500 units, with a base case forecast of 1,000 units.
Structure A: Fixed annual contract. A Buyer commits to $500K regardless of how many units they actually consume. Your Costs are $200K fixed plus $150 per unit of actual consumption (support, infrastructure).
Structure B: Per-unit Pricing at $600 per unit. Your Costs are $350K fixed (higher upfront investment in infrastructure) plus $50 per unit.
Calculate Profit at 500, 1,000, and 1,500 units for each structure. Which is convex? Which carries negative convexity? Which structure should you prefer if you expect high Volatility in Demand?
Hint: For each structure, compute Profit at all three Demand levels. Then check: does the gain from 1,000 to 1,500 exceed the loss from 1,000 to 500? That is the convexity test.
Structure A (fixed contract Revenue):
As Demand increases, Profit decreases because Revenue is fixed while Costs rise. The structure has a hard ceiling on upside (Revenue never exceeds $500K) while Costs are open-ended. At 2,000 units Profit hits $0. At 2,500 units the Operator loses money despite having a paying customer. Negative convexity accelerates beyond the modeled range.
Structure B (per-unit Pricing):
Every additional unit contributes $600 Revenue minus $50 Cost = $550 marginal contribution. The $350K Fixed Cost is already sunk. Below break-even (about 637 units), your loss is bounded - you cannot lose more than $350K even at zero units. Above break-even, every unit adds $550 to Profit with no ceiling. Bounded downside, open upside - the defining signature of convexity.
Under high Volatility, choose Structure B. If Demand could range from 500 to 1,500 or wider, Structure B's Expected Payoff improves as the range widens because gains in high-Demand scenarios grow faster than losses in low-Demand scenarios. Structure A's Expected Payoff deteriorates as the range widens because the Operator absorbs all the Cost Volatility without capturing any Revenue upside.
Convexity builds directly on your two prerequisites. Expected Value taught you to collapse uncertain outcomes into a single number - but that number implicitly assumes the relationship between the input and Profit is linear. Convexity is what happens when it is not: the average Profit across uncertain scenarios can be higher (convex) or lower (negative convexity) than the Profit at the base case input. A single-point Expected Value calculation - evaluating Profit at your base case forecast only - systematically underprices convex bets and overprices those with negative convexity. Expected Value calculated properly across weighted scenarios already captures curvature; the error is in point-estimate thinking, not in Expected Value methodology itself. Volatility taught you that spread around the mean matters - but it did not tell you whether that spread helps or hurts. Convexity answers exactly that question: positive curvature makes Volatility an Asset, negative curvature makes it a liability.
Looking ahead, convexity connects to several concepts you will encounter. Option Pricing is the formal math of convex positions - every options contract is a convex position by construction, and understanding why requires exactly the intuition you built here. Bet Sizing should account for curvature - convex bets justify larger Allocation than their base case Expected Value suggests, because the shape pays you back under uncertainty. Tail Risk becomes more intuitive once you see that negative convexity is where extreme outcomes do the most damage, while convex positions are where they create the most value. Capital Allocation is fundamentally about mixing convex and negative-convexity exposures to get the aggregate shape you want across your entire P&L. And Risk-Adjusted Return metrics like the Sharpe Ratio begin to break down for positions with significant convexity in either direction because they assume symmetric Returns - a limitation you will learn to identify and correct.
Disclaimer: This content is for educational and informational purposes only and does not constitute financial, investment, tax, or legal advice. It is not a recommendation to buy, sell, or hold any security or financial product. You should consult a qualified financial advisor, tax professional, or attorney before making financial decisions. Past performance is not indicative of future results. The author is not a registered investment advisor, broker-dealer, or financial planner.