The most defensible revenue most companies have is not the product they sell. It's the exhaust that product throws off while it runs — the payment data, the audience, the internal tooling, the trust — a resource that accumulates only because you are doing the main thing, at near-zero marginal cost to you and high replication cost to everyone else. The product earns the right to exist. The byproduct earns the margin.
This is not a metaphor about "adding a data play." It is a specific claim about where durable economics come from, and most founders walk past it every day because the byproduct doesn't show up on the invoice. You priced and sold the product. Nobody priced the exhaust, so nobody looked at it, so it compounds silently in a corner until a competitor — or you, on a good day — notices it was the real asset all along.
What actually counts as a byproduct
Three conditions have to hold, and the third is the whole game.
First, the resource accumulates as a side effect of an activity you would run anyway. You are not spending to create it; it falls out of doing the main thing. Second, the marginal cost of producing one more unit of it is close to zero, because the core activity already paid for that unit. Third — and this is the one that separates a moat from a slide deck — a competitor can only reproduce the resource by reproducing the activity. They cannot buy it as a shortcut.
That third property is what makes a byproduct defensible in a way a feature never is. Your cost to acquire it is a sunk fraction of operating cost you were paying regardless. A competitor's cost to acquire it is the full operating cost of running your business, for as long as you have run it. The byproduct's price is quoted in their currency, and their currency is time they don't have. A feature can be cloned in a quarter. Six years of accumulated payment behavior cannot be cloned at all — it can only be lived through.
That asymmetry is the reason the exhaust so often out-earns the engine. Same activity, two revenue streams: the thin margin on the product, and the fat margin on a resource whose input cost was already expensed to the product's P&L.
The arithmetic, on cash-on-delivery logistics
Take the business I know from the inside. Running Kommerce, a commerce OS for cash-on-delivery markets, means operating the least glamorous thing imaginable: physical delivery of goods that get paid for in cash at the door. The unit economics of that activity are brutal and everybody knows it.
Illustratively — round numbers to show the shape, not audited figures — deliver a $30 order and you might charge the merchant a $2 fulfillment fee against $1.50 of real cost. Fifty cents of margin on thirty dollars of goods: 1.7% on GMV, in a market where a meaningful share of COD orders are refused at the door and shipped back at a loss. As a standalone business, low-trust logistics is a knife fight for pennies.
Now look at what the activity produces for free. Every delivery attempt is a costly signal — the buyer either put cash in a courier's hand or didn't. Multiply that by every order, every neighborhood, every phone number, every merchant, over years. What accumulates is a map of who actually pays: default propensity by area, by merchant, by order pattern, by hour. No survey produces this. No data vendor sells it. It exists only as residue of having run the deliveries.
That residue underwrites a second business the logistics can't touch on margin. Once you can price default risk, you can offer a merchant guaranteed remittance for a risk-priced premium, or a cash advance against receivables you can actually assess, or insurance on the shipment itself. The product side of that — lending, guarantees — carries no cost of goods beyond the cost of capital. The expensive input, the data, was already paid for by the delivery you were running anyway.
So the same $30 delivery now throws off two revenues. The fifty cents of logistics margin, and a risk premium you can price because you know a default curve your competitors can only guess at. A rival can undercut your delivery fee in an afternoon. To undercut your underwriting, they would first have to run your delivery volume for years to earn the same map. The logistics is the product. The default map is the byproduct. The byproduct is the business.
Trust is the second layer of the same exhaust. Rails you built to move cash reliably — reconciliation, courier reliability scoring, remittance guarantees — become infrastructure other merchants will pay to plug into. You built the trust to sell your thing; it turns out to be worth more as a platform others rent.
The pattern repeats across categories
This isn't a fintech quirk. It is a shape, and once you see it you see it everywhere:
- Internal tooling becomes the product. The infrastructure a team builds to run its own business — because nothing off the shelf worked at its scale — is, by construction, hardened at a scale nobody else has reached. A large fraction of the developer-tools and cloud-infrastructure industry is this exact move: the byproduct of operating at scale, sold to everyone who wants to operate at scale.
- The audience out-values the product. Companies build an audience to sell product A and discover the audience is worth more than A ever was. The product becomes the reason the audience assembles; the audience becomes the asset.
- Transaction exhaust becomes underwriting, insurance, or benchmarking. Anyone sitting on a proprietary flow of transactions can turn the aggregate into a business the individual transactions can't fund — lending, fraud scoring, the industry benchmark everyone else has to buy back from you.
- Aggregated demand becomes leverage. Pool enough buyers doing one thing and the pooled demand becomes a lever over suppliers, a group-buying business, a marketplace — a resource none of the individual buyers had alone.
Real byproducts versus wishful ones
Here is where most people ruin the idea. They hear "monetize your data exhaust," look at their generic web analytics, and imagine a business that isn't there. So make the distinction operational. Score every candidate on two axes, 0 to 3.
Proprietary — do only you accumulate this, or does everyone in your category get the same thing for free? Replication cost — how long, and how much, for a well-funded competitor to reproduce it from scratch?
| Byproduct candidate | Proprietary | Hard to replicate | Verdict |
|---|---|---|---|
| Years of your own default/payment outcomes | High | High | Real — underwrite on it |
| Internal tool proven at a scale others can't reach | High | High | Real — productize it |
| Generic web analytics anyone can buy | Low | Low | Wishful |
| Audience you rent on someone else's platform | Low | Medium | Fragile — you don't own it |
| Aggregated demand only you have pooled | High | Medium | Real — press it |
Only the high-high quadrant earns the word "byproduct" in the strategic sense. Everything else is either a commodity you happen to hold or an asset you are renting and could lose in one algorithm change. The audience you built on your own rails is a byproduct. The audience you built on a platform that can deplatform you is a liability wearing a byproduct's clothes.
Why this compounds when everything else decays
I've argued elsewhere that your competitive advantage has a half-life — that most edges lose a roughly constant fraction of their remaining value each period, which is why the collapse always feels sudden when it was actually on schedule. Byproducts are the rare exception, and the reason is worth stating precisely.
A feature decays because imitation erodes it and nothing replenishes it. A proprietary data byproduct decays too — last year's payment behavior slowly ages out of relevance — but it is replenished by the same activity that produced it. As long as your accumulation rate exceeds the decay rate, the stock grows and gets more predictive, not less. The half-life lengthens instead of shortening. That is the only kind of advantage that reliably widens while you sleep, and it widens fastest precisely when the core activity is running hardest.
Which reframes the audit question. It is not merely "what does my activity produce for free." It is "what does it produce that compounds faster than it decays." Those are the byproducts worth building a second business on.
The move: commoditize the product to maximize the exhaust
There is a version of this strategy that feels insane until the math clears. Joel Spolsky's rule — commoditize your complement — says drive the price of whatever is complementary to your product toward zero, so demand and margin flow to you. The byproduct lens adds a twist Spolsky didn't: sometimes the thing to commoditize is your own product. Run it at cost. Occasionally give it away. Not out of generosity, but because the product's real job is to be the cheapest possible acquisition channel for the exhaust — and the exhaust is where both the margin and the moat live.
When the product is where you make money, you ration it. When the product is how you manufacture the byproduct, you want maximum throughput, because every unit sold is another unit of proprietary resource accumulated at a competitor's unreachable cost. The product stops being the profit center and becomes the data-acquisition cost of the business hiding behind it.
So do the audit this week. Take one sheet of paper and five rows — data, audience, tooling, trust, aggregated demand. For each, write in plain words what your core activity produces for free that someone else would pay for or simply cannot get. Score it: proprietary, hard to replicate. Circle anything high on both. Then ask the uncomfortable question about the one you circled — not "how do we add a revenue stream," but whether that byproduct is quietly a better business than the product you built to generate it.
Most founders spend their lives defending the engine. The margin was in the smoke.