Essay ยท August 2026
The Revenue That Never Existed
Churn books. Fraud books. The customer who gave up at the sign-in page books nowhere.
Ask your finance team where lost revenue shows up and they will show you the accounts. Churn has a number. Refunds have a number. Fraud has a number and a whole department dedicated to it.
One loss never gets recorded. The customer who tried to sign in and gave up.
I wrote about why they give up in "You Cannot Cheat Physics". TLDR: identity that depends on a central server pays a network round trip on every sign-in, the trip runs slowest or fails outright where the customer is hardest to reach, and no upgrade to the network removes the dependency. That essay was about the physics. This one is about the ledger.
Start with what a sign-in actually gates. Not the app. The app opens fine. The sign-in gates every transaction that needs to know who the customer is before value can move. A payment tied to an account, benefit, offer, or loyalty balance needs an identified payer. A loyalty point spends only inside an authenticated session. A partner offer needs consent, and consent needs an identified person saying yes, on the record. Identification sits upstream of all of it. When it fails at the door, nothing downstream fails. Nothing downstream happens at all.
That distinction is the whole problem. A failed payment is an event. It gets logged, retried, escalated, reported to someone whose job is to care. A payment that was never attempted, because the customer never got past a spinning sign-in page, is not an event anywhere in your stack. No log line. No alert. If your funnel report starts counting after sign-in succeeds, the loss is not even in the funnel. It books nowhere, because the revenue never existed to be lost.
Loyalty currencies make it concrete. Years of work go into making points spendable like money: coffee, rides, hotel rooms, upgrades. And every point spent needs a live session, because the balance lives on a server and the server has to know who is asking. So the currency works at home on the couch and fails in exactly the places the customer is surrounded by things to spend it on. The aircraft cabin. The foreign terminal. The partner venue on saturated WiFi. A currency you cannot spend where you are standing is a smaller currency than the one on the slide.
Partnerships price this in whether anyone says so or not. A deal built on consented customer data is worth what the reachable, identified audience is worth. The model assumes one reachable audience size. Identity friction delivers a smaller one: every customer who cannot be identified at the moment the offer lands sits outside the deal.
Now put the two halves side by side. The cost side of identity is fully visible. Clusters, regions, failover drills, replication, residency compliance in every jurisdiction you copy identity data into. Multiple everything everywhere. That bill gets approved line by line and grows with each new geography. The ledger shows serious money going into identity and nothing at all coming out as identity losses, and a reasonable executive reading it concludes the thing is expensive but working. Half of that is true.
The fix is not more funnel optimization, because the customers this loses never enter the funnel. The fix is making identification stop depending on a faraway server saying yes. The mechanism is the other essay's subject, so one sentence here: a credential on the customer's device carries its own proof, and the check happens where the customer is standing, in milliseconds, with nothing traveling. And proof of who is only half of what travels. The claims that matter ride in the credential too: the tier, the entitlement, the consent on record. No live lookup for who the customer is, and no live lookup for the facts the transaction needs. What matters for the ledger is the consequence. Identification becomes a property of the customer being present, not a network event that may or may not complete. Wherever the customer is becomes a place they can be identified. Which makes it a place they can pay, spend a point, accept an offer, say yes.
The first answer most teams reach for is passkeys, or an authenticator app. Both are real upgrades over passwords. Neither closes this gap. A passkey still signs the customer in to a server, the server still has to be reachable to say yes, and every fact the transaction needs still sits on it, waiting to be looked up. An authenticator app adds a second check, but the server still has to be reachable to verify it. Stronger locks on the same faraway door. Why none of this travels is its own article.
I will not promise a number. Distrust anyone who does, because the defining feature of this loss is that nobody measured it. But its shadow shows up in data you already hold, and two questions surface it. Where do sessions die before identity is established? Where do customers spend their time, versus where authenticated value can actually move?
You cannot recover revenue that never existed. You can stop foreclosing it.