Coherence: spending with no life in it.

If you print out three years of one real household's transactions and read them like a diary, you get a strange, boring, human document. Coffee on Tuesday. A grocery run that is never the same amount twice. A utility bill that recurs on roughly the same day every month. A parking ticket. A refund. A birthday splurge. Dozens of merchants, most of them small, most of the amounts ending in odd cents. It does not optimize toward anything. It is just what a life spends money on.

Coherence is the signal that reads for that texture, and notices when it is missing.

What real spending looks like

Three things show up in genuine spending, and they show up together:

• Breadth. Real people transact across a wide, uneven set of merchants and categories. Groceries, fuel, restaurants, subscriptions, medical, retail, the odd one-off. The distribution is lumpy and long-tailed.
• Recurrence. Some of that spending recurs like an obligation, because it is one. Rent or mortgage, a phone bill, insurance, a streaming subscription, a gym. These land on a cadence and carry across months.
• Irregularity. The amounts are messy. A real grocery bill is almost never the same twice and almost never a round number; it lands at odd cents because it is the sum of whatever happened to be in the cart. Very little of it is clean, and almost none of it is designed to look like anything.

The important part is that these are hard to fake at the same time. You can stage breadth, or you can stage recurrence, but staging both, with genuine irregularity, for thirty-six months, is expensive and pointless for someone whose only goal is to look creditworthy long enough to draw down a limit.

What staged spending looks like

A cultivated synthetic's spending is optimized for a single outcome: keep the account in good standing and its score rising until the limit is worth taking. Everything not serving that goal gets left out. So the transaction history tends to be:

• Thin. Few categories, few merchants, not much of the incidental long tail that a real life generates without trying.
• Non-recurring. Very little that behaves like a real obligation. There is activity, but not many of the sticky monthly commitments a real household accumulates, because the account is not actually living anywhere or subscribed to anything it needs.
• Round. A telling share of clean, round-dollar amounts. Activity that was entered to produce a record, rather than spent to buy a thing, drifts toward round numbers, because the person creating it is thinking in round numbers.

It reads like activity staged to look like life, because that is what it is. The purpose of the spending is the record it leaves, not the goods it buys.

Why 36 months is the right window

Coherence is faint over a few weeks and loud over a few years. In a short window, a thin, optimized history and a quiet real member look similar. Over three years, a real account cannot help but accumulate texture: a move, a new job, a broken appliance, a subscription started and cancelled, a holiday. The absence of any of that, across a span where a real life would have generated plenty, is itself the signal. You are not looking for a single wrong transaction. You are looking at the shape of the whole record and asking whether a life was actually lived through this account.

Reading it against the account, not the population

Some real people genuinely spend very little through a given account, keeping it for one purpose while living out of another. That is why coherence, like every signal here, is read against the account's own history and alongside the others, not as a solo verdict. A frugal or single-purpose account is thin. A synthetic is thin and arrived without provenance and then lets its appetite outrun its income. Coherence is one line in that paragraph, not the whole sentence. The next note takes up the appetite.