The Outlier Lab

Customer Tenure and History

How long and how well an account has behaved (age, settled funding, clean outcomes), used to size exposure and exploited by farmed and bust-out accounts that manufacture it.

What it measures

"Tenure" is shorthand for several distinct features that are often collapsed into one, unhelpfully:

FeatureWhy it is different
Calendar ageCheapest to fake; an account can simply be created early and left idle
Active tenureDays or months with real activity; separates aged-but-dormant from used
Settled valueDollars of funding that cleared past the return/dispute window; this is what actually earns exposure
Clean-outcome historyPrior payouts, redemptions or claims that matured without chargeback, return, clawback or confirmed fraud
Relationship breadthDirect deposit, multiple products, recurring bill pay; costly for a fraud actor to fake at scale
Time since last trust resetDays since the last sensitive account change, recovery or dormancy break

Why it matters

Almost every exposure decision is implicitly a function of history: how much instant buying power to extend, whether to hold a check, how much cashback to pay out before a receipt is verified, whether to release a marketplace seller's funds before delivery. Regulation embeds the idea. Reg CC treats an account as "new" for its first 30 calendar days and allows extended holds on check deposits above $6,725 during that period, and Reg E extends error-resolution timelines for transfers within 30 days of an account's first deposit.

The more your controls reward history, the more adversaries invest in manufacturing it. Account Farming and Resale sells aged, verified accounts precisely because they qualify for higher limits; Bust-Out Fraud builds good standing deliberately before maxing out; Synthetic Identity Fraud identities are cultivated for months. Tenure works as a cost lever: it raises the price of fraud without proving good intent.

By industry: in Cashback, Coupons and Rewards, tenure gates payout thresholds and offer eligibility, and aged accounts with a sudden burst of high-value receipts are a known farmed-account pattern; in Neobanks and Consumer Fintech, tenure and direct-deposit history drive instant-deposit and overdraft limits; in Brokerage and Investing, tenure affects instant buying power and ACATS review; in Online Marketplaces, seller history governs payout timing and reserves.

How to collect it

  • Precompute features in a feature store with clear as-of semantics so models don't learn from future information (see Fraud Data Architecture).
  • Track settled rather than initiated funding; only funding past its realistic return window (ACH return windows, check return timelines, card dispute windows) should count toward earned trust.
  • Store prior outcomes per linked entity (bank, device, identity) as well as per account. A new account on a long-trusted device and identity is not really new. Outcome quality depends on label maturity windows (see Fraud Labeling and Feedback Loops).
  • This is almost always built in-house; consortium data (Credit Bureau and Consortium Signals) can add history from outside your platform.

How to use it

  • Exposure sizing: tie limits to settled value and clean outcomes, not calendar age (see Velocity and Exposure Limits and Funds Availability Holds).
  • Change detection rather than level: compare current behavior to the account's own history. A sudden step-change in funding size, payout destinations or redemption value after a long quiet period is more informative than tenure itself. Combine with Transaction Pattern Anomalies.
  • Dormancy breaks: long-dormant accounts that reactivate and move value quickly are red flags in FinCEN's pig butchering and elder exploitation guidance, and a common ATO and mule-recruitment pattern.
  • Trust resets: after a recovery or contact change, reduce effective tenure until the account re-establishes history under the new configuration.
  • Combine with linkage: Entity Link Analysis answers "new account, but old person?" and "old account, but new controller?"

Strengths and limitations

Strengths: strongly predictive at the population level; inexpensive; intuitive for reviewers; aligns with regulatory concepts of "new account".

Limitations: can be cultivated or purchased; it says nothing about whether the current user is the owner (Account Takeover (ATO) hits the oldest, richest accounts hardest); calendar-age rules create predictable cliffs. Compensate with gradual, value-based trust accrual, continuous re-evaluation, and links to other entities.

False positives and fairness

  • New customers are mostly good customers; tenure-heavy models impose friction on growth cohorts and newcomers to the country with no history elsewhere.
  • Young adults, recent immigrants and people returning to banking after a gap all look "thin".
  • Seasonal users (tax refunds, holiday shoppers, annual traders) look dormant then suddenly active.
  • Small businesses and side-hustle sellers often scale quickly in a legitimate way.

Monitor good-customer friction by cohort and use soft controls (holds, gradual limits) rather than denials for thin-file customers (see Friction and Customer Experience Tradeoffs).

Privacy and compliance considerations

History-based limits and holds must follow disclosed policies; Reg CC exception holds require notices. Avoid using age of the customer (as opposed to age of the account) as a feature without fair lending/ECOA review where credit is involved. See Privacy and Data Use in Fraud Prevention.

Mechanisms it helps detect

Bust-Out Fraud · Account Farming and Resale · Instant Funding Abuse (Kiting) · Sign-Up Bonus Abuse · Money Mules (new accounts with immediate high pass-through) · First-Party Fraud

Funding Source Signals · Dispute and Return History · Fast In Fast Out Patterns · Velocity Signals · Reward Vesting and Clawback · Playbook - Bust-Out Detection Members · Playbook - Instant Funding Abuse Members

Sources and further reading