How are peer benchmarks computed?
Methodology overview: cohort definition (city × industry × age), nightly computation, sample-size thresholds, privacy guarantees. Full statistical detail at /methodology.
Keeping Up's defining feature is showing you how you compare to your peers — not the generic US average. Here's how that works.
Cohort definition: when you sign up, you tell us three things — your city (metro area), industry, and age range. We use these to assign you to a cohort — a group of users with the same combination. For example: NYC × Finance × 30-37. There are over 2,000 distinct cohorts.
What gets compared: your net worth, savings rate, monthly spending (by category), and a few derived metrics like "months of emergency fund." The comparison is always against your specific cohort, not a generic average.
What's never compared: your name, email, exact dollar figures of friends or strangers, individual transactions, or anything that could identify you or another specific person. Everything is aggregated.
How the math works: we compute percentile distributions (p10, p25, p50, p75, p90) per cohort nightly. When you load your dashboard, we look up your cohort's distribution and show you where you fall — "Top 12% of NYC Finance 30-37 peers" means your net worth is higher than 88% of other users in that same cohort.
Data quality: small cohorts (under 30 users) show a "confidence indicator" — we still display the comparison but mark it as approximate. As more users join, the cohort distributions tighten and we remove the approximate marker.
Full methodology: for the deep dive — cohort enumeration, sample-size thresholds, outlier handling, statistical methods, and auditability — see /methodology. The Privacy Policy covers data handling.