Federal Sentencing Explorer
The Guidelines Are Not the Sentence: What 800,000 Federal Cases Actually Show
By Eric Rosen, Founding Partner, Dynamis LLP
Every federal defendant hears the same number first: the guideline range. A loss table, a criminal history score, a grid — and a range that sounds like a verdict. It is not. In fiscal year 2025, federal judges sentenced within the guideline range just 51.5% of the time. The other half of the time, the sentence landed somewhere else — usually below.
We built a tool that shows exactly where. The Dynamis Federal Sentencing Explorer puts twelve years of U.S. Sentencing Commission data — 805,875 sentenced defendants, fiscal years 2014 through 2025 — behind a set of filters any lawyer or client can use. Pick the offense, the final offense level, the criminal history category, and the district. The tool returns what actually happened to comparable defendants: medians, percentiles, probation rates, and how often judges went below the range on their own initiative.
In fraud cases, the range is where negotiation starts
Consider the last five fiscal years of fraud, theft, and embezzlement sentencings — 25,511 cases nationwide. The median sentence was 12 months. Nearly 30% of defendants received no prison time at all. And in 29.1% of cases, the judge imposed a below-range sentence without any government motion — no cooperation agreement, no 5K1.1 letter, just a judge concluding that the guideline number overstated the conduct.
Those are not anomalies. They are the operating reality of 18 U.S.C. § 3553(a), and they are precisely the numbers a sentencing memorandum should cite.
Where you are sentenced matters as much as what you did
The national numbers conceal enormous district-level variation. Among districts with at least 200 fraud sentencings since FY2021, the median sentence runs from 3 months in the Eastern District of New York to 24 months in the Southern District of Florida and the Northern District of Georgia. Same guidelines. Same offense conduct categories. An eight-fold difference in outcomes.
Judges know their own districts. A sentencing memo that shows the court its district's actual record — alongside the national picture — speaks in the court's own language.
Cooperators are counted separately, because they should be
Averages lie when the pool is mixed. Roughly one in seven fraud defendants over the past five years received a §5K1.1 substantial-assistance departure; in securities fraud cases, more than one in five did. Blend those cases into your comparator pool and every benchmark drops — which is exactly the argument a prosecutor would make against your numbers.
The Explorer codes substantial-assistance cases separately. One checkbox removes them, leaving a clean pool: defendants who did not cooperate, sentenced for comparable conduct, at comparable offense levels. When we tell a court that the median non-cooperating securities fraud defendant received a below-guideline sentence, the data behind that sentence is filtered the way an honest comparison requires.
Numbers do not replace advocacy. They arm it.
A sentencing decision is a human judgment about a human being. No dataset changes that. But every judge who imposes a sentence does so knowing what other judges have done — and a defendant whose lawyer knows those numbers too negotiates from higher ground, sets realistic expectations, and gives the court a principled basis to impose the sentence the case deserves.
The data comes from the Sentencing Commission's public Individual Offender Datafiles, harmonized across twelve fiscal years, including the Commission's economic-crime files that break fraud cases into subtypes — securities, healthcare, government benefits, embezzlement, and more. The tool is free, and it runs in your browser: sentencing.dynamisllp.com.
If you or a client is facing a federal sentencing, the numbers are only the beginning of the story. Contact Dynamis LLP — we have told that story to judges across the country.
Dynamis LLP is a litigation boutique focused on white-collar criminal and regulatory defense, with offices in Boston, New York, Miami, New Jersey, and Los Angeles. This post is attorney advertising and does not constitute legal advice; historical sentencing statistics are not a prediction of any outcome.
