This is not a claim that Utah's roster, schedule or quarterback situation can be copied onto Michigan. It is a descriptive audit of two 2025 offenses using the same SOAR classifiers. The useful question is narrower: where did the raw efficiency gap expand, and are those areas relevant to what Jason Beck inherits in Ann Arbor?
Rush/dropback families come from SOAR's locked play classifications, not film-charted intent. Scrambles, RPO decisions, exact distance, defensive front and game state can all affect the observed split. The matrix below is a review map, not a causal grade.
The offense wasn't bad everywhere.
Michigan's base efficiency was close to Utah's. The separation grew as the offense moved into situations where the defense had more information and the field compressed.
That distinction matters for 2026. If Beck preserves Michigan's rushing floor and improves the offense when it falls behind schedule or reaches the red zone, the unit can change substantially without becoming pass-heavy.
Use the matrix as a map, not a verdict.
Every regular-season offensive snap is bucketed by down and distance, then split into SOAR's rush and dropback families. Short = 1–3 yards, medium = 4–7, long = 8+. A cell only highlights an observed family edge when both sides have at least 15 plays and the success-rate difference is at least eight percentage points.
Michigan · Lindsey (12 games)
ACTUAL · 2025 REGULAR SEASONUtah · Beck (12 games)
ACTUAL · 2025 REGULAR SEASONThe 20-point observed split is worth reviewing on film. It is not evidence that Michigan “should have run” every medium late down: 3rd and 4th down are grouped in the current artifact, and exact distance, scramble outcomes and defensive structure are not controlled here.
Did the offense actually improve?
The maize line is overall Success Rate v1. Blue dots are passing-down success. The Utah benchmark is context, not a target line: Michigan faced a different schedule, and single-game situational samples can swing hard.
Michigan's best overall success game. Underwood also produced his biggest rushing day.
The floor against an elite defense. Raw success fell almost 39 points from the CMU peak.
Casula's interim game is visually separated because it is not Lindsey's playcalling sample.
Beck does not need Michigan to become pass-happy.
Michigan and Utah both finished the 2025 regular season at essentially a 58.5% rush share. The intriguing part of Beck's Utah offense was not raw pass volume. It was how much more productive the offense remained in leverage situations.
Utah +6.1 percentage points.
PPA/play: +0.26 Michigan · +1.06 Utah.
Research split: 2nd-and-8+ / 3rd-and-5+ / 4th-and-5+.
Beck's 2025 Utah quarterbacks were real run threats, which changed the arithmetic for the defense and helped create throws off the same run-first structure. Underwood is not Devon Dampier. Michigan's own quarterback coaches have described him as more naturally pocket-oriented, even though his freshman season showed legitimate rushing ability.
Beck has publicly framed the install around conforming the system to Underwood. That is the right standard for judging the hire: adaptation, not imitation.
Three reasons not to overfit Utah onto Michigan.
Different quarterback geometry
Utah's system forced defenses to account for Dampier as a frequent designed-run threat. Underwood can run, but his default style is different. The same concepts may create different defensive reactions.
Different personnel answers
Beck's ‘positionless’ Utah offense found unusual roles for that specific roster. Michigan needs its own mismatches; those cannot be assumed from Utah's box score.
Different opponents, different game states
None of the Michigan-vs-Utah raw splits are opponent-adjusted here. They are diagnostic clues. They are not proof that the same calls would produce the same result in the Big Ten.
Michigan already had the snap-to-snap foundation. Beck's Utah data is compelling because its biggest advantage appeared after the easy answers disappeared. That is exactly where Michigan has room to grow — and exactly where 2026 has to prove the comparison travels.