IMU Sensor Placement: Best Practices for Reliable Movement Data

A strength coach at a college weight room once ran the same countermovement jump test twice in one session, once with the sensor clipped to a waist belt and once tucked into a compression short at the sacrum. The jump height numbers differed by almost 3cm on the same athlete, same jump, ninety seconds apart. Nothing about the athlete's movement changed. What changed was the sensor's distance from the body's center of mass and how much it moved relative to the pelvis during flight. That gap is the entire subject of this guide: placement is not a footnote in the manual, it is one of the largest controllable sources of error in any IMU-based testing program.

Picerno (2017) reviewed 25 years of inertial sensor research for lower-limb kinematics and concluded that sensor-to-segment misalignment and inconsistent mounting were the dominant sources of error across studies, often larger than the sensor's own electronic noise. Camomilla, Bergamini, Fantozzi, and Vannozzi (2018), in a systematic review of wearable inertial sensors in sport, made a similar point from a different angle: comparability across sessions and across athletes collapses when mounting site, strap tension, and orientation are not standardized, even when the underlying hardware is identical. Neither paper claims placement is the only variable that matters, and both note that soft tissue artifact interacts with body composition and clothing in ways that are hard to fully eliminate. What follows is a practical translation of that research into placement choices you can actually implement on a Tuesday afternoon with fifteen athletes waiting.

An IMU reports acceleration and angular velocity at the exact point it is attached, not at the athlete's center of mass. Every centimeter between the sensor and the segment you actually care about introduces a lever arm, and every gram of soft tissue between the sensor and bone introduces a spring. Both distort the signal in ways that are predictable in direction but variable in magnitude between athletes.

Two mechanisms do most of the damage. The first is soft tissue artifact: skin and muscle move relative to the underlying skeleton during high-acceleration movements, so a sensor strapped over the gastrocnemius wobbles independently of the tibia it is meant to represent. The effect is worst over muscle bellies and smallest over bony prominences like the sacrum, sternum, or distal tibia. The second is orientation drift relative to the axis of interest: if a sensor is mounted even 10-15 degrees off the sagittal plane, vertical acceleration during a jump partially aliases into the mediolateral axis, and jump height calculated from a single axis will read low.

Neither mechanism is exotic or hard to control once you know it exists. The problem in most facilities is not physics, it is habit. Whoever set up the sensor mounts three years ago picked a spot that seemed convenient, and nobody has re-examined it since. The next section gives metric-specific starting points rather than a single universal answer, because there is no single correct mounting site for an IMU.

The right mount depends on what you are trying to measure, not on what is most convenient to strap on. The table below reflects common practice in applied sport science and the general logic from Picerno's review: mount as close as possible to the segment of interest, over bone rather than muscle where the movement allows it.

Note the pattern: bone over muscle, close to the segment of interest, and consistent orientation relative to an anatomical plane. When a movement makes the ideal site impractical, for example a wrist mount during Olympic lifts because a sacral sensor cannot see the pull phase, document the substitution and treat cross-comparisons with sacral-mounted data as unreliable rather than pretending the two datasets are equivalent.

Consistency matters more than perfection. An imperfect mount applied identically every session produces trustworthy trend data; a theoretically perfect mount applied inconsistently produces noise dressed up as signal. Use this five-step protocol for jump and general movement testing, then adapt the anatomical landmarks for other metrics using the table above.

Step 1: Locate the landmark by palpation, not by eye. For sacral placement, find the posterior superior iliac spines and center the sensor on the midline between them. Mark the spot with a skin-safe pen for repeat sessions with the same athlete during a training block.

Step 2: Mount directly on skin or a single thin compression layer. Loose clothing between sensor and body adds an independent oscillating mass. If skin contact is not practical, a snug compression garment is the next-best option; a baggy shirt is not.

Step 3: Align the primary axis before strapping down. Most IMU housings mark their vertical or long axis on the case. Align this mark to the sagittal plane using a plumb line or the wall as a vertical reference, then tighten. Recheck alignment after the strap is fully tightened, since tension can rotate the housing slightly.

Step 4: Apply consistent strap tension. Too loose and the sensor bounces independently of the body; too tight and circulation complaints during a 90-minute testing session will end your protocol early. A useful field check is two fingers sliding under the strap with light resistance.

Step 5: Record the exact mount in the athlete's file. Photo or written note, landmark used, strap tension setting, and clothing layer. Six months later when someone asks why an athlete's numbers shifted, this record separates a real physiological change from a mounting change. For the statistical side of separating signal from mounting noise, the IMU validation for coaching guide covers the reliability testing that should follow a placement change.

Most placement errors are invisible in the moment because the sensor still produces a number. The number is just wrong, or wrong in a way that only shows up months later as a confusing trend.

The most common mistake is switching mount location mid-season without flagging it. A team that starts the season with waist-belt sensors and switches to compression-short pockets in week eight will see what looks like a real jump-height improvement across the whole roster. It is not athletic improvement, it is a mounting artifact, and coaches who do not catch it make training decisions based on noise.

The second is inconsistent orientation between athletes on the same team. If one athlete's sensor sits rotated 20 degrees relative to another's, the two are not directly comparable even with an identical mount location, and team-wide leaderboards built on this data are misleading.

The third is mounting over muscle bulk rather than bone when bone access is available. This shows up most often with shank-mounted sensors during running assessments, where sensors placed over the belly of the tibialis anterior report substantially noisier signal than sensors placed on the anteromedial tibial crest a few centimeters away.

The fourth is failing to re-secure the sensor between warmup and the actual test set. An athlete who adjusts a waistband or retucks a shirt mid-session can shift the sensor without anyone noticing, and the pre/post comparison built on that session is comparing two different mounting conditions rather than two different physiological states.

The fifth, and the one that surprises people most, is assuming a single mount works for both jump testing and sprint testing. A sacral mount that is excellent for vertical displacement during a jump is a poor choice for horizontal acceleration profiling during a sprint, where a sternal or upper-back mount better tracks trunk lean angle. Pulling numbers from one protocol's mount location into another protocol's normative table produces comparisons that look plausible and are not.

For a squad of 20-plus athletes tested weekly, individual judgment calls about mounting do not scale. Write a one-page standard operating procedure with a photo of the correct mount for each test type, the exact landmark description, and the strap tension check. Post it at the testing station, not in a shared drive nobody opens before practice.

Assign mounting to a small, consistent group of staff rather than rotating the job among all assistant coaches and interns. Camomilla et al. (2018) noted that inter-rater variability in sensor application is itself a source of error distinct from the sensor's technical accuracy, and this is the practical lever coaches can pull without buying new equipment.

Run a quick monthly spot check: have two staff members independently mount a sensor on the same athlete using the written SOP and compare the resulting jump-height or velocity reading. Agreement within your program's established error band (see the IMU validation guide for how to establish that band) confirms the SOP is being followed. A gap wider than expected usually traces back to a landmark being interpreted differently by different staff, not a hardware problem. Pairing this with the interpretation framework in IMU data interpretation for coaches and, where a lab reference exists, cross-checking against IMU vs. linear position transducer data closes the loop between mounting discipline and trustworthy decisions.