Trainers have opinions about movement. They have to — the daily work of improving how a horse carries a rider depends on reading the animal in real time, adjusting, and reading it again. Most trainers also, after enough years, develop strong intuitions about which parts of a horse's movement they trust their eye on and which they do not.

This post is written for the trainer who has started using biomechanics data and wants to know what to actually do with it. Not the marketing version. The working version.

What the data is good at

Three things, mainly.

Detecting drift. If you have a stable baseline for a horse — six weeks of consistent readings — the platform will flag changes before you can feel them from the saddle or see them from the rail. The most useful flags are stance-time asymmetry (the horse standing milliseconds longer on one diagonal), fetlock load asymmetry (one limb absorbing more impact than the other), and head-nod amplitude (the classic forelimb offloading signal, measured in vertical millimetres rather than inferred from feel).

Quantifying progression. When you introduce a new exercise — lateral work, pole work, hill work, interval training — biomechanics gives you a before-and-after on the specific metric you were trying to improve. If you added shoulder-in to develop hind-end engagement, you should see hindlimb stance time lengthen and pelvic asymmetry reduce. If the numbers do not move, the exercise is not doing what you hoped.

Making the rider conversation sharper. This is the underrated use case. A rider tells you the horse feels "a bit sticky on the left rein." You have two choices. You can agree, adjust the programme, and hope. Or you can pull up the last ten readings, show the rider that left-lead canter has a twelve-percent shorter suspension phase than right-lead, and have a concrete thing to work on. The second conversation ends differently.

What the data is bad at

Judging overall quality. A horse can have numerically clean biomechanics and still be a dull mover, and a horse can have interesting data and still feel rough under saddle. The data is about symmetry, load distribution, and change — not aesthetics. Judges and trainers evaluate quality; the platform evaluates structure.

Attributing cause. An offloading flag on the right fore does not tell you why. It does not distinguish between a sore hoof, a suspensory strain, a poorly fitting shoe, and a horse that is uncomfortable in a new arena. Cause attribution is diagnostic work — yours, your farrier's, your vet's. The data is the symptom, not the diagnosis.

Predicting future performance. A horse with clean biomechanics today may not stay clean tomorrow, and a horse with asymmetry today may be stable at that asymmetry for a decade. The platform tells you what the horse is doing now and how that compares to its own history. It does not forecast.

Baselines that are actually useful

The single most common way to get poor data out of a biomechanics platform is to give it inconsistent inputs. A baseline built from clips on five different surfaces, with three different riders, at unknown warm-up states, is a baseline with most of its sensitivity eaten by noise.

A useful baseline looks like this:

Same gait each time. Trot on a straight line, toward or away from the camera, is the most diagnostic. Pick one gait-and-geometry combination and make it your standard.
Same surface when possible. Your home arena is fine. An unfamiliar clinic surface produces readings that look like the horse changed, when actually the ground changed.
Same warm-up state. Five minutes into the ride, after the horse has stretched through a walk and a trot lap, is the usual sweet spot. A cold horse reads differently than a warmed-up one.
Clean sessions only. A session where the horse is off, or on pain medication, or in an unusual context, gets tagged as such and held out of the baseline. Uploading it untagged shifts the reference in directions you will regret later.

Ten sessions fitting those constraints will give you a baseline that is more informative than a hundred random clips. Sensitivity is built from consistency, not volume.

Reading the metrics — what each one is for

Rough rules of thumb, in decreasing order of how often they produce actionable flags:

Head-nod asymmetry. The classic forelimb offloading signal. If the head vertical displacement on one diagonal is more than a few millimetres larger than the other, and the trend is new, look at the forelimb that is on the ground during the smaller displacement — that is the one being spared. Most useful metric the platform produces.
Stance-time asymmetry. Time the limb spends loaded. A consistently shorter stance on one side often means that limb is uncomfortable at the moment of peak load. Works well for hindlimb issues that head-nod misses.
Fetlock load asymmetry. Peak fetlock extension angle, which maps approximately to peak load. Drift here shows up before the horse is visibly lame and is often the earliest signal for developing suspensory or soft-tissue issues.
Stride length and frequency. Less diagnostic for asymmetry but excellent for tracking progression. A horse building fitness will lengthen stride at the same frequency; a horse losing fitness or comfort will shorten stride or increase frequency.
Joint range of motion. Individual baselines matter more here than almost anywhere else. ROM varies enormously horse-to-horse. Treat your own horse's historical ROM as the only meaningful reference.
Beat separation (Icelandic gaits). For tölt and pace specifically. Measures how evenly the four beats are spaced. Used to quantify tölt quality and detect drift toward pace or trot.

What to do when you see a flag

The pattern that works, in the order you should try it:

Upload another clean clip. Confirm the flag. Short-clip noise is real; repeatable flags are real-er.
Check the obvious. Hoof, shoe, saddle fit, recent workload. These account for more flags than pathology does.
Call the farrier. Trimming and shoeing drifts show up in biomechanics fast. Often the fix is a ten-minute adjustment at the next appointment.
Rest and re-measure. A few days off with turnout, then another clip. If the flag clears, you caught a use-load issue. If it persists, it is a different problem.
Call the vet. With the report attached. A vet who can see the biomechanics will start the examination in a different place than one working from "the horse feels sticky."

The goal is not to replace the vet. The goal is to get the horse in front of the vet earlier, with better evidence, and to keep the training conversation grounded in data that both parties can see.

The conversation with the owner

One more thing worth naming. Owners of horses in training increasingly expect to see the numbers. A monthly biomechanics summary, sent with a short note on what changed and what you are working on, turns the trainer-owner relationship into something closer to how a physiotherapist communicates with a patient. Professional. Specific. Written down.

That reporting is not busywork. It is how trust gets built when the owner is not there every day. The platform produces the data; the trainer provides the judgement about what it means. Both are required. Neither replaces the other.