Bike Fit App | how accurate are they?

The Bike Fit App story

Over the years I've gotten asked about all sorts of bike fitting tools and systems (and more recently a bike fit app).  It started with basic formulas -- take your inseam, multiply by a constant, like 0.883 and this will determine saddle height.  Then more complex versions of this came about -- take a whole bunch of body measurements and plug it into this computer program and this will tell you completely how your bike ought to be set up.The problem with systems like this is that they don't work well.  They don't take into account individual differences and often don't do much more than provide a rough blueprint for bike setup for a narrow band of the population.Now there are smart phone apps designed to get your perfect bike fit.  One I've heard about a few times is the Bike Fast Fit app available on Apple's iOs devices.  It makes use of the digital video capabilities of your phone or tablet and provides some instruction on measuring joint angles and distances with some nifty features much like you may have seen on Dartfish-style or other movement analysis software and apps.  There are other apps available in Android and Apple stores for movement analysis but this one seems to be the only one strictly geared towards bike fitting.Since this is what I do for a living, I was curious about the app and decided to check it out.  I decided I would run the app on a couple different devices - my wife's iPad and my son's iPod Touch and see what kind of results I would get.How would I test or vet it?  I use an infrared motion capture system in my bike fitting practice and the app's data will be compared to the actual values I get from the infrared.  The infrared motion capture technology is the gold standard for most bike fitting measurements (I do have a video that explains where it can fall short in a few areas, see it here) with accuracy to 0.2 mm in all three axes.  I've used it for over a decade now and the data is consistent, repeatable, and objective which create the best case scenario for dynamic accuracy.A couple of things before I go further:I do realize that it's not fair to compare a $10 app to a $15,000 bike fitting system.  I don't think that this app can hope to achieve the accuracy and repeat-ability of this very expensive system.  That said, the app developers are marketing the "Pro" version to professional bike fitters --

"We listened to your requests and developed Bike Fast Fit Pro with the professional bike fitter in mind.....you can add business name, address, contact information, and logo to bike fitting reports and send PDF bike fitting reports right to your clients."

so I feel it's fair to look in depth at the app and try to determine how useful it is.  Does it provide good data and then (if you're not a pro bike fitter) does it guide you in the right direction with this data by providing "normal ranges" that will help you improve your bike fit?  (More on the normal ranges later)So how can this app be successful?Based on the needs and requirements of an effective bike fit, this app could have four level, or tiers, if you will, of usefulness, starting with the most powerful and useful:

1 Is it capable of being used in a professional bike fitting practice as a full time tool?

As I mentioned I don't have much hope that it can do this, and it doesn't have to in order for me to consider it successful in positively changing the way a rider interacts with their bike.  But in order to achieve this level, the data it gives would have to be spot on with all the infrared measurements and it would have to do this repeatedly and on multiple devices (different iPads or iPhones).

2 Can it be used part time or occasionally by a pro bike fitter when a client is remote and can't get into the fit studio?

This would mean that the data would have  to line up with the infrared data, but perhaps it can't render all the measurements that the fitter's full-time system does.  If the main measurements are possible this would be great.

3 Can it be used by an athlete to check their bike fit in an absolute way?

Meaning that measurements taken can be compared to "norms" so that the athlete can get an idea of where they might fall in the ranges -- is their saddle height on the high side?  Low side?  Right in the middle?  EtcThis is actually a pretty high bar still.  In order to be considered "good enough", then angles would need to be all within about 1° of the "actual" measurement to provide any real value.  Any more deviation than this, and the margin of error grows enough that adequate decisions can't be made about the rider's fit.This point I can't stress enough -- since each measurement is used in conjunction with others to make sound decisions, 2-3° of deviation gets magnified into much more.For instance, finding proper saddle height: if the knee extension measurement we get through the app is 140° and it might be off by 2°-3°, just one of the other measurements we have to look at is the position and motion of the ankle.  If the ankle might also be off by 2°-3°, AND we don't know which direction it's off then this makes it almost impossible to make a good decision about whether the saddle is too high, too low, or just fine.  This is definitely enough error that could convince someone to raise a saddle when they shouldn't.  And remember, this is just one measurement -- if other measurements have similar deviations, the inter-connectedness of bike fitting makes drawing conclusions from this faulty data impossible because one measurement relies on another

4 Can it be used by an athlete to check their bike fit in a relative way?

If it's not accurate enough to provide data that can reliably be compared to averages in a meaningful way, then it would need to be consistent enough for the individual to look at their own fit at different times.  If they can assess changes over time, and provide context for equipment changes (new stem or saddle for instance)  In order to do this then it would have to be extremely consistent between capture periods, meaning if I set it up and take measurements on two separate occasions (let's say a day apart so there's no significant change to the rider over time) it should provide nearly identical data at worst.  It can be successful in this way even if it's not consistent between devices  -- recording one day with one iPad and the next day with a different iPad would hopefully produce the same results, but it's not necessary for success at this level.I won't bury the lead any further.  So where did the app fall?Honestly I have a hard time saying that it even performs at the fourth level with any reliability.So how did I come to this conclusion?As I mentioned, I used the app to take numerous measurements of different angles and lengths of my body and of the bike that are the same types of measurements that my infrared system takes.  In fact the visuals and reports it generates are strikingly similar in formatting to those produced by my infrared system -- almost as if they intentionally made to look like this.  Hmm.  But I digress...So it was quite easy to compare the results.  One thing to note is that in order to get the best results possible, I took multiple measurements with both the app and the infrared, that way any variance in either could be taken into account.  I did this to give the app the best chance to prove itself....if a measurement was off but after looking at multiple pedal strokes, it all "averaged out" then I would consider this a win for the app.  To keep things simple, what you will see in the results is a single number where I've tried to give the app the benefit of the doubt and present the best possible scenario for the data.Without further ado...the data:

Knee Extension

• Infrared = 141°
• iPad = 142.7°
• iPod = 139.5°

Knee Flexion

• Infrared = 67°
• iPad = 69.06°
• iPod = 71.49°

Hip to Wrist (horizontal)

• Infrared = 67.5 cm
• Pad = 70.7 cm
• iPod = 69.9 cm

Knee Lateral Travel

• Infrared = 14 mm
• iPad = na
• iPod = 40 mm

Frame Reach

• Infrared(and actual from bike manufacturer's website) = 378 mm
• iPad = 394.4 mm
• iPod = 371.3 mm

Frame Stack

• Infrared = 551 mm
• iPad = na
• iPod = 547.9 mm

Torso Angle

• Infrared = 46°
• iPad = na
• iPod = 43°

I took more measurements than this, but the rest of them don't add appreciably to the story here.Notice that I did include a few that I only took with one device or the other -- I included them just to show a few instances where perhaps the app got close.  Yes, it would have been great to see what the other device showed for that measurement but by the time I went through all this I had already spent a lot of time putting this all together -- probably too much.Of note was that on the smaller measurements, like trying to measure knee lateral travel, hip vertical motion, and especially ankle motion, the margin of error was much greater.  If the "big" measurements (knee extension, knee flexion, hip flexion, torso angle, hip to wrist horizontal, etc) were more accurate, then the smaller measurements might be over-looked, but since they're not it begins to relegate the overall process to a bit of a guessing game.Looking at this it's clear to me that the app has fairly limited capabilities which is why I relegated it to the fourth tier.  Tier 1 (use as a full time bike fitting tool) is clearly out.  It's not accurate or consistent enough for tier 2.  It isn't consistent enough between devices for tier 3.  I think it barely meets the requirements for this fourth level as well.A case could be made that it is inaccurate and inconsistent enough to merely be thought of as an "Entertainment" app rather than a tool.  "If you like bikes and have a couple hours to kill, try this."  That sort of thing.But I'll give it the benefit of the doubt and say that an athlete might get something out of using it on their own bike.  As long as the data is taken for what it is -- a relative measurement rather than an absolute one -- an athlete might be able to make a decision or two about how to change their bike setup.

But is 2°  really that big of deal?

I know 2° of error doesn't seem like that big of a deal, and in some applications it wouldn't be.  But, as I mentioned when you are stacking measurements on the back of other measurements as you do in a a normal bike fit, 2° becomes unmanageable.  If I took the measurements from my bike fit with the app and ran with them I might make some bad decisions from the data.  For instance, when it showed my knee extension at the bottom of the stroke as 142.7°, and my total knee lateral travel of 40 mm.  I might look at these and decide that my saddle is too high and it could be causing some excessive translation of the knee laterally.  But since I now know that the real measurements are 141° and 14 mm, this would be a mistake in my case.

But wait there's more...remember the bike fit "normal ranges"

Okay, now that we've addressed the accuracy and consistency of the data let's look at the next step in any bike fit -- taking the data and comparing it to some reference values and being able to make good decisions about the fit.  Even many inexperienced bike fitter "paint by numbers" and just compare the data to stated ranges  and make changes in order to put each rider in this range.  This is an okay, if limited way to go about it.Better, more experienced bike fitters certainly use reference ranges as a guide but understand that other factors, like the rider's physical limitations and blind spots need to be accounted for and understand where and when to leave someone outside of a range.So how good is the information provided by the app to allow a lay person to "paint by numbers" at least and make some fit decisions?  The app's bike fit guidelines are pretty bad.Let's say by some miracle you get accurate data, the recommended ranges they give for some of the most important measurements (like knee extension for one) are just plain wrong in many cases.  They are providing ranges that are in line with the older (and inaccurate) static bike fit measurements rather than the dynamic ones that would apply to this type of bike fit.  (Dynamic measurements are taken while they rider is pedaling (yes, the video is stopped to take the measurement but the rider is in the act of pedaling).I could go through and line out where they should have those measurements, but I'm not going to make it that easy for them.  They need to do more of their own homework.

Finally...

So the data isn't perfect, or even good.  I would grade it as "fair" in the context of bike fitting.  The normal ranges are bad....sometimes really bad.  These two things combined make it difficult to say this app is worth $10 (or even the $5 for the regular version).  If you love bikes and think trying this might make for an entertaining afternoon, then great, go for it.  Otherwise, I'd save the money or even just use some of the free tools whether they be app or PC based.I have a training module (free) that's geared towards using digital (via your smart phone, tablet, DSLR, point and shoot, GoPro, etc) that shows you how to use some very powerful and free software to do your own bike fit.  All of this information is free and available on this website; just sign in here.  When I use my camera or GoPro and the free PC-based software, my data is consistently closer to the infrared's data than the app was.  So you can effectively do the same thing as the app, for free, and more accurately.