How accurate can frame geometry be (angles, length)?

I’m making a touring bike with a 70.5 HTA. I’m not welding the frame I’m designing myself. The manufacturer said that they can guarantee a ±1° HTA and a ± 3mm chainstay accuracy. That seems quite the range. A 1° HTA change changes mechanical trail by about 5.5mm (for example from 59.8mm to 54.2mm) and ground trail by 6.4mm (for example from 60.1mm to 57.7mm). Those bikes would ride very different.

Is the degree of accuracy that can be achieved a matter of effort/time invested or is it limited by the quality of the jig or is it more fundamental (i.e. no matter what time and no matter how much effort you put in accuracy stays roughly the same)?

By the way there is a neat video on the accuracy of geo charts here:


Steve at Hardtail Party is a great guy and I love his enthusiasm and I love that put the time into making that device (the GeoMeter). I watched right through the process of measuring the first bike, the Santa Cruz, and it looks like he’s got different size tyres on the front and rear. That will definitely affect the final numbers.

All lower headset cups and crown races do not have the same stack. And all X mm travel forks do not have the same length and offset, so unless he’s using the same headset and fork that the company has modeled their bike around, the numbers will differ.

I’d be inclined to say that the tool he created is not as useful to check if geo charts are accurate as it is to check the resulting geo of the bike with the parts chosen for the build.

To get a true representation of how accurate the geo chart is, he’d have to eliminate variables and verify some of the “hidden” numbers not always communicated via a geo chart.


Im chiming in because it appears you’re looking for an array of comparisons, not because I’m trying to say what’s best, or normal, but provide you a point of reference.

I design and build to linear measurements rather than measured angles, but I do my best to build to a front centre measurement +/- 1mm, which hasn’t been unusual when i’ve asked other hobby and professional frame builders. though its also not unusual for professionals to say they work to even tighter; many say 1mm “total” as in , the sum of all their tolerances, but this doesn’t really appeal to me, some also say “within the mm” which I take to mean +/-0.5mm.

for me this +/-1mm involves a fair amount of back and forth between welding and measuring stations, and usually cold setting and re-checking throughout the welding out of the frame.

I can only imagine I would work much faster if front centre tolerances were +/- 5mm;
EDIT: to answer your question more clearly, I imagine almost any reasonable tooling situation can be made to work more accurately than the offered tolerances, given skill and time.

NB: I say do my best, because if I’m building something absolutely mental (like mtb with a straight gauge 51mm downtube) and it pulls in a couple mm during welding, I might compromise another couple mm of front centre if its symmetrical, but the angle is fighting back when I try and set it further) so its possible that if you’ve designed something the manufacturer doesn’t usually do, they may be offering you wider tolerances so as to play it safe with something unknown.


To give another example because I definitely don’t think what I do is normal: I build very accurate subassemblies using a sine bar and bevel protractor and hold everything with a mill table. and then whatever I get as an end result, I get. I don’t cold set whatsoever, since I don’t have a way to support the brazed joint and bend the tube.
I put the finished frame in a stand, set the head angle to the design with a digital angle finder, and verify corresponding angles. I’m good to plus minus .2 or better .

Plus or minus .5 degrees is probably a CYA number and 3mm of chainstay length is also, depending on if you measure horizontally or bb to axle… I bet their actual tolerances are better.


…that’s exactly what I was looking for! Thanks.

± 1mm of front center length is equal to ±0.1° of head tube angle (at least at the HTA range I’m looking at). That is what I was hoping for but then I was a little shocked to hear about the ±1° HTA accuracy. In a different context another frame builder said he could do a ±0.1% HTA (= ~ ±0.07°) but I had assumed that he was just quoting me the nominal/official accuracy of his jig which I guess might be quite different from the real accuracy that comes out at the end. Anyway good to know that 0.1° HTA is within the realm of possibility.

1 Like

I also thought that it’s a “cover your ass” (had to look CYA up :slight_smile: ) number. But the measurements in this article Just how accurate are bike geometry charts anyway? - MBR suggest either that the ±1° HTA accuracy number is realistic and not only a “cover your ass” number or that the way measure these angles has a ~±1° accuracy :slight_smile: In other words such a test is worthless for the reader if the accuracy of the measuring instrument is unknown. The article doesn’t describe the measurement method.

I didn’t read too closely. I’m pretty skeptical of people who don’t build bike frames and propose to measure them. The MBR article is looking at mass-produced full suspension frames with hydrofromed shapes and all sorts of features that will cause variations and make them hard to fixture or measure (I also wonder about some of the crazy variance in STAs… There’s no way a frame could be off by 10 degrees… I don’t think the manufacturers are all reporting the same number MBR is measuring)

A traditional diamond frame with straight frame tubes is going to be much more accurate than a FS MTB.

1 Like

Agreed. Half of the bikes are carbon fiber. At the bottom of the article they say that the carbon fiber bikes would be a lot more accurate because of the different production method. But MBR’s measurements of the carbon fiber bikes are not a lot more accurate then those of the aluminum bikes. That’s what made me doubt the measurement accuracy of this article.

This info is from ORA Engineering, who IMO is one of the best frame manufacturers, and who have a digital frame measurement jig:

These are their production tolerances:

HTA .5deg
STA .5deg
BBD 2mm
CS 1mm
STK (driven)
RE 2mm

A few points:

  • Tolerances are a distribution. 72+/-1deg does not mean your HTA will be 71 or 73deg. It means on average it will be 72deg, and the maximum acceptable deviation is 71 or 73 deg.
  • All the dimensions are coupled together: you can’t change the HTA without changing the BBD. This is why the BBD is +/-2mm. The designer and manufacturer need to decide which tolerance is important, and hold the frame to those standards. That is what the toleranced drawings are for.

I have a (somewhat wide ranging) sense of how much accuracy is possible now.

But what I still don’t fully understand is what the determining factors are for accuracy and whether a very high accuracy can be achieved by investing more time / effort. This is of practical relevance because after our discussion here I offered the manufacturer to pay extra for a very high accuracy (0.1° head tube angle, 0.5mm chainstay length) and they said that that is possible for the extra payment. If high accuracy can actually be achieved by investing a lot of time/effort then this is credible and I would agree to this. But if time/effort are not the main factors and because head tube angle and all the other dimensions as Daniel said are distributed normally, all I’m buying is the willingness of the manufacturer to deal with my refund/complaint if one or more dimension ends up on either end of the distribution AND I do actually measure it… :slight_smile:

Tolerances are not about time and effort. Tolerance is the reality of manufacturing - temperature changes, material variation, tool wear, human touch etc… For metal bike frames, most of the variation will come from welding distortion.

Engineering and manufacturing are not about getting the exact results, it’s about finding a process that is good enough (meets the design requirements) and minimizes the time/cost. You could get tighter tolerances on your frame, but it would cost way more and/or it means you end up tossing all the frames that don’t meet spec. Your contract manufacturer probably does not have the equipment nor the desire to build to that high tolerance.

If one of the best factories in the world with over a million dollars of capital equipment builds to the standards I posted above, then that’s probably the ceiling of accuracy for a bike frame.


Sure would be a lot easier with a wire EDM though!

I have done a lot wrong, but still haven’t managed to make a bicycle that rode poorly, even with the crooked ones. I know 77 degrees would probably be bad, and 57 degrees would also probably be bad. Beyond that, I haven’t really learned much about bicycle design. Any number I choose is practically arbitrary. It’d be foolish to sweat a degree, at least in my case


Is there a difference in terms of achievable accuracy between steel and titanium (because one or the other distorts less)?

Either this is right or the accuracies mentioned above by @crowe-molybdenum (0.1° accuracy) and @Nick2 (0.2°) are unrealistic.

One theory that would allign all of the observations made here is that big manufacturers like ORA build to “cover your ass” standards as suggested by Nick2 above so that they can guarantee that 99.9% of their frames are within the ±0.5° range. It could well be that 80% of the frames are ±0.2° and if someone like @Nick2 or @crowe-molybdenum observes a 0.2° or a 0.1° accuracy that is believable IF the quantity of frames they build is sufficiently low. Just my statistics approach to bring together all the observations of this thread.

What you’re not factoring in is time… what is realistic for a person trying to make themselves a bike is not realistic for a person trying this make money. I can spend months on a frame if I want while a factory can’t even spend the afternoon.


I was hoping that time would be a major factor but that contradicts:

I’m not talking about welded bike frames.

Edit: if you are only talking about welded frames, sure. I think that for most people, a certain level of distortion is inherent in the welding process. I can’t speak to that. Maybe there are some people with the skills, knowledge and intuition to mitigate distortion. I am not.

I do know that as a medium-skill brazer with a healthy amount of free-time, it is possible to largely eliminate distortion in a fillet brazed, bronze joint.

It seems like you’re struggling to believe the factory, and now struggling to believe some people on this forum, but I don’t think this is really that complex;

Following a process that reliably yields a greater level of accuracy can take more time than following a process that doesn’t.

Paying more for tighter tolerances is extremely normal, often in part, because producing a thing to tighter tolerances can take longer than producing it to looser tolerances.

Im lucky if I make one frame a month. But I’ve possibly chosen to make my frames with some tighter tolerances than some other people have decided to make their frames to; because our priorities are different; and what we think of as “good enough” may be different.

I tig weld steel frames and there are many things I do to make sure they come out as I like; if I didn’t care for 5mm of front centre, I could spend less time doing those things, and the frame would be built more quickly, though I couldn’t guarantee it would be as accurate.

A high level of accuracy is definitely achievable given time and money/resources. My jig is not the stiffest around. It flexes if you lean on it which means I have to fabircate everything accurately so that is just sits in the jig. All mitres/copes/joints are built to a super tight gap so that there is plenty of metal on metal contact before any joining. Gaps are the killer of accuracy. They will pull close on cooling in either welded or brazed joints. I can put my frames back in the jig without any force after brazing because of the time I spend getting the joints right. I rarely cold set an of my frames. The rear tips occasionally need a slight tweak but usually only when the brake bridge is brazed in. I don’t weld so I can’t comment on how much that moves around.

3mm on Chainstay length really isn’t going to affect the way the bike rides. 1 degreee on HTA will though and I’d be asking for a tighter tolerance to be honest. It is achievable.

1 Like

I wouldn’t pay extra for those tolerances and I definitely wouldn’t request a refund if a builder is not able to hold them, assuming the frame is not egregiously out of spec. I totally understand the desire to have an accurately constructed frame, but if you care about 0.1° on the head tube angle and 0.5mm on the chain stay for performance or handling reasons, then you’re so far down the rabbit hole that you should also be considering water bottle mass/placement, front and rear tire pressure/compression (so the bike doesn’t sit at a slight angle), whether you’re riding the bike in the morning or evening (body weight can fluctuate a surprising amount throughout the day), whether the frame is loaded or unloaded, how the load changes over the course of a ride, the different surfaces you’ll be riding on, etc. etc.

Bicycles handle the way they do because of the forces acting on them. Geometry charts can give you insight into roughly how a bike will ride, but they are an inherently incomplete picture of the dynamics because they lack the forces, system stiffness, system mass, exact weight distribution, etc. No two riders will experience the same bike in exactly the same way, and there are probably significant differences in the way riders between sizes experience the “same” bike from a company’s line up.

These things are fun to think about, and even more fun the experiment with (that’s one of the great things about custom frame building :slight_smile: ), but at some point you’re really splitting hairs. If you want to determine what actually matters it is going to take your own ride testing to figure that out.