Really interesting data, thanks! However, I think cost per gram metric is the wrong way to indicate the diminishing returns. Whenever I’m making a purchase, whether it’s a new bike or a new component, I consider instead the price per gram saved.
As an example, using your price per gram metric, an $7000 7kg bike would have the same price per gram value (1$ per g) as a $8000 8kg bike wouldn’t it? In that hypothetical situation, the $7000 bike would obviously be much better value for money, if bike weight was the buyer’s priority, even though they share that same $ per gram value.
Calculating price per gram saved gives a more meaningful indicator I think. Of course, you’d need to choose a reference weight for the weight saving, perhaps the median weight.
Hm… Good point. I’m open to suggestions on how to improve those stats with the data I have. Maybe @jfurner will know. He seemed to be more adept at data than I am.
This is the first time I’ve posted on a TR forum, and I I do so now because I’m really disappointed in this community. FFS people. The OP put in over 40hrs work and presented us all with an interesting data set. All you have all done is criticise them. Saying “oh the sales price stats are misleading”, or you’ve done something else wrong. Trying to flex your intellectual muscles and criticise the stats and the exercise. I also work in stats and data analytics and you know what, I have just enjoyed reading the OPs work. It was fun and interesting and I got something out of it. As you are all so good at interpreting the data, just do that, read it correctly, and enjoy it for what it is. Pull yourselves together.
I give the OP a round of applause for doing this. Thanks for your time and effort. A real good breakfast read for me. Cheers
Thank you, Nate, it means a lot. I honestly felt surprised and disappointed by the reactions as well.
Of course, I am open to constructive criticism to improve my work, but the amount of work put into this was huge. I basically doubled the dataset + added more info (about the groupsets). So yeah, some comments were hard to swallow.
I’ve stayed quiet, but been thinking the same. “I’ll bet OP is more than a little disappointed with the response”. People can be hard, and even harder on the internet. Having said that, I think MOST of the critical replies are meant as constructive criticism, and while they come across as harsh because it’s the internet, I hope you don’t take it as a personal attack on your hard work. I’ve been happy to see your openness to improvements and asking for collaborative thought.
I come from a BIG projects background, where I managed extremely large IT efforts. It never felt good to have my leaders (or peers or staff or users) criticize our work, I often had to remind myself that 99% of them were just trying to ensure we put out the best possible work that added the most value and that it was almost never personal.
Thanks again for the work and for sharing. I know I took away a few good pieces of info from reading it.
Very interesting overall, with a few surprises for me. Keep up the good work I’m sure i’ll come back to this for reference, hopefully you can refresh this every year!
If you’re looking to continue working on this, one thing I’d like to see on weight - instead of $/gram, it would be cool to see $/gram saved vs the heaviest bike. For example if the heaviest bike is 10kg and $1000, how much extra would you spend to have a 9kg bike.
Thanks for putting this together. The most interesting aspect was the totality of both carbon fiber, and electronic groupsets. I had suspected the percentage was high for both but so high in this sample is surprising.
Yeah, that surprised me as well. Unfortunately, the new Shimano mechanical 105 was introduced, so in the following months, we will see many new bikes with it, making my research obsolete already.
What counts as ‘a bike’ here? If a brand offers a model in multiple configurations, does each of those count as ‘a bike’ for these purposes? If not, how do you determine which price, weight, etc. to take as representative?
If they do count separately, I would imagine that would tend to skew things towards carbon, electronic, higher prices, etc. as higher end bikes are probably more likely to offer more variants, whereas lower end models would be much more ‘take it leave it’. I’m not sure exactly how that could be addressed, but some way of weighting them accordingly in the stats would be interesting.
Hi David,
As I wrote in the notes in the article:
“Every bike is included once (same bike in different colors = one bike). This does not apply to bikes with different groupsets (the same bike with Shimano Ultegra or Ultegra Di2 = two bikes).”
Nice to look at! What I found interesting is that people buy carbon bikes without knowing the weight?
Some tips for data representation if you’re trying to learn from this for the future.
on your question on “cost per gram” above, plot weight (x) vs the value (weight/cost, y).
As somebody said above, it would be clearer to update the plots with “index number” of the bike on the x axis to be histograms. Extra star if you put in vertical lines for median and mean.
And as above, the bar charts for “min, avg, max” are better represented as box plots where you show all the data. Extra star for labels for the actual median/mean prize to the side of the plot.
I honestly felt surprised and disappointed by the reactions as well. 
Thanks for the suggestion.