The absolute best way to deal with this is to take design off the table and make it about usability and things you can tie to metrics. Then you track and report on every metric imaginable. You already know where issues will be in the UI/UX, finding data supports this will be trivial.

Marissa Mayer is actually a great resource on stuff like this. She famously got into an argument trying to test 41 shades of blue (or something like that) on the Google homepage and had a designer quit because he thought it was stupid.

If you aren't tracking/testing/iterating on user experiences than you are not thinking about them correctly, imo. Designs should be a means to an end and never the end in itself.

+1

the practical problem with this, though, it that doing proper AB testing requires that you actually design 2+ versions of something, which can be infeasible because of time constraints. not to mention setting it up. so yeah, it's 100% the right way to approach it, but i find you often have no choice but to make your best guess about something. so guessing well and having good instincts becomes really important in practice.

So what is a career path for someone into one of these mid to high six figure salary positions? What are the educational and experience requirements, etc, or a path like "entry level: years 1-3, junior: years 3-8, senior: years 9+", and so on?

Maybe this question doesn't make sense. Probably doesn't.

So we had a old version and the new one's been in beta that users can just click into. We actually have a ton of data for user metrics, but the PM said we're releasing the new version regardless, because the process has been a year or something and needs to be wrapped up asap.

Also we stopped gathering user data recently, apparently because we're rolling our own infra and can't handle the volume.

I'm still junior so wanted to check if all this is relatively common.

Yes, it is all extremely common.

It is not what top-tier companies are doing, but you have to remember there are stats like 70% of IT-related projects fail, so shipping anything that is an improvement in I'm sure at least several ways over the current system is a success.

Good systems should be able to be rapidly A/B/multivariate tested basically all the time. We build websites for modern companies so in my world it is absolutely a design pillar to have a system where you can rapidly set up tests.

We just built a site that is going to be plugged into a "machine-learning personalization algorithm" and different panels of content, images, copy, etc. will be showed to different people based on segments that connect to a major marketing automation application.

All of these pages are also being analyzed with heat maps and behavior as well, etc. It is really ****ing fun in my opinion as well to create these cool marketing experiences and then see the results work.

Thanks it was something even weirder than this. When in doubt just recreate the project. Frickin VS. It doesn't like when you drag and drop cpp files into a project. Gotta copy paste the code and then it works for some reason. This ONLY happens on my setup too, I can do that same thing at school with no issues.

Different VS version?

Any resources to learn the proper way to do this type of stuff? I'm building a site and have no idea where to start on ux analysis and a/b testing

I found this site a few days ago and thought I would link it here because I find it neat.
Site: https://www.shortcutfoo.com

Useful for getting faster or just learning shortcuts for IDEs/text editors that you probably use and its sort of a cool concept.

Book marked. Thank you iosys

+1, it feels like we're trying to do the right things, but setup too ad-hoc and basically there's no follow through.

I'm trying to write a good response and I'm having difficulty because I've been immersed by this for the last 4+ years. I've learned from reading hundreds of articles, dozens of case studies, conferences, and this is what our company does.

I think in general building a good UX is very simple. The biggest issues are procedural in terms of red tape, and getting teams to work together and agree on what is important.

1) You have to know your audience. You have to understand what they are going to want to do. Then you need to come up with the things you want them to do, and you want it to take 3 clicks at most (in typical scenarios).

2) Then you need to set up tracking analytics. Google analytics is a product worth knowing, if you don't like google then you can use Piwik.

3) Analyze the data. What is it telling you? Where are users coming from, what pages are they landing on, where are they then clicking on that page, are they scrolling to the bottom of the page, are they leaving your site, etc. Search for things you don't understand, "What should my bounce rate be?" "What is the difference between organic and referral traffic behaviors?" Unless you are super experienced, you don't even need to take action on this, just use it to create testing criteria. If you think X, test it to confirm it.

4) Testing. If you think users would click a larger, more prominent button, then set up a test. Clone the page and create a duplicate with the new button. You can use Google Experiments (part of analytics-free) or something like Optimizely to send a % of traffic to this new page. Or you can use paid-tier Optimizely or similar tools to do dom-replacement A/B testing and they will basically just edit html on the fly for you and give you results.

You can test small or large. You can show users a whole new page and track several metrics at once, or you can show them a single new button.

There's a lot of art and science in this bullet point. It's a talent you can't really teach imo. Many many people who are gainfully employed doing it are very bad at it.

I think this is a topic in general that if you have a natural talent at it, you create such a massive amount of efficiency compared to the alternative ways.

On the audience point. Say you find yourself as a Director of UX and you have not a clue about your audience. Well you can build reports on it, you can figure out your demographics, etc. and then you can read about those demographics, or even conduct surveys to understand what the users want. There are ample tools/consultants that will take your money and provide data about your audience.

But you can easily save tons of time and money by just kinda knowing this yourself intuitively. This is why people like Sam Altman say that college kids are in such a great spot to figure out trends, you are in an environment with tons of people not likely to accept the status quo and likely to experience new things and see trends appearing.

However, people who are not in college, and even people who sit in a room all day by themselves are very capable of building a super impactful app/website/etc. You can immerse yourself in the data and figure it out because you are swimming in it, or you can kinda just understand intuitively what you need to do.

The next step is understanding technical limitations in relation to UX and designing solutions that are easily implemented. It is easy for me to create a list of demands on what I want this awesome web platform to do, but actually getting it done in a reasonable amount of resources is another thing.

This is a great response. Its a bit frustrating how diverse experience in just web app building can be, but I guess thats why programming has such a big job market.

As for tracking user activity on a page, is this also something included in google-analytics free? Were looking to do essentially everything you listed for minimal cost. Any other software/services you recommend for a small scale startup in this realm?

I feel sorry for the early day UX designers of the web.
Rough life in browser world couple years back.

Also I get the general impression, to make it as a good UX designer for the web, you have to be good at designing and programming.
Sort of a rarity to have good programmers that are also good at design.

Google Analytics free-tier will provide you 99% of what you need.

It is imo the greatest free product available on the web, and its not even close.

Thanks for all your help sir. Its interesting getting some experience with all aspects of product development after a few years in regular old server programming

i'm sure it's possible to have multiple github accounts run from one computer, right? Do you need to set up multiple identities?

thanks so much for this. the funny thing is i think we do all of this, but the problem is some teams make use a lot if it (data/search/etc), unfortunately some don't.

when it comes to major updates (say 3-4 months of engineering), how do you effectively AB test? i mean the new flow takes that long to develop, so you can't test parts of it. by the end of the 3-4 months if you test the whole thing and get negative reviews, are you just gonna not ship?

Yep.

We had one comany we created a new homepage for and they tracked us to 18 different metrics. 5 of them were basically show stoppers if we didn't improve them. These 5 were "analysis of your [major marketing service] account", "free trial", "request a demo", "bounce rate", and some other stuff.

We took them from 80(!) different buttons/links/options on their homepage, down to 15-20 while vastly updating the design to look like a modern SaaS company. They had just raised a 10-15m Series B.

We started with 25% of homepage traffic and then gradually went up to 50%, 75% and then cut over completely when we continued to beat the old site, we made some iterations in between there as well.

We started working together and had the new homepage live in 4 weeks. Their director of enginneering took responsibility for the infrastructure since their software had like 20+ aws instances. It was after a couple months later and a full site launch on a new version of the cms that we found out he thought "caching was only to be used as a bandaid" and put the site up with zero caching, I mean none. This resulted in emails to their blog ddod'ing their site. They freaked out at us, we had to fight for 3 weeks to get the infras set up with multiple layers of caching/cdn, etc.

Because of that bs we terminated them and aren't allowed to speak about it publicly and in our last conversation I told them that they are basically idiots and I expected better from a company pretending to be a tech company, which I was told you apparently aren't allowed to tell a client.

I think they were likely being sketchy and wanted to take credit for our work. Since this is a company that shouts all the time about what sick ninja-hacker-wizards they are.

But yea the moral of it is, they would have cut the project if the new designs couldn't be quickly iterated on to beat the metrics.

Even if it looks better, is better to maintain and add features, when you are getting conversions at scale (they were 7 figures in monthly views and 4-5 figures in monthly leads) you can't degrade that and hurt your overall growth rates.

I see both my personal one and my work one when I log in, not sure how they got linked but it lets me choose which accounts repos I see.

i tried adding the same ssh key to another account and github complained. I guess I can generate more keys though, so that's probably what I needed to do.

Also, looks like I'm project manager for my capstone class by default. :x

I'm considering a job with a different group in our company. The work seems much more interesting and career-expanding than the foreseeable future with my group now. The drawback: they adhere to the ungodly abomination of 4-space tabs. I don't know if I can do it.

Furthermore they all still use Eclipse as their primary IDE. So you're already looking at your code in this little window, then add in 4-space tabs and you can barely see any of it w/o tons of right-left scrolling. I don't get it. People on Eclipse should want 2-space tabs more than anyone.