There's many services that provide this these days, including HotJar (paid) and Microsoft Clarity (free).

Being able to see users actually use your site, combined with heatmaps and more really lets your team focus on usability issues, or help direct attention to more important features.

Things like, do users struggle to find the shopping cart button? Are they confused about how to create an account? Is it frustrating to make changes to something, etc. Engineers are really bad at understanding usability issues unless they see it for themselves.

However, it does tend to be something you review/monitor sporadically, not every day. You have to have enough samples to see if changes have had a significant impact or not.

The ones we've used have all had the ability to censor what the tool believes to be sensitive information, such as passwords, credit card numbers, etc. In our experience, these tools overzealously censor on-page text out of an abundance of caution - which is just fine. Rarely are you interested in reading the text on screen, rather, you're watching how the user interacts with the page.

It's kind of a ridiculous notion that a site operator cannot monitor how users use the site... this is not wiretapping and I assume (hope) it gets thrown out.

Such types of analytics tools are useful for certain purposes, however I do think we need some sort of regulation, or monitoring to decide where the line is, and how far any company should be able to go with this type of thing. It's a bit disconcerting because without such technical regulation, where does the line get drawn? Will be the norm transition to discreetly recording audio/video as well?

> It's kind of a ridiculous notion that a site operator cannot monitor how users use the site... this is not wiretapping and I assume (hope) it gets thrown out.

Putting aside how many other ways there are to determine how users use a website without needing to resort to this level of intrusion, the issue is that almost always the user isn't aware this level of monitoring of their actions is happening. Sure, it may be buried deep in some privacy policy or terms and conditions no one ever reads, but that's not exactly what I'd call informed consent. This is where regulation to at least require websites notify their uses that there's literally a recording being taken of their screen interacting with the website would be helpful.

I don't think it matters at all. There is no expectation of privacy when browsing a website (from the website operator). We can see everything you do... obviously.

Also, it's not a recording of their screen that is being taken, but rather it's mouse coordinates and keystrokes which then re-assemble a replay of the actions. Just like a video game replay.

I can’t talk for the US, but in the EU under GDPR there absolutely is an expectation of privacy.

This is not obvious to the average person outside of the industry. For example, most people would expect that text typed into a form field without clicking submit is private and has not been sent to anyone. It is obvious to a software developer that secretly sending the text back to a server while the user types is easy to implement, but posting an agent to read over someone's shoulder while they write a draft which they may choose not to send at all does not become morally acceptable just because modern technology has made this more practical than hiring a PI to watch through someone's house window from a public street.

> I don't think it matters at all. There is no expectation of privacy when browsing a website (from the website operator). We can see everything you do... obviously.

Server logs will of course tell the operator what buttons are being clicked on and which pages are being requested. But once those pages are on my computer, I disagree, I do have an expectation of privacy around my mouse movements and keystrokes on my own computer. If I choose to send those to some third party server so be it, but having those collected without consent is a problem.

As engineers we may assume this is being done regardless and accept it as a reality of the web, but I'd imagine the average person would be quite surprised to see what is a recording of their browser window reconstructed without being informed that was happening. They may rightful have that expectation of privacy here, which is being broken by these tools.

> Also, it's not a recording of their screen that is being taken, but rather it's mouse coordinates and keystrokes which then re-assemble a replay of the actions. Just like a video game replay.

It doesn't matter what method is taken to generate a video if the end result is the same as if you were recording my actual browser window.

How would this be any different if I were to design my site to use a websocket and stream mouse events and keystrokes to figure out if you are clicking a button, versus using a javascript event handler or something? In fact, a lot of your browser games operate this way (using coordinates).

There is no one single way to design a website.

The very same non-technical folks that you allege are misled here also have no clue how buttons work on a site either... it's all magic to those folks. Are we not supposed to do any logging now? How is it any different to only log when a user presses a button?

Your TV tracks what shows you watch, what buttons you press, how often you skip commercials, etc. Your smart fridge tracks how often you use the ice feature, your car tracks how often you drive aggressively, etc. No one is alleging these devices are "wiretapping"...

> How would this be any different if I were to design my site to use a websocket and stream mouse events and keystrokes to figure out if you are clicking a button, versus using a javascript event handler or something?

It shouldn't matter the collection method. The point is that you'd be collecting these keystroke and mouse events as they happen in an environment that I, the user, may believe they are not leaving my system.

> The very same non-technical folks that you allege are misled here also have no clue how buttons work on a site either... it's all magic to those folks. Are we not supposed to do any logging now?

That's a fair point. Some users will have no idea at all what data they're sending to the likes of Facebook, for example; it's just a magic box that shows them photos of their grandkids. The test here is given the average person's understanding of the products they are using, is it reasonable to assume the data being generated is not leaving your local system? I think even the average user would assume mouse movements are not being sent away to reconstruct their session on a website down to that level of granularity. You may disagree but that's the tricky issue here: where does the line of an expectation of privacy get drawn? From a technical point of view, I'd argue it's at the separation of client-side and server-side interactions where anything not needed for the functioning of the product on the client-side remains client-side. Hence, something like my mouse movements on a run of the mill website for analytics is across that line but something like mouse movements in a multiplayer video game is not.

> Your TV tracks what shows you watch, what buttons you press, how often you skip commercials, etc.Your smart fridge tracks how often you use the ice feature, your car tracks how often you drive aggressively, etc.

The point is where does this data get sent to? Admittedly I don't know exactly what data my car collects on my driving habits, but I'm pretty sure it doesn't leave the car in that it doesn't have an Onstar-like system installed in it.

If this data is used to essentially be an observer sitting over my shoulder without telling me when I don't have a reasonable expectation that it's happening then, yes, that's a violation of an expectation of privacy. The fact that all of this is legal really demonstrates how far behind laws are to the issue of data collection around usage of services and products.

---

edit: thread depth limit reached so I'll just say the following:

> Your car collects huge amounts of data about your driving habits. A Tesla, for example, might auto-report that data, but your "dumb" car also provides that data to an ODBII reader. In fact, that was the premise for those Progressive (I think) Car Insurance things they wanted you to plug into your car to "prove" you were a safe driver.

Right, and the difference is that the data provided through the OBD port requires me to physically plug something in and upload it elsewhere. I have control over that data. This is different from these websites that are collecting data like my mouse movements and immediately sending it off without notification or consent.

It's very common to log when a button is clicked. Which means we can already reconstruct a user's session because we know what coordinates buttons have. We might miss out on the random movements and such, but that might not be a big deal if we also log the timestamp when a button was clicked.

Again, if I designed my site to use a websocket instead of event handlers - the result is the same. Are we proposing we outlaw this? Or inversely, placing in law the only accepted method of website design? That seems foolhardy to me.

Your car collects huge amounts of data about your driving habits. A Tesla, for example, might auto-report that data, but your "dumb" car also provides that data to an ODBII reader. In fact, that was the premise for those Progressive (I think) Car Insurance things they wanted you to plug into your car to "prove" you were a safe driver. Any time your car is in for service... this data can be freely read. Your insurance company will also obtain this information after an accident in some cases as well.

We use mouse/key movements for a lot of things, not just reconstructing sessions. Google's newer "I'm not a Robot" captcha's just ask you to click a single button for example. A ton of anti-fraud/abuse systems do this as well, particularly for credit card checkout forms (anti-bot measures). The ramifications of a bad decision for this suit are huge.

> Again, if I designed my site to use a websocket instead of event handlers - the result is the same. Are we proposing we outlaw this?

yes?

I can stab my knife into flesh all day, if that flesh is another human being it turns out it's outlawed.

There is no requirement that the law be applied consistently for some action. Insider Trading doesn't imply all trading is illegal. Being able to pop off fireworks doesn't imply doing so anywhere is legal.

That would then be the first law of it's kind, to mandate how a website is to operate on a technical level.

Even the existing ADA laws do not specify technical details (which is why they are abused in frivolous lawsuits to often).

The opposite actually, if the law says you cannot murder, it doesn't become legal because you used a hammer instead of a gun.

Your question is essentially "well what if I use a knife, because knives can be used for skinning, which is a legitimate use!".

It doesn't work that way.

Like I said, that would be the first law of it's kind. We have zero laws that specify how a website is operated on the technical level. Or how it's engineered or designed. Nothing exists that is comparable to what is being proposed.

That's true, I think the point is that such laws need to start getting created because the compute and personal device issue is legitimately different.

The problem with this, is the technology evolves far quicker than our laws do. The ADA law is a glaring example of this, unfortunately.

Apparently if you outlaw murder, the fact that people are constantly creating new ways to murder makes it confusing as to whether or not murder is outlawed.

No, because murder is still illegal.

In a technical setting, we can't even reliably describe how things will operate 5 years from now. Imagine a law that specified how frames were to operate... and now no one uses frames.

Just imagine a law that says "no tracking". What is tracking? How is it done? How do we know it when we see it? All of this must be defined... and these definitions are fuzzy and change rapidly.

The ADA law specifically says you must make your website accessible... but fails to specify what accessible even means in a modern setting. That's the pitfalls of trying to regulate how things are designed in a technical setting. Perhaps it's possible, but I have not yet seen a good law that made an attempt.

Docker is 9 years old, try again.

No, actually, don't. I reject everything about your premise.

"What even IS ACCESSIBILITY, OMG, you can't explain that!?!?!".

I mean ... maybe YOU can't, but plenty of people can.

No, nobody can. Not even the folks that wrote the law. Not even the folks that enforce the law (DOJ) (referencing the ADA law in regards to websites). You can go look for yourself... it's created an entire cottage industry of would-be vexatious litigants, if any of the suits actually went to court.

So no, it is not easy to define or describe. There is an incredible amount of detail that must be hammered out to make it easy for everyone to become compliant.

We can also reference the EU's GDPR, or California's CCPA/CPRA. All of which are struggling to actually clearly define what they mean. What is an essential cookie? What does it mean to sell someone's personal data? Neither are what you might actually believe...

You can reject it all you want, but that does not change the reality I'm afraid.

What you're describing is called interpretation of the law, and weirdly enough, legal systems have been doing that for the better part of 600 years.

> It's very common to log when a button is clicked. Which means we can already reconstruct a user's session because we know what coordinates buttons have. We might miss out on the random movements and such, but that might not be a big deal if we also log the timestamp when a button was clicked.

The point I was making above was that there's a difference between purely client-side events like mouse movements and keystrokes and server-side events like button clicks and page requests. This may be a trivial difference for most websites, but I have an expectation of privacy around client-side events. Collecting them without the user's consent violates that regardless of the technical mechanism of implementation of that collection.

> Your car collects huge amounts of data about your driving habits. A Tesla, for example, might auto-report that data, but your "dumb" car also provides that data to an ODBII reader. In fact, that was the premise for those Progressive (I think) Car Insurance things they wanted you to plug into your car to "prove" you were a safe driver.

Right, and the difference is that the data provided through the OBD port requires me to physically plug something in and upload it elsewhere. I have control over that data. This is different from these websites that are collecting data like my mouse movements and immediately sending it off without notification or consent.

If I were asked for consent for all the things you listed, I'd withhold it.

Sort of like how Apple gave users the ability to consent and they overwhelmingly withheld it (to the chagrin of FB).

If someone learns they're watched via cameras while in a grocery store, they may not be happy about it but they'll also generally recognize they're on _STORE PROPERTY_.

The key difference here is that this is being done on the _USERS_ property and you're acting as if they shouldn't have to right to withhold that consent.

---

Everyone accepts that an employer may look through company emails of its employees because it goes through company equipment.

These same companies then have the unadulterated gall to turn around and try and claim they have those very same rights on _OTHER PEOPLES EQUIPMENT_.

They're laying claim to other peoples compute.

The reality is you wouldn't be allowed to use the site at all then. As invisible as you might want to be, the system does need to track you and your actions to do anything useful at all. We're not talking about a blog or static site here...

If someone asks to borrow my car to run to the grocery store and they drive 100 miles away to another state, it's theft.

If I consent to have someone use my compute for the operation of their website on my device and they decide to start recording tracking information, it's theft.

There is no dichotomy here, it's possible to consent to one use and not another.

I think we are in a changepoint time where we need to think about how to make privacy serve the user. By that, I mean it's no longer acceptable to cite the legal standard, but the ethical one. If you collect data, it should be to serve the user. Additionally, any data not used operationally (keep the service running) should have either be limited by purpose ("we will not use your data to target anything to you") or via true anonymization (differential privacy).

If it's not beneficial to the user for you to have this data, you shouldn't have it.

It is equivalent to following someone around a store and writing down how long they look at each item on the shelves, what they pick up, what they put back, etc.

There are (multiple) companies in the US who already do this via in-store cameras and secret shopper monitoring. I know this because I've worked with them.

How is it ridiculous? This seems like an argument more like because you can do it, you should do it. What if car manufacturers had similar ideas to "focus on usability issues" and recorded every driver as they're driving their car. You could even use the same argument, if you're out in public there's no expectation of privacy, and you don't own the tech in the car, you just license it. But somehow we would all consider that to be crossing a red line and to be super creepy.

I’m probably happy to provide that information but I want to know when that happens, and not by fine print in “terms” somewhere.

My browser should render a page, then if I write an angry review and edit it to a less angry one before I submit, I expect my angry review to be logged nowhere at all.

This reminds me of days of unstable internet connections & form submissions not working randomly. In those days, I would type my comment in Notepad first and then copy, paste, and submit. Perhaps, it is time to restart that practice.

When a user goes to your website, the user is trusting you to run you run your software on their hardware. You should respect that.

A few years ago I was at a startup that integration one of these video playback services (Fullstory) into their product. At the time the PM team was all jazzed up about the idea and engineers were rushing to get it implemented. I was the only person to bring up potential privacy concerns which were immediately dismissed under the excuse that the PMs absolutely needed this level of functionality to view individual user's full sessions on the website.

Anyway, Ublock blocks Fullstory's JS from running by default so at the very least your ad-blocker is potentially saving you from this level of intrusion.

I don't think this is entirely accurate. I was at FullStory, as a director of development and founding engineer on the mobile analytics, and we were focused on our privacy-by-default initiative that blocks out user PII while still giving you PM-level information.

The founders of the company were invested in privacy and PBD, and we put a lot of effort into balancing data with privacy. One of the founders even spent a great deal of time on the initial implementation.

This work started years ago and I can say that everyone was interested in privacy, from recording/playback engineers to execs.

I didn't work on the integration directly at the time so it's certainly possible some of my details are incorrect. That said, from what I remember of the demos that were given there was the ability to set regions of pages which contained PII that were not collected but this was entirely up to us to configure. Otherwise, it was possible to watch an individual user's full session on the website.

This was years ago too so maybe it has changed since, but my issue is with the whole premise of products like Fullstory that make it possible to essentially watch your user's over their shoulder without being given proper notification that level of monitoring is occurring.

The potential is there for sure, but the company believed in giving customers the tools to ensure that PII was not recorded, including go as far as giving internal engineers tools required to scrub data permanently that was accidentally recorded.

I no longer work at FullStory, but I believe very strongly that the culture of privacy still exists.

Senior individuals at FS, myself obviously included, felt very strongly that we wanted to avoid any of the creepiness factor, and it was important to us that we felt comfortable with the technology. Discussions on privacy were paramount in all designs.

We demo’d Fullstory recently and decided not to go with it because of user privacy concerns, but there is a “private by default” setting now that masks all fields except ones you explicitly allow.

I mean, yeah, isn't this normal behavior for every major video game as well?