Hacker News Reader

Comment by salicideblock | Hacker News Reader

Seeing that Plausible Analytics is not using cookies and instead hashes fields including IP address - the measure of "unique site visitors" could be a proxy for false uniques by Plausible Analytics. For example, an user visits once via his data connection and once via wifi - Plausible may see different users where Google's cookie will pin it as one user.

Another interesting result is the 16% of total page views - I don't think false uniques could changes this. But it will be distorted by correlation of views-per-user with blocking settings.

View on HN

I think this is probably correct. I wrote my own hit-tracker for my website and I spent a lot of time looking at logs to see if I could reliably track unique users by hashing headers and the IP address. My conclusion is that you can't - you will always double count some users.

It is no uncommon for consecutive requests from obviously the same user to come in from different IP addresses. Usually these are either cell phone users whose network changes from second to second, or browsers behind big proxies with multiple servers fetching content.

For example, I would see this quite often (all fields made up):

  75.34.234.71 index.html mobile-browser-v25
  75.34.234.75 style.css mobile-browser-v25
  75.34.234.75 header.jpg mobile-browser-v25
  75.34.234.72 code.js mobile-browser-v25

Plausible would plausibly see that as 3 different users.

Another problem is bots. So many bots. Some that take the trouble to mimic real browsers - they don't know they are bots. And they have a plan.

Most bots don't bother to execute javascript but analytics that rely on single-pixel webbugs and the like will likely show inflated numbers.

Business users behind an HTTP proxy will be another problem, as they will all have the same IP, and be using the same OS and browser (mostly).

That turned out to be less of a problem than I anticipated. Browsers spew all kinds of mostly unique information into the User-Agent header. This is slowly getting better as the browser companies wise up.

Aren't Google planning on killing the User-Agent header altogether?

> they don't know they are bots. And they have a plan.

Nice subtle Battlestar Galactica reference there.

I don't see why the focus is on "unique" visitors. To compare to older media, if someone buys two newspapers then they are a double-good customer. Either they are spending so much time reading it themselves that they need two copies (ie reading via two devices/IP addresses) or they are giving that second copy to a friend (two people, one device). So imho page view should be valid regardless of uniqueness (absent bot detection etc).

This depends on monetization.

If you're selling copies of newspapers, 2 purchases by 1 customer = 2 purchases. The number of customers is irrelevant.

If you're monetizing via on-page ads or affiliate links (as the author is), revenue is much more likely to correlate to unique visitors than to page views.

>> revenue is much more likely to correlate to unique visitors

That's the bit I don't understand. Why does an advertiser not value a customer seeing an ad twice, or for a longer time? In fact I would say that those customers are much more valuable because their multiple views bookend an extended period of viewing an ad. If I view a website at home in the morning, then again while at work, I have been exposed to that ad for many minutes, as opposed to one-time readers who click away within seconds. That has to be worth something more.

You're on to more than you think.

In traditional media, it's pretty much an axiom that people have to see the same ad at least seven times in order to get the message. There is no sin in showing the same advertisement more than once. It's why, for example, advertisers pay extra to have their ad appear at both the beginning and end of a commercial break on television. It's called "bookending."

When digital advertising started, the ad companies decided that someone seeing the same ad more than once was a bad thing, and then convinced their advertisers this was true.

I don't know why this happened. Maybe the digital people had some data that proved it. Maybe the digital people were just computer people and not advertising people and so didn't know about the decades of prior research into this.

But it's where we are now. Digital advertising companies value unique visitors, so web site owners do, too.

I always thought this was mostly to prevent fraud. If you pay for multiple views by the same user, it's much easier to generate revenue from fake views.

I dont know why either, but to me viewing traditional TV with same ads every 15 min or 30 min is not an annoyance, but doing it on Youtube or Website is god damn annoying. To the point where I think the brand is trying too hard and I have negative feelings with it.

There are huge difference in Online / Digital Media consumption compared to traditional media like Tv and NewsPaper. And may be I am old I tend to value the latter a lot more.

The value is screen time. If a user spends ten minutes on a site and comes back on a different IP for another ten minutes they're twice as valuable even without knowing they're the same person.

Well, if I were buying an ad I guess I would pay more to reach 10 people one time, rather than 1 person 10 times. It would increase my chances of getting a receptive customer.

If the purpose of your website is to get people to sign up for the trail of your SaaS product, then you have very different reactions to "ten people looked at my product once and one signed up" vs "one person looked at my product ten times before finally signing up." In the first case, you work on improving your landing page. In the second case, you let people sign up for an email newsletter.

Afaik if you visit the site from your phone and laptop on same wifi they will count as 1, which wouldnt be the case with google' way

Reply on news.ycombinator.com