Trust them as much as you like

[pecunium]

6A announced, back in January, that they meant to implement some analytics.

They claim it's benign, me; given the recent flaps, I don't really care to have people poking about in my affairs.

This is the announcement in Lj Biz.

With this new change, we'll now also use Omniture on a very small random sampling (about 5%) of journals and communities, including profile pages, friends pages and comment pages. This change will take place on or after September 27, 2007.

Omniture is a website analytics service. The system will collect information that's pretty straightforward, including what browser you're using, what site scheme you use, your window size, how people travel through the site (what are the common links, where are people going after viewing their friends page, what people are or aren't clicking on), and things like how many page views different parts of the site get.

With this change we will be able to learn more about how you use the site and what areas are confusing or are in need of improvement. We'll also have a good way to help prioritize all of your suggestions based on what people actually use.

Some key points:

* We're only going to apply the cookie to a very small random sampling of users, about 5%.
* We're using the resulting stats to find out what to focus on in the future for LJ.
* The Omniture code doesn’t capture any private data such as payment information provided in the Gift Shop.
* Omniture does not have access to friends-only or private entries.
* You can opt out, and if you've already opted out, you'll stay that way.

As always, we are providing a way for any user to opt out of contributing to the stats-gathering (even though we know it runs the risk of statistically biasing our results). If you’d like to opt out, go to the Admin Console and type "set opt_exclude_stats 1". This opt out applies to the entire implementation of Omniture -- site-schemed pages and the new inclusion of journals, profiles and communities. If you've already opted out, you don't need to do so again.

We're looking forward to having more detailed data to help us make decisions about the best ways to improve the site!

If you aren't all that happy with the idea of Ominiture poking about when you are using other sites who happen to be clients of Omniture, go to this page and set the opt-out cookie.

For Firefox I had to manually set the cookie in my options (I've had some other places where cookies wouldn't set either).

That might seem belt and braces, but there you go.

Comments

[clownburner.livejournal.com]

If you're using Adblock, you can also just block everything from 2o7.net, and then you won't load their code no matter where it appears.

It's safer than the cookie, which you have to trust them to honor.

[starcat-jewel.livejournal.com]

Hmmm. Would adding that domain to my Hosts file accomplish the same thing?

[clownburner.livejournal.com]

no; you can't add the whole domain to the hosts file, so that's impractical. If you run your own DNS server, you could black hole it that way.

[starcat-jewel.livejournal.com]

I checked with my partner, who's the hardware-savvy one in the household. He says that if you have a router, you can identify the IP ranges used by 2o7.net and add them to the router's block-list -- and that he'll be doing that as soon as we're back from this weekend's con. The effect of doing this is to send any attempt at communication between your system and 2o7.net, from either side, straight into the bit-bucket.

[clownburner.livejournal.com]

That'll work, but since your browser doesn't know that 2o7 won't be responding, it may wait for those connections for a while, slowing performance.

Not happy.

[starcat-jewel.livejournal.com]

In order to opt out from Omniture, I would have to add them to my "trusted sites" list, which I am reluctant to do. And doing it in Firefox won't help, because I normally browse using IE.

[kodi]

5% is a "very small random sampling"?

[antonia-tiger.livejournal.com]

Nope.

Thank you

[dragonet2.livejournal.com]

done.

CDK, if there are enough members, 5% can be a very good sample.

I'd rather not have them ruffling about my files. I use Safari and am not sure what can and can't happen that will go through my browser firewall.

[kodi]

I'm on board with "very good sample." I'm puzzled at how it can be a "very small sample." Does that make 25% a "small sample"?

[pecunium.livejournal.com]

It's a piece of PR. If they said they were collecting information from a "statistically repressentative random sample," people would go, "huh?".

If they say, "1 in 20", people might go..., "whoa!"

But, "a very small sample" makes people feel secure that it won't be them.

TK

[antonia-tiger.livejournal.com]

There's something either incompetent or fishy in this.

I went to the Omniture opt-out link you provided.

When I tried to set a cookie, it told me my browser wasn't set to accept cookies, so I checked.

It is, and my browser shows 29 cookies from them. I have set my firewall.

At least the admin console seems to have worked without errors/ Am I too paranoid to think that Mandy Rice Davies applies?

[don-fitch.livejournal.com]

My guess/assumption is that their intent is relatively benign (if "relatively benign" includes "figuring out where best to put paid advertisements") and might even be part of an attempt to improve their services by discovering how people use their lj accounts.

My impression is that The6A are Dedicated Computerist types who'd rather do such investigating with automated software than by something a human as just asking people what they use lj for and what changes or improvements they'd like. (Granted, a random sampling (5% sounds reasonable, to this non-statistician) makes sense if they're planning on doing something to attract many random New Users, but I'm not sure that would be a good idea.)

Me, I haven't even looked at my lj site more than half-a-dozen times since I got it some years ago to avoid having to post comments anonymously and to be able to read a few "friends-locked" postings. But I do find myself echoing Teresa Nielsen Hayden's remark "I deeply resent the way this administration makes me feel like a nutbar conspiracy theorist."

[yuripup.livejournal.com]

The usual political poll you see in the news usually contacts no more than 2000 people. Using, as a ball park, that there are 300,000,000 people in the US and of that 225,000,000 are 18 or over (guesstimate from the CIA Factbook) that means for +/-3% accuracy they are using 0.00000009% of a percent for their results. (Polling 5% of the US population would mean asking over 15,000,000.)

A number that is somewhat smaller than 5%.

There is a flipside. I am sure the LJ population is smaller than 225 mil--so even a same sized sample will be bigger percent wise. And I am sure they want a large enough sample so they can break down subgroupings of LJ users into statistically relevant chunks.

[don-fitch.livejournal.com]

Everything you say here sounds reasonable and is (AFAIK) accurate, but ... I'm thinking along somewhat different lines.

Calling this a "poll" seems to be imprecise and possibly misleading. It is, to my mind, going to be more of an "analytical study" -- the kind of thing with which the more factual data you have (up to the point of absolutely ridiculous redundancy), the better (and more impressive).

Political (&cet.) polls, as I understand it, usually involve paying people to conduct interviews, or expending money on postage, so there's a significant cost-per-response factor and the numbers are deliberately kept as low as possible consonant with estimated (or plausible) reliability, with considerable effort going into selecting a representative, rather than purely-random, sampling.

The Project we're dealing with here seems to be almost entirely a mechanical/electronic collection of factual data. The Big Expense -- possibly 90+% of it -- will lie in developing and setting-up the software. This is a Fixed Cost whether two accounts or 200,000 are to be examined. I think it's obvious that I know little about computers, statistics, & sampling and am only speculating, but logic suggests that it would be necessary to follow the selected Accounts 24/7 for at least two weeks (maybe three, to include a Holiday Weekend), that a single dedicated Big/Mainframe computer might be able to handle up to about 5% of the lj Accounts, that this was the cheapest package the survey organization offered, and that The6A would opt for as much data as they could get for the price. Both the logic and the speculation may, of course, be wildly wrong, but I don't see this as being the particularly important aspect of the situation.

Whether The6A are going to be asking the right questions is another kettle of fish, and crucial if, for example, they include searching for Content at all -- such as using a List of Naughty Words, or flagging the use of f*ck*ng asterisks mid-word. Most likely, presenting a longish list of things they will be searching for, and leaving it open-ended, was (typical) incompetence/cluelessness re nuances of verbal usage[1], but it does permit a certain amount of Creative Interpretation.


[1] After all, we're dealing with six people who insist on defining "interested in" as meaning "support and practice". Sheesh! Hey, I'm interested in Republican Party Politics, -- but purely as a matter of self-protection.

[pecunium.livejournal.com]

But that poll is expensive, and tailored.

They get a list of names, which are distributed accross the country in question (to get regional demographics). They look to economic distribution. They look to educational variety.

And they have a bunch of names in each of those categories, when they get enough responses, they stop asking that chunk of the list.

What 6A wants is a broader sort of thing. They aren't looking for a set of answers to a specific set of questions, but rather of broad trends (at least that's the desires, as advertised).

Because self-reporting is skewed (most people aren't willing to cop to things they think the questioner will think "wrong" which is how "push-polling" works), 6A has to have a larger sample; and not rely on response (if they happen, by chance, to get three people who are all into David Hume Kennerly[of whom there are four on Lj], the resulst would be off (esp. if they were having a conversation about him).

So a wider net has to be cast, but I don't know if 1 in 20 is a reasonable number.

I happen to think not.

Flip side, if all the people like me opt out (and Optimune can't tell 6A what percentage has opted out) then we will be invisible to 6A, and things like the breastfeeding icon flap (which shows some of the problems in 6A arbitration of complaints) and the Great Banning (which highlights a couple of different problems, related to agenda driven campaigns on content, and the CYA corporate structure of the present enterprise).

I think I ought to make a poll; non-relevant as it might be.

TK

[cjmr.livejournal.com]

Okay, so if I have three LJs, does that mean I have to do this three times--once logged in as each me? Or can I use one of the other methods above to have Firefox do it for me for all three at once?

I use LJ over one of the other blogging sites because I'm not especially computer literate and it has been easy to use, damnit!

[delicata77.livejournal.com]

Thanks so much for posting this.

[yuripup.livejournal.com]

I am iffy on opting out. Its not like I am paying for what I am doing here--and since LJ is a business they have a reasonable expectation to profit off my being here.

That said, were this a paid account I would opt out in a second.

[don-fitch.livejournal.com]

We might /a/s/s/u/m/e/ hope that the presence of your comments, and mine, and those of other Free Account members, contribute sufficiently to the consumer satisfaction of enough Paid Account Members that we are of some value (if only marginal) to both the Group and the Organization. But, yeah, that's one of the reasons (aside from being lazy) that I'm inactive om my Home Page.

[pecunium.livejournal.com]

The problem with this is, so far as can tell ist that Moniture isn't limitig itself, to just LJ.

Checking my cookies folder (the one I was told omniture couldn't load a cookie into, until I made it an "allowed" site), I had about a dozen cookies from them.

And they aren't collecting just on/for Lj usage.

So it's not so much that I don't want 6A to be looking at my Lj usage (which is one thing) as I don't want Omniture looking at my web usage, just as I don't want people looking at the books I check out of the Library.

TK

I've gone from fat dumb and happy

[dragonet2.livejournal.com]

to aware. My Safari cookies folder was immense. I just nuked it. I wish there was a way to program it to ask me if I WANTED to accept a cookie, there doesn't appear to be one. going to firefox now to see if that's an option. If it is, bye-bye Safari.