you’re right. Server activity and audience research are different things… Too often online automated tracking data, and server logs are called ‘research’ but actually, they should be called something else. Activity data? I see it like putting a ticker on the M50 toll bridge, another on the Tunnel, and some more on the major roads… and using this information, trying to describe how big and busy Dublin is. You have to exclude those who pop in and out, cars and lorries and try to work out the number of people. Its all there for the counting, and you’ve counted everything but its very difficult get meaning from it. Compare this with door to door research, or a census, and you have sample research, and registration on a site. Audience measurement typically use samples (as opposed to panels) and these all do different things and tell different stories.

Activity data and research should also be broken down by category. Offline, does a bank compare itself to telephone usage, or road usage, or magazine readership and The Late Late? Course not. So why do Comscore et al do this online. Why have search, banking online, recruitment, youtube, msn, ireland.com and RTE together. They are different. Different markets, models and users.

Banks should be with banks, search engines together, UGC and blogs in another category. Then travel and tourism… and so on. We have to learn to treat online with the same cop on as offline, or soon becomes ridiculous.

I’d say Northern Rock has a lot of ‘traffic’ at the moment…, and Lloyds less so. What’s the real story?

@Emmet – I’ve actually visited deviantart, but I can’t see how the average user would have heard of it!

Perhaps we are looking at two different things: one is web user behaviour and the other is web user activity. Web user activity is easier to measure from the webserver log files. Behaviour is not so easy to measure so it is necessary to extrapolate from existing data. I am aware of how audience and market research works.

The technology and software to track users across sites exists and it is relatively simple. The problem is getting access to the webserver log files. It would result in a large database but it there are certain optimisations that would work. The increasingly fixed nature of IPs make tracking usage easier.

I think that getting an article like that in the Sunday Tribune news section was the real intention of Cybercom. There was a balancing article by Damien in the business section but few of the people who believed the news section article will have read it.

As regards the Irish web, I probably would have a far better view of it than most advertising research companies. I would have serious reservations about trusting Comscore’s data when it claims that the nih.gov is one of the most popular sites for Irish visitors. Perhaps you had better publish your top 30 quickly. However for the Irish media, it had better be printed on drool proof paper.

Two things. What’s bursty? Secondly, tried and tested mathematically sound audience research methods have little or nothing to do with sociology. It’s just market research. Circa 2 billion euros was spent last year in the display advertising market, with similar predictive research approaches. On radio (day after recall with very large sample), TV (small sample of TV tracking), newspapers (very large sample primary research of header recognition and recall) and outdoor (large offline ad recognition and recall). These other media use this research to provide demographic insight into media consumption and percentages within an acceptable statistical range. If you call anything other than log files ‘just sociology’… so be it. It’s just not though. (cf. http://en.wikipedia.org/wiki/Sociology). It’s called audience research, and it’s what advertisers and media owners use and need. Its broad brush stuff, there’s no doubt, but it works really well, and the findings are right within one or two percent. The demographic insight is invaluable. 35 million euros will be spent on online display advertising this year for the Irish market (not including search and classifieds), so we need this insight for informed media decisions.

Statistics are a bit like magic. You talk to 1200 people the right way, and their opinion will be true for a huge population. When we hear Gordon Brown has grown in popularity by 1% it’s based on this small sample but will be true for a population of over 50 million. Even if several companies do similar research they get the same findings, within a few percentage points. The bigger the sample, the more you can drill down. Strange but true.

Don’t get me wrong… Research isn’t everything. Server log activity does have its place as a measure of activity, but it won’t tell you how many people use a site, or anything about those people. NB use server logs to get a measure of how busy site servers are, so we can gauge what impression will be used if we make a booking, and then tracking takes over. Bigger markets which don’t have a JIC, just let the experience of site advertising command the price of that advertising, and we do that too. We know that even if a site has a gazillion impressions, it may give few clicks per thousand and post click and view behaviour may not be what the client needs, so spend will migrate to perhaps a less busy site, with more suitable user behaviour. It’s a natural process of optimization, rather than one of research. It’s not predictive either, it’s anecdotal. So, for the Internet, research is only a small but important part of the big picture of the large scale digital media buying process. It is a digital medium after all. We can’t do this with other media, and if we could, we would, and the predictive measures that exist for these media wouldn’t be so important.

You are also sort of right about the number of sites we track. Its such a fragmented space that we can’t do in-depth demographic exploration of the many sites with a tiny number of Irish visitors, but, as least we know they exist, because the market told us, and we know that they are smaller than others in terms of Irish Internet traffic. If we place ads on them, and it works. Great. When it comes to the larger entities though, like messenger, ethnic sites, banks and insurance companies, search engines news and information site, travel and airlines, sport and the many other major categories, we have considerable insight from this research approach. Insight that server logs couldn’t touch.

Server logs tell so little. There’s no demographic information; you can’t compare server logs across sites and unique users aren’t people. Even if you could fix these three things (which you can’t), logs aren’t available independently across the many sites you’d be interested in. Net Ratings and ABC work hard to fix this fact, but it’s still a fact, and working with many small ethnic sites shows us it’s likely to remain a fact unless some new magic technology comes along to sort it all out.

Digital advertisers need to know how many people visit their site, their competitor’s sites, and how many Irish people exist in their market online. They also need to know lots about these people. Do they drive? Are they students? Do they have kids? And then, how many of their market visit the sites they might place advertising on. I only know one way to provide these insights, and if there is a better way, cool. I’ll take whatever is out there that makes some sense of behemoth of a medium. But I would say… the methods we’re using are so similar to other media. Everyone, except you, can’t be wrong.

What annoyed me about this article is that the headline said it had information on Irish surfers. It promised accurate information on how Irish people use the web, but it just didn’t give it. To the general reader, it was confusing and misleading and the data had no source. Talking about Disney, a paper doll site and the US department of health and saying this is what Irish people spend most of their time doing online is just really, really silly and it made the Irish Internet user look silly. It did little absolutely nothing for medium or those involved, and it just didn’t describe Irish Internet behaviour at all. I mean seriously, I have to ask. Have you ever been king.com, or stardoll.com, or deviantart.com? Do you know anyone who has? This article gave the impression that you should have been there at least once, along with all your mates for between 20 and 30 minutes (how recently it doesn’t say). If you have been to these site, I’d be somewhat worried actually. Oh, and HMV.co.uk isn’t as important a site as iTunes or CDWow for the Irish user. It just isn’t. This article reported measures of the wrong thing and it communicated nothing. And just at a time when advertisers are spending for Christmas too.

I can see just see it. The US department of health suddenly notice an unexplained spike in advertising queries. ‘What does this graph mean Smithers?’ ‘It’s the Irish Sir…they say we’re really popular over there…’ ‘They’re trying to get citizenship by pretending to be sick are they? Sneaking in through the banner ads eh?’

I advise damage limitation at this stage. Maybe a little clarification, keep the head down and wait for the passing of time.

I’ve been in this game a long time, and run Monitrack Internet Research, which originated the JNIR and Net Behaviour, Ireland’s largest online media buyer. Here’s my two cents.

I find the numbers in the Tribune article totally incredible, and there’s no methods given. They also seem to be counting the wrong thing and confusing people.

For example, NIH.gov (no.5) is a US health site. Irish people spend more time on it? Really? How many people? Maybe a few sick people, but very few. What is Stardoll.com? (10 years as an online researcher, and never heard of it and here’s why…). Its a paperdoll site.

This isn’t valuabe information and the lists also don’t make any intuitive sense. As there’s no methods given, its difficult to see exactly what’s wrong, but there is a problem. With the most visited websites the problem is the usefullness of the information and again, methods. For example, if you key in Google.com, you get bounced to Google.ie automatically, so why are they both there? Am I counted twice? Who knows.

We in Net Behaviour have been doing a survey over the last year and now track 500 sites. We use it to help with media buying for several of Irelands largest agencies and advertisers. We’ll be launching the data soon, but in the interim, you might like to get some toplines from our blog http://www.thetannoy.com, on Ireland’s top 30 and other bits. We’re showing the total number of people who visit sites in a given month (not unique users, or alexa browsers, or ABC audited server data). The data is based on over 1200 questionnaires from collected from 30 random online sampling points (sites and search engines). Its real time data too.

In my experience, advertisers and marketers are interested in customers and readers, and these are people… Unique users aren’t people as one person can be several unique users. Alexa measures are based on the number of people with Alexa browser downloads, so its not random, mostly US too, and has no demographics. Net Ratings don’t really do demographic data, though they have a go at sex, age, and location (based on probability and household composition). Comscore is the same.

The net is a truly global medium, and any research has to count ALL sites, not just the ones in Ireland to be valid. It also needs full demographic information to give meaning. So, the net isn’t like Irish newspapers, or radio research or other joint national research. It is also a space that has tectonic changes every six months or so. Bebo and YouTube.. came out of nowhere when you think of the net three years ago and Google bidding and Adsense changed net advertising totally in about a year. We also advertise on MSN chat, podcasts, and recently mobile phones, so, now we have to research all of these things alongside news and infomation sites for display advertising.

Any digital or Internet research has to aim for total flexibility, and absolute independence to stand a chance of staying impartial while keeping pace with the changes online. It must stay statistically valid, useful and meaningful. The methods of data collection need to correct, robust and transparent. If not, data becomes a misleading mish-mash of errors, half-truths and sales patter, confusing people and damaging the credibility of the medium.

Emmet

What comScore is doing is estimating (or exatrapolating) web vistors based on the activity of its panel. The website owners via log file analysis provide actual web visitor statistics. There is a bit of a disconnection between the two. One is an extrapolation that might provide some trends and indications – the other is hard data. The comScore data may well be more accurate than the JNIR figures – I think that JNIR had to restate figures. Because I work with web/domain stats on a daily basis, I have an inherent distrust of extrapolations especially when it comes to web traffic.

Perhaps contacting websites and requesting Irish visitor figures and then comparing the comScore figures might work better than providing the comScore figures on their own.