Accuracy: It’s Not Just A Song By The Cure

If you’ve ever extolled the virtues of the total accuracy of online metrics to anyone, you’ve been guilty of lying and you’re going to burn in hell. I’ll be there too though, so try and find some good indie music and maybe some cheese dip and try and drag those along.

In web metrics, discrepancies abound. If you’ve ever run two sets of metrics on the exact same data, you’ve most likely received two totally different sets of numbers. There are reasons for that, of course, but they may not make sense to someone who isn’t overly technical (or bright, like many of my past clients.) So when we’re telling someone that yes, he or she should funnel more cash into online marketing because it’s really easy to get accurate stats, what we’re really saying is that yes, he or she should funnel more cash into online marketing because the stats are probably more accurate than offline measurements, but they’re still not REALLY really accurate.

Like many of our SEO arguments that we make to each other and to our clients, the accurate measurement argument is a relative one. You obviously cannot accurately measure the number of viewings of a highway billboard, for example, even if you knew the exact number of people who’d driven past it in a certain period of time. If any of those drivers/riders were like me, they’d have been paying strict attention to the immediate dangers of highway life (bad song on the radio, band of wandering midgets who’ve gotten loose from the circus, those dusty plant things that roll down the streets in Westerns, etc.) You may have circulation numbers of a periodical but you have no clue as to whether the subcriber even opened it. You can’t tell when someone’s turned down the volume on the radio when a commercial is aired.

Online measurements, at least, are a bit more easy to conduct, since a user has to do something in order to trigger his or her action to be recorded. However, there’s quite a lot that goes on with the programs used to measure this raw data. Depending on your session length, for instance, you could count a visitor once or twice. When I used to, ahem, cloak my sites, I realized that the analytics program I used was counting each hit twice, counting the cloaked page once and the actual page served to the user once. Another analytics program I used on the exact same sites did not do this, though.

So, if we can agree that web metrics are MORE accurate than most other forms of marketing measurements, what difference does it really make if they’re still potentially not actually all that accurate? If we’re used to estimates from keyword software and we use those as a guideline for how to best utilize our content to capture the highest ranking spots and the most converting users, why can’t we be happy with inaccuracy in terms of how many people visited our site this month?

The New York Times has a great article about the dangers of online web metric inaccuracy, one which lays out the major problem with this: advertiser money. The following bits are the most telling…

“Many advertisers pay Web publishers each time their ad gets an impression, meaning that it is viewed by a reader, but each company uses its own methodology to count impressions.” So think about it: if you’re being told that your ad has been viewed 10 million times, you’re going to be charged for that. What if it had truly only been viewed 8 million times? You’re paying for 2 million views that never happened.

For major advertisers being lured by sites with the premise that their viewership is massive, this kind of discrepancy can also add up to tons of lost money, whether it’s being lost by inflated stats or by the promise of a high readership that turns out to be untrue. Speaking about site owners attempting to gain major advertisers, Gian M. Fulgoni, the chairman of ComScore, states that “It’s in their interest to make their audience look as big as possible.”

“To make matters more complicated, consumers who delete cookies — small bits of computer code that track their online wanderings — are also overcounted by publishers’ servers, by most accounts. Some news sites have tried to improve their systems by asking their visitors to register, but many people refuse.” I delete my cookies all the time, especially when I get bored. I also refuse to register on most of the sites that ask for it just to be contrary. I’m obviously adding to the problem here but I see no way around that.

Should we have to settle for constant inaccuracy? Probably so, if we’re dealing with any form of marketing. I’m not really all that concerned if I get two sets of visitor numbers for the same site, but what DOES seriously concern me is that my conversions may not be accurate. Currently I run conversion tracking on a small set of paid ads for a client, and he can verify that those numbers are correct because the conversion is on a contact us form that he receives. Currently, that’s about the only bit of accuracy that I’m enjoying though. It would be nice if that changed, but I’m not too optimistic about it right now.

Liked this? View all posts in Web Analytics

One Response to “Accuracy: It’s Not Just A Song By The Cure”

  1. Congrats. I think focusing on our BS – albeit airing our dirty laundry – is a great thing. Here’s my pet peeve about lousy statistics. Much of the SEO industry seems to be built around the WordTracker methodology. But there are two nonsense pieces of statistics in it.
    1. Their data on searches comes from extremely minor (weird) sampling (follow my signature link to read about it). So it’s not representative of the general population. While it’s better than guessing, it’s far from accurate.
    2. Their whole theory that google references equates to competitiveness of term is….silly. I haven’t blogged on this but it’s easy to disprove. Pick a color. If you google “red”, you’ll find there are over 1 trillion references. Sounds very competitive. Yet, there is not ONE paid ad for that term.

    I couldn’t be the first one to have noticed this, can I?

Leave a Reply