Counting The Numbers That Count

Image courtesy

The internet is good at numbers. You'll never find yourself short of data in digital. So in amongst a blizzard of stats, the real job is often to understand the metrics that really matter - in other words to count the numbers that really count.

And the internet is big - it's easy and cheap to achieve scale which can also mean that it is easy to chase volume over meaning. If you are looking to attract a relevant, quality audience to your website, is it better to have a high proportion of your user base driven by Search Engine Optimisation, or to be facilitating (and an active part of) a vibrant, engaged community meaning your users visit more often, interact more, contribute more, and spend more time. If you are looking to make money out of your website, is it better to have a significant proportion of your traffic driven by relatively low value content such as photo galleries, or engaging high value content such as features, interactive tools, community content.

The internet is a does medium. People do stuff on it. Meaning that the best metrics are often those that relate to people doing what the internet is best at - interacting. But human interaction is complex, so the way in which we measure it cannot be over-simplified. Take the much maligned banner ad. When I was working in digital ten years ago click rates were typically around 1%, sometimes even reaching 2%. Within a year or two that had declined to more like 0.5%. By 2004, typical click rates were somewhere around the 0.3% mark. A study just released by ADTECH of more than 10 billion banner inquiries across Europe, reveals that the average click-through rate now fluctuates between 0.11% and 0.19%.

One way of looking at this is out of 1000 ad impressions you serve you're still getting 1 or 2 people clicking. And if you want to buy 1000 ad impressions you can do so very cheaply. Another way of looking at this is that a 0.11% click rate is a 99.89% failure rate.

You could conclude from this that banner ads just don't work. Or maybe they're just not working hard enough. The idea of having commercial real estate positioned within the context of editorialised content is not such a bad one after all. Either way, making the formats more interuptive (by placing them right in the middle of feature text for example) is not the answer. But perhaps making them more relevant and more useful may be. If I'd been on the look out for when my favourite band was touring for example, an ad that was served to me telling me that they were playing at my local venue, and tickets had just gone on sale, and here's where I can buy them, would be useful.

So layers of targeting that benefit from sophisticated use of contextual, behavioural and profile data  might actually make this kind of advertising genuinely relevant. Some affiliate arrangements already deploy enormously sophisticated levels of tracking and optimisation against defined cost-per-acquisition metrics. But even then, we are in danger of attributing everything to the last click, and very little to any other form of effect, or to any brand-influence or other advertising that user may have been exposed to. Too many banner campaigns are measured on click-through alone. It is not all about the click, and the last click is certainly not everything.

So think twice before taking the big shiny number that sits right in front of you. Data tells it's own story. It might just be that you're stopping at page one.

Comments

[Jon]

'If I'd been on the look out for when my favourite band was touring for example, an ad that was served to me telling me that they were playing at my local venue, and tickets had just gone on sale, and here's where I can buy them, would be useful.'

This would probably require using a different model for online marketing. Instead of a Broadcast strategy (the 'natural' evolution of TV ads) we would need a strategy much like direct sales.

Consumers would be more likely to click a banner ad from just about anybody if they perceived it to be one that they themselves had requested. If you could tell the web that generally, you're interested in music and concerts and these bands then you might see an ad you'd want to click.

The problem with the current intrinsically determined advertising is that it's missing the mark. I would absolutely love to see a Steely Dan show, but I will never see an ad for one because my browsing history doesn't reflect that. If I could open up my last.fm data to Google Adsense (for example), then my chances might go up but if I could Explicitly state that I like Concerts and specifically 70s bands it's a near surefire thing.

Er, excuse the ramble. Enjoyed the article.

[eaon]

some good points.
also i agree with the commenter above that releasing data from the likes of last fm/spotify or other online apps could deliver more relevant ads.
however, i keep coming back to permission. there's still a big role for 'brand' advertising provided the person being advertised to has granted permission to receive it.
I love the Nike big brand films for example, and through the web and/or recommendations from my network I can easily seek them out.

[Le'Nise Brothers]

I agree that increasing the level of targeting can lead to better quality conversions and instant post-click activity and it is definitely true CTRs are falling and that the majority of clicks come from a small minority.

However, something else that needs to be considered is the role that display ads and indeed, other digital channels make in driving the awareness that leads to the desired end result (ie. traffic / sales / sign-ups).

If you look at display advertising in isolation, the results will look weak. However, moving away from 'last click wins' thinking towards path to conversion / engagement analysis will assign a value to each channel and depict a true picture of what led a person to complete the desired action.

[neilperkin]

Some good points here - thanks.

Eaon, Jon - I've seen some interesting technologies that use people's profile and preferences (like Last FM data) to enable them to shape what other content (like editorial content) comes to them

Le'Nise - Good point and I agree - plenty of research already around to support the role display has in amplifying search.