Necessary Waste in Advertising

In online display (banner) advertising, over 99% of advertising impressions do not lead to a click. Some percentage of impressions (let’s say 5% in lieu of more rigorous analysis just yet) are probably noticed by user and have some impact on their subsequent behavior. But that still means at least 90-95% of the advertising that is put out there can be judged by the advertiser to be “waste”.

My company, CPM Advisors, is in the business of helping advertisers reduce that waste and find media that will perform for them per their goals. We are usually able to do a very good job in limiting waste through a variety of techniques that i will not expand upon here at length, but encompass choosing the correct sites and types of placements on those sites, turning things off quickly that do not work, aiming for the right user frequency, etc. etc. and doing this in a rapid and coordinated fashion using technology. It is true, however, that some waste is always going to be necessary for the advertiser to be sure they are maximizing their potential to find their target customer.

If everything seems under control, you’re just not going fast enough. – Mario Andretti

Unfortunately there is no perfect information about what users want or are likely to buy. Thus every presentation of a good or service that is put before a consumer is a probabilistic question – what is the likelihood that this person is interested in my service (or, to be more clear – start with – what is the chance this person is going to see my message at all?)? This applies to search keyword purchases too – though there is a lot more “signal” that the user provides about their intentions and state of mind. Many companies make a living by analyzing and exploting the differences between things like someone typing “car insurance plan” and “car insurance plans” – each one may have a very different probability of click and conversion. And so, really the story of advertising is a story about estimated probability of action, and the margin of error of those estimates.

When you slice and dice your campaign – I want to show these different ads with different messages, on different sites, at different times, you cut up the data and you then require more time (and usually, more money) to be able to gather enough data to be able to say something conclusive about performance outside of the margin of error. If you cut things up too many ways too quickly, you lose a lot of signal and are awash in noise.

Don’t get me wrong – we see REAL waste all the time in advertising that we fight against and work to stamp out. Ads that are never seen by a user, or are in a bad place that never gets any clicks¬† (we literally had one wellknown top-20 site that we discovered never gets any clicks on its ads — never, no exaggeration, and we checked if it was a technical issue and it wasn’t — so our system won’t buy there anymore). There is enough of that inefficiency and waste to be driven out of the system to keep us all busy for a good deal of time, but we won’t be able to get rid of all waste since that will mean we’re not doing our best to find customers for our advertisers.

We now need to apply our minds to thinking about how we can figure out what the “right” amount or an “allowable amount” of waste might be, to set client expectations accordingly and work to build on top of actual performance, meaningful data that can cut down future waste not just for a client in specific but across all companies that advertise online.