News & Press Releases
Hitting the sweet spot: going big vs. being effective
By Alex White
What makes a great communicator? Is it mastery of all the pertinent facts? The skill of a turn of phrase or use of engaging metaphors? By all means these are skills necessary to be good. But greatness comes when each listener feels that the speaker is communicating “directly to me,” even surrounded by even thousands of people.
That’s kind of the way it is with online advertising. Marketers want to go big, reach the biggest audience possible. But it has to be the right audience receiving the message otherwise those ad dollars will go to waste. Understanding and mastering that tension – macro/micro, buckshot/laser, Yao Ming/Spud Webb – is key to targeting success in RTB.
In my current role I often hear companies ask, “How many categories is the right amount in a taxonomy?” This is an interesting question and there are a ton of opinions out there. The IAB has put forth a recommendation as to the structure and makeup of a taxonomy that aims for standardization. It is recommended that companies in the contextual or semantic space follow at least the first two levels in a hierarchical structure, and further granularity is at the partners’ discretion. Google (of course) has their own approach, a taxonomy that’s derived from their keyword search products. What they both have in common is that their taxonomies go deep – too deep.
The nodes have it, or do they?
Companies often tout the size of their taxonomy, seemingly for bragging rights. To me it makes little sense. Say you could find some category in a logical tree structure (if it’s presented that way) out of some 3000 categories. The tiny volume of impressions that can be purchased by targeting the last node in the tree does not make it worth anyone’s time or money. Back in the day I worked for a network that offered a self-service interface for buying contextual categories. We had over 5000 categories in our taxonomy, but no scale in any one category other than the top levels. As a network this made it challenging to engage with clients effectively, and there was not significant repeat business for the product.
Now, things are a little different with RTB. Scale is less of an issue than it has ever been, however if you think about it, having excessive granularity just makes no sense to an advertiser. True, it seems appealing to target only the narrowest topics, but at end of the day all you really succeeded in doing is wasting time and energy setting up campaigns that will fail from the get go. Granularity becomes the enemy of scale. There needs to be a balance between the number of categories and appropriate granularity. When you need a magnifying glass, you don’t use an electron microscope. Talk about overkill.
The semantic advantage
You see, the problem is that those massive taxonomies are actually a legacy of the keyword-targeting approach to online advertising. Once you realize that it becomes clear why these (oftentimes competing and inconsistent) taxonomies have gone haywire. There is a limitless set of keywords that advertisers can choose to target, so it’s axiomatic that the taxonomies that support that approach have to be pumped on steroids.
In contrast, semantic based methods used to derive categories are more accurate and actionable than keyword based methods. Because this approach looks at the entire page, it is able understand the difference between phrases that challenge contextual targeting, like ‘Paris Hilton’ vs. ‘the Hilton in Paris’. This is because semantics looks at the words and how they are structured in a sentence and paragraph to identify patterns of how the words relate to one another
Once you understand the value of a semantic based system vs. a keyword based system, you will understand that keyword based systems are actually less useful to advertisers.
The bigger the nodes, the harder they fail…
Take the 80/20 rule (please!) Let’s say 80% of your scale is coming from 20% of your categories. That means that you have 60 categories of scale (or value), to an advertiser.
There are roughly 10B impressions available through RTB. Even if you could classify 90% of that (I only know of one that can and does), your looking at 9B impressions per day. Doing the math, you’ll see that 80% of those impressions will be contained in the 60 or so categories of scale, 25 of those will be top level. That leaves you with 35 meaningful categories and 2940 fragments without meaningful scale or value.
So why the obsession with nodes? It comes from a few things. One is that people like to think that ‘bigger is better’ or ‘the more the merrier’ and so on, and this may be the case with many things, like scale of impressions, or the number of QPS that can be achieved. Another contributor may be from a legacy of keyword targeting. Once buyers become more comfortable with newer, more efficient data to base their decisions on, we’ll see that expansive taxonomies, with their largely irrelevant deep-level categories, will be considered what they actually are: useless.
From iMediaConnection