Display advertising, which at one time looked like the web’s
glorious channel is now a glory hole for direct marketing. In many respects
brand advertisers have themselves to blame, not believing that a click is
branding and fumbling around with poor microsite driven strategies that have
zero ROI. The fact of the matter is an interaction experience anywhere (in our case the click) is an opportunity to
create an emotional attachment with your product or service.
So while half the blame for the display implosion rests with strategy the other half rests with technology. Listening to the display
industry of late you might believe that the worst is over – that a second
coming of display is upon us. Emerging from the wreckage are two saviors that marry new technology with emergent strategy to blend performance with brand. Let’s call them Thing 1 and Thing 2.
Thing 1: The new mode of buying display ads is about buying an
Thing 2: The new mode of buying display ads is about using data
If you recall in The
Cat in the Hat, Thing 1 and Thing 2 fly kites around to amuse the advertisers err, children. Ultimately the kids capture the Things and the Cat in the Hat cleans
up the mess that was created. Then they all disappear.
In our story Thing 1 and Thing 2 will not clean up the mess that
is display advertising either and the disappearing part is ominous. I will follow up soon with a post on issues around data and optimization (in the meantime here is a great look at the challenge from AdExchanger.com) but first I want to first examine Thing 1 – the idea of buying an audience vs. the old way of
is a group of people collected together in a single place and time. They don’t
need to have similar attributes but they share some common interest defined by
expressed actions or attendance. The presence of audience is by itself worthless
(unless they have bought a ticket – subscription & pay4play models). The
value resides in the emotional state that allows for the success of persuasive
efforts. This disposition is almost always temporal in nature and as such is
something that is most effectively targeted in realtime systems (like search). This
is one reason why display consistently fails.
Audience information is useful to target messaging that
triggers a response but the underlying emotional state is best mined deep
inside the publisher’s content. Search works in a large part because it is
content driven. But make no mistake, the dynamic content delivery controlled by
the user is still present in publisher environments. What is not built in is
effective solutions to derive revenue from the goals of the user and leverage
that moment (the kairos).
So if at its core an audience is about interest, intent and
action how does this fit in with the buying models now being touted? Data and observation
have taught us that online demographics & personas are meaningless for optimizing
performance. The web is a user-controlled medium so just like the actions that
define people we need to segment and target based on what people are doing, not
who they are (as defined by what they’ve done).
Don Norman calls this Activity-Centered Design. He also rightly points out the weakness of human centered design approaches to dynamic
sequences — the very backbone of
the event driven medium that is the web. My own optimization work has only
validated these same ideas qualitatively and quantitatively across dynamic
content and advertising. People are not predictable, but their actions can be a
window of predisposition to certain messages and information.
we are being honest with ourselves as an industry the only reason we are moving
to audience buying after a dozen years is that impression-based buying/selling
has run its course. The problem with buying impressions in display is only that
no one actually clicks on the ads. The strategies and platform are of a
different time and medium and consequently the ads suck. If we knew CTR would
be this bad no one would have ever built this type of system.
On the back-end most of the problems stem from the different
goals each participant (publisher, advertiser, visitor) has. This causes
multiple inefficiencies or optimization conflict. Display as it exists now is
further expanding the delta between the value of publisher’s visitors &
content with the revenue generated from third parties. This is not sustainable
business for anyone. Eventually in digital optimization no one wins unless
The funny thing with impressions is that when the ads are
good you can actually give away the impression for free. In Search the goals of
the users (relevance) the advertiser (ROI) and the publisher (revenue) work in
lockstep, each helps optimize the success of the other. Search validates the
idea of activity targeting vs. audience targeting. This is realtime and dynamic
— what people are doing and what is effective in persuading them to do
something. Proof that audience
targeting isn’t needed? Besides language the Google homepage is the same in all
233 countries and has grown to dominate global search.
Cleaning Up: As the
web has become more complex band-aid after band-aid has been added to try and
make display work. Everyone has
known about its poor performance for a decade yet few have done anything about
it. Those few that have successfully created new ad platforms have not
surprisingly had huge exits. However that innovation is now 5-10 years old. The
“lost decade” for VC in digital advertising was due to investment in layered approaches to an underlying faulty system rather than brand new ad models (though entrepreneurs surely share some of the blame).
It’s never too late to build new systems and it seems the need is becoming
more obvious. Just yesterday Bernard Lunn of RWW made his siren call. It is my belief that in order to work new solutions must have a ground up approach where
the core inputs are content and realtime events and the trigger mechanisms are visitor attributes. Just like the timing needed for ad persuasion, the current rise in APIs, realtime data, IR, semantic tools and desperate
publishers signal that the time may now be right for a different conversion event.