Category Archives: waa

New for 2011! Standardizing the Definition of View-through

It has been over 17 years since the advent of the Netscape Web browser in 1993 and almost as many years since the first AT & T banner ads were served on Hotwired.com. Back then, the Internet was heralded as the most accountable medium ever.

Fast-forward to 2010: the digital advertising industry has gone mainstream and will likely generate more than $25 Billion (US only). At the same time, a subject that concerns far too many people is declining or flat click-through rates. Last week’s gushing “news” from a rich media vendor that clickthrough rates have supposedly leveled off after years of decline is a good example. 

Definition of Insanity
In a business that obsesses about such meaningless metrics, the digital advertising industry simply cannot continue worrying about click-through rates. Although most will recognize that the novelty of clicking banner ads has largely worn off, this measure which still provides almost no insight on the effectiveness of most campaigns just won’t go away – regardless of marketing objectives.

In the no man’s land somewhere between the ad server and site tracking is an analytics oddity called viewthough: a useful albeit tortured metric. Testament to this is thats not one of the major industry trade groups recognizes viewthrough by including it in their standards glossaries:  IAB, OPA, ARF or the WAA.

Despite early work by DoubleClick, plenty of practioner interest and ongoing research by ComScore, the digital advertising industry still somehow lacks an official definition of a viewthrough. At times, it seems like we’re all too often measuring what is easy or expedient. And, clearly that is not working for the industry.

Here is a sampling of viewthrough articles over the last 10 years:

  • Lilypad White Paper Response Assessment in the Web Site Promotional Mix (2/1997) A very early attempt by the author to describe the phenomenon in the context of measuring ad response; see the diagram.Online Awareness Model of Banner Advertising Promotional Models. At that time there was not yet a way to measure such passive behavior.
  • Conversion Measurement: Definitions and Discussions (9/2003) An early article that focused on “people that ultimately convert but did not click”. Technically, viewthroughs are not people and are probably better described as visits. Also, depending on the campaign objective, a conversion event is optional.
  • Neglecting Non-Click Conversions (11/2003) A pretty thorough piece on the subject, although the term “viewthrough” is not used and there is again an emphasis on conversion.
  • Lies, Damn Lies and Viewthroughs (8/2005) Again, the focus is exclusively on viewthrough conversion, which is clearly a trend. However, it is misleading as it misses all the non-converter traffic.
  • The Most Measurable Medium? We Still Have A Lot To Do! (9/2007) David Smith actually made a literal plea for the industry trade groups to define viewthrough. A great idea, unfortunately it fell on deaf ears and several years later not much has changed.
  • Why view-through is a key component of campaign ROI (9/2010) provides a more balanced look at what viewthrough is but still brings up conversion. Also, th acronym “VTR” is confusing as that is what most people might consider a viewthrough rate similar to how a “CTR” means clickthrough rate.
  • Different Views of View-Through Tracking (10/2010) More of the same, although this article actually quotes Wikipedia (scary) and further convolutes the matter referencing a Google Display Network definition that focuses on viewthrough conversion. Consistent with the theme, the term VTR is used to mean viewthrough conversion rate not view-through rate – two very different measures. On the upside the potential of viewthrough for media planning and optimization was right on.

Curiously absent from the ongoing discussion is what viewthrough inherently represents: measurable incremental value from an affirmative self-directed post-exposure response. With just syndicated panels and qualitative market research to divine results, traditional electronic media could never quantify this .

At the same time, the advertising industry now has over 10 years of similar “directional” qualitative research focused on the familiar yet ephemeral measurements of post-exposure attitudes and intentions (notoriously unreliable). Many see these brand lift studies as rife with data collection challenges and ultimately of dubious value. Just this year, Professor Paul Lavrakas on behalf of the IAB released a critical assessment of the rampant practice.

Parsing The Metrics
It is bizarre that many digital marketers insist on defining viewthrough rates in conversion terms while clickthrough rates are always measured separate from subsequent conversion rates. Mixing metrics has confused the matter but effectively left viewthrough to be held to the higher standard of conversion. Ironic, since very few clickthrough (in volume and rate terms) even result in conversion.

While “clickthrough rate” is always understood to be relative to impressions (# of clicks / # of impressions), “viewthrough rate” seems to have skipped the middle response step and gone all the way to conversion. That doesn’t make sense when there are so many other factors that influence the purchase decision after arriving on a Web site.

To be very specific, viewthrough rate (VTR) should be similarly calculated, i.e. # of (logical) viewthroughs / # of impressions. “Logical” means that the viewthrough is observed where a branded post-exposure visit is most likely to happen analogous to the target landing page of a click-through;usually this means the brand.com home page. 

Measurement Details
The real problem underlying the apparent confusion is that viewthrough measurement invokes several raging and simultaneous, inter-related and often technical debates: branding vs. response, optimization, cookie-deletion, cookie-stuffing, panel recruitment bias, correlation vs. causation and last-click attribution. Anyone one of these arguments can cause a fight.

Nonetheless, in defining what viewthrough actually means it would be helpful to overview the two basic ways of measuring view-through:

  1. Cookie-based: This a browser-server technique that relies on cookie synchronization between the ad server and the target brand site. When the user receives the ad, a cookie is set on their browser that is later recognized upon visit to the target site, which is then matched via speical page tags back to the associated campaign. There are several ways this can be done, e.g.  DART for Agencies (DFA)/Atlas/Mediaplex page tags, ad server integrations (Omniture) or ad unit ridealong pixel tracking (Coremetrics). Optionally, PSA campaigns can be run alongside a camapign for a simultaneous test-control comparison of viewthrough “true” lift;essentially you can measure a baseline amount of viewthrough traffic that would end-up at the site anyway. Downside: subject to browser cookie limitations.
  2. Panel-based: Alternately, a standing Internet behavioral panel can be utilized, e.g. ComScore and Compete. In this approach, two comparable groups are observed: an exposed test group and an unexposed control group that represents the baseline viewthrough. The difference between the rate by which the test group (exposed to ad campaign) and the control group (received PSA or other’s ads) subsequently visit the target site reveals the lift that is explained by the presence of display advertising. This method may also include ad or page tracking, but does not require cookies. Downside: subject to panel bias.

The Impact of Time
Next, an additional layer to viewthrough measurement that is worth mentioning is time, i.e. delayed response. Like traditional advertising media, display ads exhibit an asynchronous response curve where the effect of the advertising decreases over time.In our real-time data collection world, it seems the common sense realities of human behavior are often overlooked.

Many factors can impact the viewthrough response curve, including messaging, frequency, share of voice and creative execution to start. And, one size does not fit all: a considered purchases could reasonably have longer shopping cycles than CPGs. Depending on the method of measuring view-through, typically 30 days or 4 weeks are often used as initial “lookback windows.”

Et Cui Bono?
Although that was fairly straightforward, as soon as viewthrough is connected to a site conversion (through deeper page tracking), the thorny issue of attribution arises (and cookie-based measurement is implied). Viewthrough measurement often goes off on a tangent t this point because there are two layers to attribution.

  • Channel attribution is simply put: which digital channel is assignedtr credit for the conversion event? Measuring display advertising happens to be more complex and most site metrics tools punted on tracking this capability. That means that simpler response channels like paid search, natural search, affiliates, CSEs and email to receive last credit as a default. For many marketers, measuring conversion attribution or participation gets complex and often political very quickly.
  • Media attribution gets really contentious, especially for lead generation and ecommerce-oriented marketers. Performance ad networks often insist on having their own special page tag in place where the conversion event occurs;in this way they can independently measure conversions and potentially optimize their ad delivery. The problem is that there usually are multiple ad network vendor tags on the conversion event page and all of them will count the page load as a conversion. Worse, this is an easy way for the ad network to shoehorn themselves a retargeting cookie pool. Unchallenged, media vendors may claim credit for everything such that marketers end up overpaying for the same conversion. Alternately, some very Byzantine schemes have arisen to guestimate credit. 

Despite all of the above, here is a working definition of a viewthrough for 2011:

Definition of View-through
Viewthrough is a measure of the passive but self-directed impact from a partiucular display ad unit (banner, rich media, video or audio). The viewthrough event follows one or more ad exposures and when the ad unit is clickable can be post-click (initial click visit timed-out) or post-impression (with no click). Importantly, a viewthrough may or may not be associated with a purchase conversion event but must be associated with a target page load or other high-value action. VTR or viewthrough rate is calculated as # of viewthrough / # of impressions.

Viewthroughs decay over time from ad exposure. In-flight viewthrough are observed during the live ad campaign while the post-flight “vapor-trail” begins immediately after the associated ad is served.

Don’t like this definition? Come up with a better one or edit the above…and, the sooner the better or the industry might get stuck with this sketchy Wikipedia entry.

Chicago Analytics?

I’m always interested in connecting (and reconnecting) with colleagues in the Chicago area; especially those working in the crazy field of online measurement. The other day I received an urgent request from Meetup.com, warning that the Chicago Data and Strategy Consortium meet-up was about to be canceled! So, I volunteered myself to prevent that from happening…

Lincoln Park Lagoon looking southeast.

Why? I personally, would like to see an informal group of professionals working in this nascent field that isn’t always served by other local groups like Chicago AMA, CIMA, national groups like IAB, OPA and the more software-oriented or social/networking groups. Ideally, less drinking and more learning. Thin overhead and easy to manage…maybe even invitation-only?

In what areas of analytics and research are people interested?

Contact me and let me know what you think…we’re all busy people and all at different stages of our respective careers.

http://www.chicagoanalytics.org