Commentary

Duration-Weighted Impression: Is It Worth It?

A new study published this week finds that most ad execs are unaware of, or do not really understand, duration weighted impressions (DWI), a controversial provision among the industry's proposed cross-media audience measurement standards, which are under public review until June 25th.

According to the new standard, all cross-media ad impressions should be weighted by their exposure time in relation to a fixed 30-seconds denominator. In other words, a 15-second ad that was viewed to the end would account for half the impression value of a 30-second view, while a 6-second ad would register 20% of the impression value.

The controversy surrounding the new standard comes from the implied linear relationship of duration and advertising effectiveness, and the danger that such an oversimplification will take advertising backwards -- especially when it comes to marketing productivity assessment. Let me unpack these concerns.

First, from a research point of view, the relationship between “exposure time” and cross-media effectiveness is murky at best, even in some of the research cited in the new standards document.

For example, in a well-known study of TV exposure time from 1993, Surendra Singh and Catherine Cole concluded that “a blanket comparison of :15 to :30 could lead to misleading conclusions because length effects depend on message appeal.”  In 2014, in his research on the “rising cost of consumer attention,” Thales S. Teixeira emphasized that  although easier to measure, duration is not a good proxy for attention.

A year later, Goldstein, McAfee and Suri (2015) wrote that “If the total amount of time a publisher gets is divided into slots, two short ads increase memory per slot over a single, longer duration ad.”

MMA´s own SMoX research into the modern marketing mix -- multi-touch attribution (MTA) studies with 14 brands -- have found that duration doesn’t have a consistent relationship with effectiveness. For instance, in our Allstate MTA study, we observed that a 15-second mobile ad was up to twice as effective on a cost basis compared to a 30-second ad -- a finding that in and of itself invalidates DWI.

More importantly, the central theme from all the SMoX studies is that cross-media effectiveness is a complex, multidimensional relationship which cannot be captured by one factor, especially duration. For example, targeting usually works as a multiplier boosting effectiveness of the same creative by 2X-5X.

Ad placements may also impact the effectiveness of the same ad. In one study, a pre-roll video was found to be 80% more effective on a cost basis than an in-banner, auto-play option of the same video ad.

Finally, effectiveness also varies by marketing objective. When foot traffic is the goal, desktop and mobile tend to be more effective on a cost basis than TV, but the opposite is true when optimizing for brand consideration.

H.L. Mencken once said that for every complex problem there is an answer that is clear, simple and wrong. DWI falls in that category. It's appealing because it’s simpler than other approaches to assign cross-media value, but it invites advertisers to use it as a shortcut and to bypass the hard work of attribution.

That's important because as of late last year, at least 6 out of 10 marketers didn't have an MTA solution in place. Those that did were only applying MTA to one-third of their media spend.

So, although the proposed cross-media audience measurement standards clearly do not recommend using DWI as a stand-alone effectiveness measure, marketers will very likely use it as one. And those who do will sacrifice ROI because their reallocation will not be based on real effectiveness data.  

For all these reasons, we need to find alternatives to DWI that also work given the current infrastructure of media measurement.

For example, we can’t even track the second-by-second ad that viewing DWI requires in linear TV yet. So, let's continue building granular measurement for all media, including TV.

In the meantime, we need to invest more time in attribution and assess what really drives effectiveness.

4 comments about "Duration-Weighted Impression: Is It Worth It?".
Check to receive email when comments are posted.
  1. dorothy higgins from Mediabrands WW, June 21, 2019 at 11:08 a.m.

    The oddest part of the duration weighting is simply that we do not revalue impressions based upon unit length of linear TV or radio for audience measurement/delivery purposes.  So to do this with digital video is, well, interesting.  That said, we DO have a duration weighting or equivalizing approach based upon :30s for price and value comparisons.  That most certainly needs to be applied to the new digital video and OTT currency mechanisms. 

  2. Tony Jarvis from Olympic Media Consultancy, June 21, 2019 at 3:22 p.m.

    When is an Impression not an Impression?
    Sadly this commentary reflects a fundamental but critical confusion (or flaw) in terms of the interpretation (deliberate?) of the nomenclature being supported by MRC in these proposed Cross-Media Audience Measurement Standards [sic], i.e., "viewable impressions".  This term is being consistently equated with the measurement of "gross impressions" an accepted audience measure which is driven by a respondent's proven opportunity-to-see content and NOT, as with "viewable impressions", a device measure driven by the rendering of content on a screen whether in the presence of a person or not and whether that present person is "viewing" or, as used in OOH currencies, has an "Eyes-On" or not.  Yes, its subtle but a very important value difference to advertisers.
    Mr. Bakapoulos even refers to, "a 15-second ad that was VIEWED.." and then, "a 30-second VIEW"!  This dimension of actual viewing is not accounted for even if the content is correctly rendered for a couple of minutes.  "Viewing" or "Eyes-On" or "contact", notably for ads, is also only loosely supported by Nielsen people meters that merely require a person's acknowledgement of "presence" near the TV every so often.  Mr. Bakapoulos' reference to "exposure time" which could be restated as "Eyes-On time" is appropriate as long as it is exposure time for the target group versus merely content rendered time albeit a fundamental pre-requirement for a proper derivation of audience gross impressions.  
    Duration-Weighted Impressions as proposed by MRC requires considerable more study.  As  correctly pointed out "effectiveness" results in a complex mix of marketing and media factors that go so much beyond duration.
    Perhaps the industry and MRC should remember that distribution is not necessarily receipt, receipt is not necessarily an audience impression and an impression may not necessarily be an actual exposure.   So, maybe the industry needs to simply stick to mandating the measurement of exposure?  Without it there can be no effectiveness of any ad. 

  3. John Grono from GAP Research replied, June 21, 2019 at 9:18 p.m.

    As usual Tony - spot on.

    Duration weighting is essential for content.   When buying TV the duration weighting - which expresses the audience as 'average minute audience' - provides the most likely audience for your spot as the buyer (who might be buying 8-weeks out) doesn't know which break in the programme the ad will appear or which position-in-break the ad will appear, as well as relying in recent prgramme averages.

    The alterative measure would be programme reach.   You might have 10 million people watch a programme of which 1 million watch the full hour, 1 million watch for a few seconds, and the other 8 million are somewhere in between.   But if you use the 10 milion ... the buy would be a shambles.

    In essence if you used a 1-second threshold for a one-hour programme your would have a 1-in-3,600 chance of hitting the one-second viewer ... and a 1-in-60 or hitting the viewer of just one minute.

    When it comes to ads if you use a 1-second threshold you have a 1-in 15 chance for a 15-second ad, and a 1-in-30 for a 30-second ad.   Better but still not good.

    However, the 'effectiveness' of the ad can be highly driven by the creative execution.   If the 15-second ad is a teaser with a sting in the last 2-3 seconds (e.g. brand name reveal), then viewing of the first 12-seconds would be ineffective.   If it is a 'reminder' ad the a few seconds would do the job it is meant to do.

    The thing is, there is NO golden rule that would apply to all ads.

    1-second is way too short for the vast majority of ads.   15-seconds would be overkill for many ads.

    Further, to the best of my knowledge, there is no meta-data attached to online ads (or should I say non-linear distributed ads) that define how long the ad is.   Plus they don't have to be in the 15-second, 30-second, 60-second structure of TV which is encoded into the unique key-number (I know - old school, but it works for a VERY good reason).   Without verified data as to the duration of an on-line ad then duration-weighting will be problematic, but would IMHO be a usable proxy based on behavioural data.

    Personally, I like the idea of "served, rendered, completed" as key reporting parameters.

  4. Tony Jarvis from Olympic Media Consultancy, June 24, 2019 at 7:15 p.m.

    John:  At the agency we always checked that the content was delivered to spec as part of reporting to the client.  Not an audience measure of course.  There is another article on this subject by Joe Mandese with some great comments.  Commended to you!  Needs your further insights.  
    https://www.mediapost.com/publications/article/337405/duration-weighting-shouldnt-be-controversial-but.html?utm_source=newsletter&utm_medium=email&utm_content=headline&utm_campaign=114359&hashid=m1KG0ratr5oQVJUGsEAKemLsvUI

Next story loading loading..