Monday, February 25, 2013

Proposed Viewthrough Definition

Below is the proposed definition being considered by the Digital Analytics Association Standards Committee. Although officially, only DAA members feedback count - all feedback will be considered.

Please provide comments below.


  1. ADD a calculation for viewthrough visit rate = viewthrough visits/impressions delivered; viewthrough conversion rate = viewthrough conversions/impressions delivered.

    ADD two points about viewability: a) that viewthrough is commonly confused with viewthrough but that b) they are related: "Viewthrough is most meaningful as a metric when it is calculated based on viewable impressions.

  2. I agree with the prior poster, the actual calculation of the metric needs to be added in order to provide a framework for the standard. A viewthrough could be on a variety of metrics:

    Viewthrough Browsers: Browsers who viewed an impression outside and organization without clicking on the impression.

    Viewthrough Visits: The sum of visits where the Unique Browser of the Visit viewed an impression outside of the organization without clicking on the impression.

    There are multiple metrics that may be derived from the Viewthrough metrics - ratios, percentages, absolute counts (Unique Browsers) and sums (Visits.

  3. Some sort of calculation or rate seems prudent, but depends on what the metric is trying to accomplish. If the intent is to show the value of display advertising, then it may seem weird to explicitly exclude conversion events following a click. If the point is to show a rate that says X% of my impressions were followed by a conversion (where X is total impressions / total conversions preceded by at least one impression), then you at least start to get the story, but it's not a replacement for real attribution which would show the value of each. I wouldn't necessarily conclude that one placement is better than another because it's viewthrough conversion rate was higher, there could be a lot of other factors.

    I also agree that viewability could be taken into account, but if so that should be called out as a different metric, otherwise you are in danger of having vendors providing different results for the same data set.

  4. Some guidance on determining an appropriate window for Viewthrough measurement could be useful. Again determined through controlled experiment with PSA or simulated control group and studying decay curve of incrementality with time from exposure.
    A caveat on known inaccuracy of Viewthrough connect rate due to cookie deletion between exposure event and visitation or conversion, should also be kept in mind vs clickthroughs where this is less of an issue. Proof: Increased frequency tends to show increased viewthrough conversion rates even from PSA ads. Reason - with increased frequency there is lower chance of cookie deletion based disconnect.

  5. :( these comments make me sad. Keeping in mind that this was a call for input on defining standards, I expected to find well thought out and clearly explained input...

    1) how many times did I reread "...viewthrough gets confused with viewthrough"? Please don't include any such business in the definition. Don't even move in this general direction. Define it once clearly and concisely, do not depend on "is not" as part of the definition. It's at terrible rabbit hole to fall into.

    2) Do not use the term "browser" to define anything except the client-based technology used by visitors to browse or "view" the internet. This is already a well accepted definition, and viewthrough should not attempt to re-purpose existing terminology while defining itself. This rule should be applied wholesale to any predefined/accepted terms. No good can come of trying to change what's already set by the community.

    3) Perhaps I'm alone in my thinking, but I don't believe anything more than a pseudo-algebraic formula should be used for the calculation. Unless, of course, you want to have to redefine yourself and your core with each advancement in technology and technique. It immediately dates you (the way disco soundtracks immediate dated all 1970's movies). Stick to principles and concepts. Convey the meaning, don't prescribe the method of calculation. This forces you [the definition writer] to be more thoughtful of your wording and writing. It also forces us [the readers] to think a bit about what is being discussed and reduces the risk of some jackalope running off to tool it out verbatim and without forethought.

    Besides, by purposefully avoiding the prescription of how to calculate, you allow for [even encourage?] discussion on the topic, which in turn should spur challenges to common opinion, and hopefully improvement on technique.

    Bottom line: defining a concept (like viewthrough) should not be dependent on implementation.

  6. 4) I actually like Vikram's point on providing guidance for defining windows. Again, don't prescribe them, but perhaps the definition of a viewthrough should include some discussion on how factors impact it's usability, accuracy, and reliability. As far as I'm concerned, identifying dependencies and explaining their impact is wholly within scope of defining a term.

    5) Avoid naming specific technologies and whatnot when defining your own term. Partially for the aforementioned disco-plague, but also because you create gaps in principles. For example, Vikram's point of cookies and deletion may have been completely valid five years ago and considered still relevant today, it'll likely be completely irrelevant with the adoption of websockets across all major browsers. So, if you mention cookies in your definition (or any other specific technology, implementation technique, etc.) then what does it mean for other unnamed technologies/implementation techniques/etc.? Was your omission deliberate? If so, then what does that mean for viewthrough on technologies that don't utilize cookies at all? Are viewthorughs, then, impossible to achieve/measure/ocure at all in volatile (i.e. non-static, non-persistent) mobile app that uses no local storage for user data?

    Finally, some words on the existing content:
    - maybe I'm old school, but I don't think you should use short-form nomenclature in the definition (Ad vs. Advertisement).
    - Are viewthoughs limited in scope to advertisements? I'd consider promotional art, videos, or even audio tracks (think Grooveshark music) should all qualify as mediums for traffic generation.
    - to belabor the point a bit, I think you tread on prescription when you name specific mediums in association thereof
    - I'd like to see the "sending side" reworded to specify content delivery
    - and definitely the receiving side shouldn't be exemplified by a reporting tool. it should be defined by/as whatever is used to view the provided content. Reporting doesn't play a role in describing what viewthrough is - it's just how we measure it.
    - you're not writing a position paper; there's no need to convince the reader why your definition matters. You can definitely include that throughout the site, but it has no business as part of the definition itself.
    - Merriam-Webster says incrementality is not a word and I agree. Don't create new words to bypass writing things out properly.

    Hope that helps!