• http://searchengineland.com Bradd Libby

    “Consider an apparel retailer which sells several lines of clothing including menswear, womenswear, kids clothing, wedding and bridal wear, footwear etc.”

    Judging from the title of your chart, I assume that these clothing lines do not intersect. :)

  • http://www.rimmkaufman.com George Michie

    Bradd, you are funny!

    Siddharth, always nice to see folks pay attention to the data. Well done.

  • http://www.efrontier.com sidshah

    Bradd: Perhaps in the future we can do a special post on cross-dressing campaigns. Unfortunately, the volume on those is low . I wonder why :)

    George: Thanks! I think you will find the upcoming post interesting too.

  • GE.GAO

    Hey Sid,

    This is really an interesting post and I happen to be testing around this concept around the past month.

    The punchline is: the conversion duration is actually an issue of revenue distribution. in other words, how do we setup cookie life to relate a sales to an certain ad?

    Here is the problem I had during my test.
    The conversion duration is product based. We know an expensive product could take longer to convert, which can probably be measured and thus given a longer cookie life. However, the ads pointing to this conversion can vary.

    Using your example, we can setup a longer cookie life for the wedding dress campaign for sure. But the ads in this campaign may end up with an underwear conversion. On the other hand, ads on a different/broader keyword(especially brand term) may produce a wedding dress conversion. In either situation, longer cookie life in wedding dress campaign does not help in accurate revenue distribution.

    I can think of two ways to solve this problem:
    1, Accumulate enough data to discount the wedding dress campaign revenue to adjust the cookie life.
    2, Associate actual conversion revenue to the exact ads, meaning dynamic/multiple cookie life, which is technically sophisticated .

    While this is a potential can-do, whether it is worthwhile or not may vary by cases.

  • http://www.efrontier.com sidshah

    Hi Ge,
    Thanks for your comment. My suggestion is you can view the click to conversion time on a campaign by campaign basis and decide what should be the cookie life. As part of the test set a really long cookie life like 30 days and record all conversions. You can then assume the time taken to record 95% of conversions as the cookie life for the campaign. Whether your ads contains broad keywords or not is not so important as making sure you are attributing the revenue to the ad correctly on an overall basis.My next post will cover this in more detail.

  • GE.GAO

    Hmmm, failed to make myself clear.

    When I mentioned broader keywords, I meant keywords not in this Wedding Dress Campaign.

    Say the site is called Edress.com, brand keywords like “Edress” in Brand Campaign and general keywords like “dress” in General Campaign may generate a good number of wedding dress conversions. To attribute the revenue to Wedding Dress Campaign correctly, we need to filter out such revenue from actual wedding dress conversions. When Wedding Dress Campaign keywords end up with casual dress sales, which may generate less revenue, we may also need to filter out/adjust such revenues. The quest for accurate revenue distribution may be costly…

    Looking forward to read your next post on this issue. :)