What Does A Display Advertising, Retargeting Campaign Look Like?

Retargeting(Make sure to read the follow-up to this article that looks at "below the fold" impact on the AdExchanger.com retargeting campaign.)

Shield your eyes!

AdExchanger.com is "opening the kimono" on its recent retargeting – or do you say "remarketing" – campaign.

Disclaimer: the following should not be perceived as an endorsement of the vendor used (Google). It’s meant to be a simple, real world example of retargeting in action.

To reiterate, when I say "simple," I mean it.

AdExchanger.com was only driving awareness with this campaign as the bar was set very low in terms of success metrics. For vendors waiting to tell me how wonderful you are and that you can provide much better retargeting services, no doubt you can. And, that’s fine with me.

Why Remarket/Retarget?

In general, the basic goal of remarketing is to reach out to users who have already visited your website by buying display advertising placements on other websites.

In AdExchanger.com’s case, the goal was to continue to drive awareness with display ads, have a little fun, and encourage users to “like” AdExchanger.com on Facebook.  If the ad is clicked in this retargeting campaign, it takes the user directly to the AdExchanger.com Facebook page (see it here!).  But, they have to click the “like” button.   I could not track exactly how many “likes” were generated, which was not ideal. But, again, fine with me – this is a straightforward campaign about building awareness.

Here’s a smaller version of the 728x90 advertisement that AdExchanger.com used in its retargeting campaign (still running as of today!):

Banner

Certainly, whether for retargeting or not, awareness campaigns can go much deeper in providing useful statistics and insights such as those provided by any number of brand response vendors (Dynamic Logic, Insight Express, Knowledge Networks, Vizu, etal).

Also, the clear winner in remarketing are eCommerce retail sites that can run direct response retargeting campaigns such as those you might see for Amazon, Ebay and others and where a direct line can be drawn between an impression, a click and a purchase – cookie wars excluded.

Pixeling For Browsers

In order to create a targetable instance known as “the cookie,”  AdExchanger.com has a piece of Javascript provided through Google AdWords which is embedded in the HTML of the site and sets a cookie in the user’s browser once the user visits AdExchanger.com.

Assuming the AdExchanger.com cookie is not deleted or blocked, whenever the user’s browser visits a website with a placement that is using a Google AdSense or DoubleClick For Publisher 728x90 display ad tag (which is enabled for the auction), the placement then "recognizes" that the browser cookie matches AdExchanger.com’s request to retarget and AdExchanger.com enters the display ad auction for that placement.

Bidding

AdExchanger.com bids at a fixed CPM of $20+ for all display ad placements.  Why so high? First, this is a second price auction.  If AdExchanger.com wins, it will pay the second highest bidders bid. So if the second-place bidder in an auction for a single website ad impression is a $3 CPM, AdExchanger.com would pay a $3 CPM.

Next, AdExchanger.com’s users are a (in Internet terms) small, yet powerful group numbering 35,000 uniques in October.  AdExchanger.com wants to reach them – bid high I say!

Bidding

Finally, there’s no way to bid according to recency. The "recency" parameter tells the bidder where in the session the user is – did he or she just sign on to their web browser? If so, they might be more prone to clicking or seeing ads.  AdExchanger.com’s theory is that with no ability to understand recency through AdWords, bid high!  Perhaps Google or the retargeting vendor will show AdExchanger.com’s retargeting ad sooner?  But, that’s really a guess on my part.

A brief riff on "recency": this parameter could end up being another unwinding of publisher CPMs, too - perhaps this is why Google doesn't address it through AdWords (let alone another bidding parameter which adds more complexity). If bidders are only willing to bid for impressions early in the session, they may not bid later. Of course, one could say that media buyers will bid higher than ever for early session impressions so it all comes out even in the end? Who knows!?

Cookie Lexicon: Users and Membership

Let me tell ya…. Cookies are deleted by the user or degrade (whatever you want to call it) at a very high rate.

For example, at the end of November, the total number of cookies that were targetable through AdExchanger.com’s AdWords account was approximately 150,000 (called “Number of Users” in AdWords) which is significantly more than the aforementioned 35,000 monthly uniques. AdExchanger.com began pixeling users for retargeting purposes in January or February of 2010 give or take a month.

Also, another important variable to consider when retargeting is a  “lookback window” or as Google calls it with its white-glove lexicon: “Membership duration.”  This is the number of days that a retargeter wants to keep the cookies it has collected and try to retarget the display placements which the cookies can help identify. Google lets you store for a maximum of 540 days. From my perspective, with a site like AdExchanger.com and its simple goal of driving awareness, there’s no reason to not try and retarget a cookie for the maximum amount of time allowed.

Once again, if this was Amazon.com with a direct response focus, Amazon likely has a good idea that you can keep a cookie around for 4 days for retargeting a potential book buyer, or 7 days for a blender buyer, or two weeks for a PC laptop buyer, etc.

More Targeting Params!

In that AdExchanger.com likes to reach all of its users worldwide, nearly every AdExchanger.com user possible is targeted via geo-targeting parameters set in Google AdWords.

Also, depending on the robustness of the advertising platform, a retargeter gets to decide how often a user will be shown an ad - a.k.a. frequency capping.  You can target every day with multiple ads if you like and it’s apparent how this could get freaky for the end user as it seems possible to spam the hell out of people. (I always think Google has solutions lurking that they do not divulge - hence "seems".) On the other hand, do web surfers just get used to it OR do they need controls which let them opt-out? OR all of the above? All of the above!

AdExchanger.com chose to target with one ad per week per user.  In other words, not much – enough to keep the awareness bubbling.

The Results

In seven days, from November 28 to December 4, 2010, the results were as follows:

  • 14,255 impressions bought
  • Nearly 3/4 of all impressions bought were 728x90
  • 89 clicks for a clickthrough rate (CTR) of .62%
  • Average CPM of $6.95
  • Average CPC of $1.11
  • Total Spend $99.06

So with over 150,000 user, Google AdWords cookies available since January-ish, only 14,255 impressions were bought. Whoa!  That’s a lot of useless cookies. Or maybe Google didn't find all the cookies that were available? -that seems less likely. Google, like any ad exchange or display ad supply source, wants me to spend! Or maybe AdExchanger.com was losing a bunch of auctions? ...Ya think? Nah.

One way to have increased the ability to target users is to obviously have targeted across multiple ad exchanges or supply sources (a separate pixel for each) through a DSP or ad network which will manage your cookie (or “membership”) pool across all supply. Still, this result suggests the limited value of cookies over a long time period.

And how about that CPM of $6.95? Seems pricey except that we’re talking about a spend of only $99 in a week. It would follow that as the long tail of websites with valuable audiences gets accustomed to retargeting, they will be bidding high with little impact on their bottom line and as a part of their overall marketing toolkit.  The aggregate of long tail retargeter spend could drive up publisher CPMs – as long as the publishers have the valuable audience. Same holds true for the impact of mid and short/fat/wide end of the tail retargeting spend. The mid and short ends can pick their spots for retargeting whether on behalf of an advertiser (reach extension) or for the site itself.

But looking at the $6.95 another way - hey, there are a lot of second price auction bidders who are bidding high! - at least for AdExchanger.com's audience on the DoubleClick Ad Exchange (the auction marketplace which is accessed through AdWords).

The 728x90 versus 300x250 fill rate is also an interesting story. With 75% of all ad impressions bought going to the 728x90 display ads, it would appear that DoubleClick Ad Exchange inventory is heavy on the 728x90. But, that's another guess. For those who are curious, there were 4 728x90s and 5 300x250s running in this campaign with the ad rotation set to "Show ads more evenly." This should force even delivery of all ads available for each size as opposed to taking advantage of optimization available through AdWords.

Finally, whether you like clickthrough rate data or not, the .65% CTR data is heartening and one of the reasons any marketing services company such as a media agency will institute a retargeting campaign as a first step with any client. The results look good!  But, it’s the scale of the campaign that remains the challenge.

By John Ebbert

10 Comments

  1. John...thanks for the insight...always helpful to see what others see!

    Do you have any idea of how many uniques you actually reached....how many of the 150K Google was able to reach?

    Oh and I'm pretty sure users who read adexchanger.com might not be a normal sample when it comes to cookie attrition 🙂

    Reply
    • Anonymous

      Thanks for the question, Matt. If Google frequency capping is perfect, I would guess that I reached 14,255 uniques (taken from 14,255 impressions) for the entire week given that i had the campaign frequency capping set at one ad per week.

      Reply
      • No it was not.

        Frequency capping was 1 impression per "ad group" per week. There is/was only one ad group containing all the ads.

  2. You've hit on one of the dirty little secrets of audience-based advertising... the difficulty in finding users on whom you've previously dropped a cookie. To extend the lessons learned, this is why you should never buy data on a "per stamp" basis... because most of the users (or stamps) won't ever be found again.

    Reply
  3. You've hit on one of the dirty little secrets of audience-based advertising... the difficulty in finding users on whom you've previously dropped a cookie. To extend the lessons learned, this is why you should never buy data on a "per stamp" basis... because most of the users (or stamps) won't ever be found again.

    Reply
  4. Eran Metzer

    Few points to consider:
    1. Do advertisers see the competition on every auction to know that a true 2nd price auction is the case? For the above example, a $20 bid resulted in $3 can hint that some sort of rounding was done by the exchange (modified 2nd price auction) as most predictive algorithm rarely predict a bid float bid with zero cents.
    2. Recency is mainly referring to how recent did the user see the same ad, that is being retargeted. in many cases this is a critical attribute - as we all know, there is data decay and data saturation. You don't want to hit the same user 30 seconds after they saw the same creative, yet you don't want to limit that for 3 days.
    3. Frequency is coming to help here, but most of the frequency is measured as how frequently the user browsed through sits that are using the same ad-server as the buyer. You rarely encounter true user frequency on the internet (the overall exposure since the user opened the browser)

    All of these are addressed by some advance technologies that exist today - yet some of the players don't really have them and are hiding behind the confusion created for marketers.

    Reply
  5. Great Post John...always great to see real numbers! I'd bet the preponderance of 728's in DFP/Adwords is due to the pubs dumping below the fold remnant into the exchange (bottom of page). Med Rec's like 300X250 often are hard to place anywhere other than above the fold/upper right...and usually sell better as result. I'd be curious to see you place a DSP pixel and acquire your same audience via their more sophisticated bidding engines (i.e using a Dataxu, etc) on a variety of exchanges (Rubicon, etc) to see how prices are affected. Also, did you (or will you) run a comparison campaign with "no-targeting"? In my experience, DR heavies often find "tonnage" is still beating targeting for price & performance (overall)...would love to see that A/B.

    Otherwise, keep up great work!

    Reply
  6. Ohad Gliksman

    I really liked the case study.
    Just a quick point that comes to mind is that a lot of adsense content units, only allow text based ads and not image ads which might severely limit the exposure you can get over Google's network. However this still does not fully explain the campaign's results.

    Reply
  7. Tom Anderson

    Regarding pacing across ad sizes, often there are two 728X90 per page and one 300X250 plus 300X250 in general receives ~20% higher CTR over other size ad units likely due to where the 300 lies in the center of most pages and often ATF.
    Many 728's are buried in footers that are rarely InView. Measuring InView at the site level is a valuable metric supplied by a couple companies and is more valuable than a simple ATF/BTF measurement.
    Eran hit squarely on the other salient points.

    Reply

Add a comment

XHTML: You can use these tags: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>