| Last
week, David drew a comprehensive outline of what
should be included in the online process audit.
On the Spin Board, only one person could think of anything to add, which leads us to believe this is the most complete list of things you should do to ensure your agency passes the audit test. "Given that this is for an online audit, I think it should clearly be spelled out in the [Terms and Conditions] about impression delivery. Specifically what does the agency and client consider the gospel numbers, the advertiser ad server or the publisher server. Also, it may be helpful to always ensure clients have access to their information on the server. This way they can sort of prove everything was eventually delivered in case of make good situations," wrote Brendan Kelly. Can you think of anything else to add? Let us know on the Spin Board. |
Monday, April 15, 2002
Accounting for the Total Value of non-CPM Buys
By David L. Smith As direct response online campaigns become more prevalent, the need for accurately attributing results to media vendors has become a concern. An estimated 30-40% of users who see an online ad do not immediately click through to the site, but return on their own at a later time. In addition, the branding effect resulting from the serving of hundreds of millions and sometimes billions of impressions is not taken into account.The post-impression visit attribution has been verified not only by Mediasmith client campaign experience, but also by the Engage Online Advertising Report Summer 2001, which puts the figure at 36%. These post-impression users are highly valuable, as the awareness level required to return of your own accord is much higher than that of those that immediately click through. The same Engage Online Advertising Report Summer 2001 showed that post-impression converters are 60% more likely to be repeat customers than post-click converters. The efficacy of ad campaigns budgeted by advertisers is not being accurately measured in many Cost Per Action and Cost Per Click deals, since these post-impression users are often not credited back to the specific media. Most ad serving systems allow for accurate tracking of these users, however the cost of serving ads in a CPA/CPC scenario is often prohibitive. (The "effective" CPM on a CPA or CPC deal is extremely low, as a function of the number of impressions that must be served to generally make a deal pay out. As a result, 3rd party serving costs can represent too large a percentage of the total buy to be affordable). CPA and CPC campaigns are generally tracked using only a unique 3rd Party URL, rather than an image redirect html tag, as in a CPM campaign. While this method allows for detailed back-end tracking at the advertiser site level, it does not account for the post-impression user, as mentioned.
Leaving these users unaccounted for has implications for the advertiser, the agency, the 3rd party ad server and the sites.
The overall performance of an ad campaign can be improved dramatically by the identification of sites that deliver post-impression users, as it allows for more intelligent optimization, renewal decisions, and planning. These combine to improve the ROI through increased margins for the marketer/advertiser. This therefore helps to rationalize Interactive ad budgets in the long run. In addition, the long (and even short) term branding effect of the Web is being increasingly documented. Any CPA or CPC advertiser will verify the jump in "non-attributable" traffic that happens when a new campaign is launched.
Overlooking the post-impression data has led to the exclusion of significant commissions for the agency, effectively eliminating the profit margin on many Direct Response campaigns.
The wide variation from site to site of the portion of activities conducted post-impression can often lead to a miscalculation of site performance. Sites may out-perform others on a buy by a large gap when post-impression is taken into account, however these sites appear to be under par when looking only at post-click data. As mentioned in the advertiser section above, this impacts optimization, renewal, and future planning. Additionally, sites are not being fairly compensated for the activity they drive.
As pricing structures and campaign objectives continue to evolve in the online medium, we are always seeking solutions to improve ROI and accurately and fairly compensate all parties involved. The following details a process for ensuring this in CPA/CPC campaigns.
OBJECTIVE
Accurately account for the post-impression activity that occurs in a direct response campaign Account for the branding benefits that occur in a direct response campaign Maintain tracking and serving costs at an acceptable level PROCEDURE
It has been determined that serving the entirety of the vast amount of impressions required to meet a CPA or CPC goal through 3rd party adservers is generally inefficient and makes it difficult to cap costs. It has also been determined that accounting for the post-impression users is beneficial to all parties involved. This leads to the process of attribution described here, a stopgap measure until an affordable technology solution can be found. (Companies like DoubleClick have stated that they are conscious of this market and are working on product solutions).
The amount of impressions that would likely be served to fulfill the Cost per Action or Cost per Click components of the campaign needs to be determined. Of this overall impression estimate, perhaps only a small percentage needs to be served to accurately project the post-impression response across the components. What is this small percentage? It is one where the impression number is large enough so that given a creative/site CTR and site conversion level determination, a statistically significant number of actions exists.
For example: Client goal is $3.00 net per registration for a service. Assuming that the creative to be used has been tested and yielded a CTR of .6%. This creative and the client's landing page together yield a conversion rate of 10% across similar media. If the site is allotted $10,000 for the campaign, it should be able to reach the registration goal with 5,555,555 impressions. The minimum we'd like to look at for purposes of statistical significance is 166K. Assuming one placement on the site, and two creative, the statistically significant level for attribution is 50 x 2 x 1 x (1/.6%) x (1/10%) = 166,666 impression, where 50 is the minimum number of actions we might consider to be statistically significant.
This pre-determined number of impressions would be served through a 3rd party adserver using an image redirect tag. This allows us to see the percentage of post-impression users attributable to the particular site for this campaign on an average daily basis. Once the efficient impression number is reached, the site will then commence with serving the impressions and a unique clickthrough will be used to track back-end actions. In the example above, given an adserving fee of $.32 CPM, the serving fees for post-impression attribution testing purposes would be $53, vs. $1,666 for serving all the impressions.
For the remainder of the campaign, the number of post-impression actions will be determined by applying the ratio between post-impression and post-click actions from the initial run on an attribution basis.
For long-term campaigns, the image redirect tags will be re-implemented and the accuracy of the attribution level will be tested from time to time. In addition, if there is a change in creative direction, a significant change in the offline portion of the campaign, or a major site change, the attribution level will be re-tested.
When there are reasons to suspect that post-impression activity may be heavily influenced by outside "noise" caused by other advertising efforts, PR, or word of mouth, a post-impression baseline should be determined. This baseline will be the amount of activity caused by non-campaign noise, and should be subtracted from post-impression reporting. The baseline is measurable by serving non-brand/product creative, such as Public Service Announcement banners, to a small percentage of the campaign's impression, across all site/section/placements to avoid skews.
For major campaigns, clients are doing themselves a disservice if they do not tap into the branding pre and post campaign awareness testing services offered by Dynamic Logic or Millward Brown Intelliquest. The cost of these services is amazingly affordable given the value that can be derived.
Technology Solutions
Some other solutions for this issue are foreseeable in the near future.
For instance, the 3rd party adserver could create a pixel that is placed on sites using the company's publisher side solution (e.g., DFP). This pixel would be unique to the advertiser's campaign and be used for tracking. This would benefit the adserver in that, a) they will generate more adserver volume, b) sites that they currently represent will have an increased incentive to use the publishing system, since their payout numbers will be greatly increased on CPA deals, and c) they may be able to get sites not currently using their system to test under the auspices of adding more dollars to the bottom line for the site.
There is also the potential within an ad serving/ tracking system to have it serve impressions only to a pre-determined level (such as calculated above). During that period, an impression projection is developed, and once that has been calculated, the serving automatically reverts to site-side (external or its own publisher solution). This solution will need some minimal code tweaks on the adserver side but it is believed to be achievable within current technology parameters.
A more affordable ad serving solution may also be the answer, as the current cost of adserving is contributing to the inefficiency of the process. A partial solution may be to decrease the amount of information being collected on the back-end (limit pixels to 1 or 2), thus decreasing data storage cost. Another may be for the ad server to change the pricing model for such deals to a percent of the spend rather than a CPM basis.
Client-side metrics may provide a solution, whereby the site serves the ad while the client-side engine does the counting. Adserving costs are therefore redirected completely to the site, leaving only the counting, software and service, to be paid for by the advertiser.
