Archive for February, 2006

in API

Release 0.12 is out

Monday, February 27th, 2006
By Ilya Martynov
February 27th, 2006

Another bugfix release. Bugs reported by Netblue were fixed (errors in date times returned by the SOAP server, Publiser.update()
not working). More fixes in InsertionOrder and LineItem services in input data validation code.

WSDLs for the production version:

WSDLs for the test version:

in Publishers

What are Some Good Ways To Increase eCPM?

Friday, February 24th, 2006
By lwalsh
February 24th, 2006

Publisher eCPMs can vary significantly, even across similar sites. So when a publisher asks me what we pay on average, I explain that it’s hard to figure out, and that price depends on a number of factors.

On the other hand, maximizing eCPM is not that difficult if you know how to manage those factors effectively. Below are some suggestions:

  • Ad placement - a poorly placed ad size will not buy that house in Maui. “Below the fold” ads are virtually invisible. The ideal placement is directly above the content. After that, the strongest positions are below (not below the fold) and to the left and upper left of content.
  • Ad noise - How many ads are on the page? 3? 6? 10,000? When there are large numbers of ads on a page, the click-through rate and conversion rate of each ad unit is lower, which generally lowers the eCPM of each individual ad placement. Remove the poorest performing ad units and then focus on how to optimize the rest.
  • No site sections - Is every site getting untargeted run of network ads? There is usually no way of knowing, but there are ways to avoid this.  Segmenting your site(s) into sections based on content can help ad servers like Yield Manager optimize for those areas and also allow advertisers interested in targeted categories find your inventory.
  • High impression frequency - Does each unique user see the same page 50x a day? Or 1x a month? If a site user has browsed past 15 different ads, and hasn’t clicked through any of them, the next ad impression they see is unlikely to present much value to an advertiser. This drives down the predicted eCPM value of that impression. Increase the number of users coming to your site, not just the number of pages each user views.
  • Clicks and conversions - Which placement is underperforming? Analyze click and conversion rates and then modify placements until you get it right.
  • Competition - allowing advertisers/networks to compete for your inventory is a sure way to command a higher price, and it’s where Right Media distinguishes itself. Putting each impression up for auction, as opposed to allocating bulks of inventory to specific networks or advertisers, ensures that if a buyer is willing to pay more for impression, that buyer has access and can bid.  If, on the other hand, you send a bulk of impressions to one network for an extended period of time, it doesn’t matter that a different advertiser is willing to pay more—it doesn’t have access to that inventory, and you leave money on the table as a result.

In a future article I’ll talk more about competition and the systems Right Media is building to help publishers.

in Direct Media Exchange

We May Have a Name

Wednesday, February 22nd, 2006
By Pat McCarthy
February 22nd, 2006

It’s been a long climb trying to find a name, but we think we may have finally made it to the top and found one.  We’re doing some trademark searches, but hopefully we can post it here shortly.

in API

Release 0.11 is out (the first production release)

Tuesday, February 21st, 2006
By Ilya Martynov
February 21st, 2006

This version doesn’t contain much visible changes since 0.10. Mostly it is bugfixes in InsertionOrder and LineItem services in input data validation code.

Major news is that it is the first version to go into the production. Since this version we’ll be running two versions of API - the production version which uses the production database and the test version which uses a copy of the production database. Please note that some APIs available in the test version are not available in the production because they haven’t been tested sufficiently yet. At the moment Creative and Campaign services are not available in the production version. Also please note that since the production version is an early beta unlike the test version it is not open for public access. Please ask on rmapi-users@rt.iponweb.net for access rights for the production version.

WSDLs for the production version:

WSDLs for the test version:

in Media Guard

Automated Creative Tester

Friday, February 17th, 2006
By Michiel Nolet
February 17th, 2006

Creative Tester is the automated testing component of Media Guard. The purpose of Creative Tester is two-fold: first, to determine whether a creative is auditable (see the prior post on auditableness); and second, to flag and detect any changes to the initial “behavior” of the creative that was observed. If the latter happens, and the creatives content changes in any way, the creative will immediately be removed from “Media Guard Classified” status until a human can re-review the creative.

After Creative Tester tests a creative and assigns the Technical Attribute tags to the creative, the creative will either go to a human review to assign the Content & Offer Attributes, or the creative will be marked as “unauditable” and will not be reviewed by a human. If a creative does not go through a human audit it will be categorized as non-Media Guard Classified content. A creative will be marked as “unauditable” by creative tester for several reasons, such as the if the creative runs Active X or has any viruses. Additionally, creatives that are 3rd party tags are given special consideration.

Determining auditableness of creatives that are 3rd party tags:

  • Is it a 3rd party tag?
  • If it is a 3rd party tag, does it always load the same creative?
  • What are the creatives being loaded?
  • Does the creative always go to the same landing page?
  • What are the landing pages the creative sends the user to?

The above tests are used determine the difference between a rotating 3rd party tag, which is not auditable, and a legitimate advertiser 3rd party tag. Creative tester will analyze the 3rd party tag behavior and record anything and everything it finds. If we determine that the creative consistently loads the same offer (say, 4 creatives for NetFlix) that go to the same site (netflix.com) we will pass the creative to our human audit phase. Of course, if we ever determine a change in the creatives downloaded or click-through URLs we will immediately flag this creative for re-review.

Other ‘Auditableness’ Characteristics:

Since we are continually loading creatives to make sure they are auditable we decided to also include some tests for malicious behavior that sometimes sneaks through 3rd party tags. Specifically we will be testing for:

  • Strange popup behavior
    • Does the creative have an exit pop?
    • Is it a banner that is trying to launch a pop?
    • Does it load multiple pops or other JS warning windows?
  • Does the creative try to initiate an Active-X install?
  • Do any of the files downloaded test positive for viruses?
  • Are there any fatal JavaScript errors?

We are still determing exactly how we will deal with some types of “behavior” found in Creative Tester and which creatives will not be submitted to a human review. Our primary objective is to keep out very malicious content or content that can change significantly and thus is impossible to classify.