OpenX Tips

In OpenX, it is possible to apply capping to banners. Capping allows you to set a maximum number of times a banner will be seen by a user. This limitation can apply:

• For the browser session — that is, the limitation on the number of times a user will see the banner will apply until the user closes their web browser, at which time the capping limitation count will be re-set; or
• For all time — that is, once the user has seen the banner the number of times specified by the limitation cap, they will not see the banner again, even if they close and restart their browser.

With either type of capping, you can specify an optional time period. After this time period, the cap will be re-set, and the user will once again be able to see the banner, until the cap is reached again.

In addition to setting capping at the banner level, capping can also be applied at the campaign level. In this case, the capping is applied to the aggregate impressions for all banners in the campaign, and once the limit is reached, no banners from the campaign will be shown.

As an example, consider the banner shown below. It has been set up with a capping level of 2 impressions for all time, with a capping re-set period of 24 hours. You have already seen the banner once now, so, if you reload this page, you will see the banner for a second time — which is the limit. Should you re-load the page a second time, you will not see the banner, as you have already reached the limit of two impressions. (But if you come back tomorrow, the banner will once again show for another two impressions.)

The delivery capping options set for the above example banner.

With regards to the above example banner, no guaranteed delivery banner has been included in the zone being displayed, which is why no banner appears once you have reached the 2 impression limit per 24 hours. This highlights why, in a real-world OpenX setup, you should always have a guaranteed delivery banner in every zone.

Why use capping?

Capping banners is not for everyone. Indeed, it makes little sense to apply capping to banners that are there to simply take up unsold inventory.

However, it is widely believed that as the number of times a user has seen a banner increases, the likelihood of that user clicking on the banner decreases¹. As a result, if you are running CPC or CPA campaigns, you may find that you will make better use of your inventory by applying capping, so that you don’t show the banners to these users too many times — and once the capping limits have been reached for a given user, then you can show the user the lower earning CPM campaigns.

Cookies & first time users

One important thing to note about banners with delivery caps applied is that they are managed using cookies. If OpenX cannot set cookies, then banners with delivery capping applied will never be shown. This is because, if capped banners were shown to users who had cookies disabled, then these users would be able to see capped campaigns infinitely. For cases where you have actually sold inventory on a capped basis, this would obviously be very bad — which is why OpenX does not do it.

However, this does mean that on the very first visit a user ever makes to your site, OpenX will not yet know if it can set a cookie or not. As a result, on this very first ever impression, OpenX will not deliver a capped banner. It will wait until the second impression to do so, when it can then be sure if it was able to set a cookie from the first impression. This may be important, if you absolutely require that a certain banner be shown to all users on their first visit to your site — in this case, you will need to use an uncapped banner.

1. It is surprisingly hard to find any actual evidence to back up this belief, though. Please comment below if you have a reference to any! [↩]

The OpenX ad server‘s Maintenance Prioritization Engine (MPE) calculates priorities for banners in Contract campaigns. (OpenX’s Contract (Exclusive) and Remnant campaign types have their banners delivered based on relative campaign weights and banner weights.)

However, if you read the fine print about how the MPE works, you may notice that every zone in OpenX has a default zone impression forecast — that is, OpenX assumes that, until it is able to learn what kind of inventory is available in a zone, that it has this number of impressions per hour.

This means that if you create a zone in OpenX, but don’t actually generate its zone tag and put it on your web site (which in turn means that OpenX will never see any requests for banners in the zone), then the OpenX ad server will continue to assume that this zone has the default number of impressions/hour.

While this might not seem like a big deal, consider this: the default is 1,000 impressions/hour. If you link a campaign to just one such “inactive” zone, then the OpenX ad server will think that there are 24,000 more impressions available per day than there really are. That’s a lot of impressions — and the more inactive zones you link your campaigns to, the more of these incorrectly assumed impressions will exist!

This will cause your Contract campaigns to under-deliver.

As a result, you should never link your campaigns or banners to zones that are inactive.

As you may have read in the OpenX 2.8.1 release notes:

Zone probabilities are distributed amongst campaigns, independent of the number of banners. OX-3095

This means that it is now possible to divide a zone’s inventory up between the linked campaigns by campaign weight, without having to re-calculate all of the weightings when a new banner is added.

As a result, Tip #8: Understand the banner delivery process has been updated to describe the different delivery process for OpenX releases before and after version 2.8.1.

Here’s a great question from reader Tabrez:

Let’s say we have a newspaper website, and we want to have a hierarchy of zones. Should we have (for example) a Sports News zone that then daisy-chains to an All News zone? Or should we use source params etc. What’s better, when?

Similar questions appear on the OpenX forum from time to time as well.

There are actually two different — although related — questions for the price of one here. The first question is, what is the best way to deal with site segments on my website? The second is, how do I deal with hierarchies on my website?

Let’s look at each question in turn.

Site segmentation

Some websites have multiple segments. For example, in Tabrez’s question above, his newspaper centric website will obviously have many different news segments — he’s highlighted that “Sports News” and “All News” as just two of these segments.

If your website has multiple segments like this, then selling advertising based on the different segments is probably a good idea, as it should allow you to provide your advertisers with a more specific target audience, and thus command better rates.

Of course, this means that you need to be able to manage your advertiser’s banners so that they show up in the appropriate segment(s) of your website. So, what’s the best way to manage this with the OpenX ad server?

As Tabrez suggests, there are essentially two different approaches — you could create a separate zone for each segment of your site (i.e. use multiple zones), or you could have just the one zone that is used across the whole site, but use delivery limitations (i.e. banner targeting) to ensure the banners are delivered in the correct segment(s) of your site.

But which of these two approaches is the “best” approach?

The answer, as so often is the case with a very powerful and flexible system like the OpenX ad server, is that there are a number of pros and cons for each of the two approaches, so which is “best” will depend on what works best for you! Here are The Guru’s thoughts on the two different approaches…

Multiple zone approach

The pros of using a different zone for each segment of your site are:

• As each segment in your site is a separate zone in OpenX, and because it is possible to view OpenX statistics by zone, this means that it is possible to view all of the OpenX delivery data in the statistics screen by site segment. This includes the ability to see how each banner is performing on a per-zone (i.e. per-site segment) basis. This information is not available when using the banner targeting approach to site segmentation.
• Provided that your zones are clearly named, ideally with some form of naming convention, then once you have configured all of the different zones, management of your advertiser’s banners should be very simple, as “targeting” banners to given segments of your site is simply a matter of linking the campaigns and/or banners to the appropriate zones. As the linking of campaigns/banners zones is something that you would do anyway in OpenX, this means that the multiple zone approach to site segmentation introduces no additional management overhead.

The cons of using a different zone for each segment of your site are:

• Although the multiple zone approach to site segmentation means there is no additional management overhead once you are up and running, there is potentially a high initial set up cost associated with the approach, in the event you have many different segments on your site. This is because you need to set up all of the separate zones in OpenX, generate the zone tags for each zone, and then ensure that these tags are placed correctly in the different segments of your site.
• If you have advertisers that run campaigns where the campaign banners are not always displayed in the same website segments, then this approach will mean that you cannot link campaigns directly to zones — you will need to manage the zone links at the banner level. This may introduce additional work, especially if there are many banners per campaign.
• If you do not use an appropriate naming convention for your zones, or you have many hundreds or thousands of site segments, then the OpenX screens for managing campaign/banner to zone linking may be slow, and it may also be difficult to find the appropriate zone(s) to link to, simply as a result of the number of zones in the system.
• If you have segments on your website that receive very small numbers of page views, then you may find that these zones do not effectively deliver Contract campaigns. This is because the OpenX Maintenance Prioritisation Engine depends on zones having a reasonable number of impressions per hour to be able to accurately prioritise banners — if there are very few impressions per hour in some of your segments, using banner targeting to segment your site may be preferable.

Banner targeting approach

The pros of using banner targeting to target banners to different segments of your site are:

• There is almost no initial setup required to start targeting banners to a segment of your website. Simply enter the required banner delivery limitations that will ensure that your banner appears only in the required site segment, and you are done¹!

The cons of using banner targeting to target banners to different segments of your site are:

• As the site segmentation is being performed with delivery limitations, and because delivery limitations can only be applied at the banner level in OpenX, this means that every single banner that is to be targeted to a site segment needs to have delivery limitations applied, which introduces an ongoing additional management overhead².
• Unlike the multiple zone approach, it will not be possible to view the OpenX delivery data for banners on a per-site segment basis. As there is just one zone across the whole site, banner delivery data will simply be logged against this zone, no matter which site segment a banner’s impression occurs in.
• If you have segments on your website that receive very small numbers of page views, then you may find that banners in Contract campaigns that are targeted to these segments can cause the OpenX Maintenance Prioritisation Engine to cause other Contract campaigns in the system to delivery incorrectly, due to issues with dealing with very highly targeted Contract campaign banners.
• Debugging delivery issues where a banner is being displayed in an incorrect site segment may be more difficult than with a multiple zone approach to site segmentation, due to the potential complexity of the banner delivery limitations.

Conclusion

Performing site segmentation is a balancing act. The pros and cons of the two different possible approaches probably seem difficult to weigh up, especially if you have not previously tried both approaches.

To make life simpler, The Guru recommends that if you have not performed site segmentation before, you select an approach based on how many site segments you have. That is, if you have only a dozen or so site segments, then use the multiple zone approach, as this will not be too difficult to setup, and with a small number of site segments, you are not likely to suffer from the many of the cons of this approach. However, if you have more site segments that this, try the banner targeting approach.

Keep in mind that if this is your first time using site segmentation, no matter which approach you first try, you can always experiment with the alternate approach later, should you find that the initial approach is not suitable for your needs for one reason or another.

Finally, it is worth pointing out that you can actually combine the two approaches. That is, you can always partially partition your site via multiple zones, and then further partition each zone into smaller segments via banner delivery limitations. This approach will of course require balancing at which level you stop multiple zone based segmentation and move to banner targeting based segmentation, but with time and practice, you will learn to get most of the pros, while avoiding most of the cons.

Zone hierarchies

The concept of a zone hierarchy is related to the idea of site segmentation. Think about it this way — if you have divided up your website into segments for the purpose of selling your inventory at a higher rate, what do you do when when you have spare inventory in a site segment?

The answer, for once, is reasonably simple. If you have used multiple zones to manage your website’s segments, then using zone chaining to create a hierarchy of zones is the correct approach. That is, if you have a zone that represents a site segment (e.g. the “Sports News” segment zone from Tarbrez’s question), and you find that that the site segment/zone has spare inventory, then the easiest way to ensure that a banner is shown is to link the zone via zone chaining to the next zone “down” in the zone hierarchy. In Tabrez’s example, this sounds like it would be the “All News” segment zone. Of course, you can have as many levels of zone changing as you need in your zone hierarchy³.

If, however, you have used banner targeting to manage your website’s segments, then a zone hierarchy doesn’t really apply. Presumably, advertisers that are buying inventory in your site segments will be doing so via the Contract (Exclusive) or Contract campaign types, and so you can simply link any Remnant campaigns, containing un-targeted banners, to the single, site-wide zone in OpenX. These Remnant campaign banners will be shown whenever there is spare inventory that is not used by your segment-specific advertiser campaigns, due to the way that the banner delivery process will only show banners from Remnant campaigns after it has determined that there are no banners to be displayed from any Contract (Exclusive) and Contract campaigns⁴.

1. Of course, some delivery limitation types may require additional work before they can be used. In the original question, it was suggested that “source parameters” would be used. To be able to target with source parameters, the Site – Source delivery limitation type requires that changes to the zone tags will need to be made so that the appropriate source parameters are passed into the OpenX ad server calls and allow the targeting to be performed. For this reason, the Site – Page URL delivery limitation type may be more suitable to perform site segmentation, assuming that you are able to accurately determine site segments via your website’s URL. [↩]
2. This can be somewhat reduced by appropriate use of targeting channels, however. [↩]
3. There is a performance penalty for each level, of course, so try not to have a zone hierarchy that is too deep. [↩]
4. If you are not using the Contract (Exclusive) or Contract campaign types for your segment targeted advertiser campaigns, then a two zone hierarchy would be appropriate — an “initial zone” that is linked to all of the campaigns containing banners that are site segment targeted, which is then linked via zone chaining to a “remnant zone”, containing all of the untargeted banners. [↩]

Check out the free replacement icon set for OpenX 2.8 from EsDeveloped.

The Guru has, and it’s rather good!