Priority Engine Always Does The Opposite Of What Is Required. Why?, More of a open letter to OpenX than a question. Options

[erictbear]

This feedback has taken time to compose. I trust that the developers of OpenX receive it in the spirit of my contributing to their fantastic software which, imho, has gone off the rails somewhere with the new priority engine.

As mentioned in previous posts, I have been using OpenX in its various incarnations for over five years - I upgraded to 2.4 from 2.0 three months ago and have been struggling with priority ever since.

I can not get anything to deliver as expected.

Situation A.
So far, every campaign that has only required low levels of hourly delivery has overdelivered by 200% to 300% and finished days or weeks early.

Situation B
So far, every campaign that has required high levels of hourly delivery has underdelivered by between 40% to 70%. This has occurred even though there has been buckets of inventory available and the competing but lower priority campaigns have overdelivered by double or triple. It has even occoured when there has been no competing campaigns and instead of the high priority campaign (HPC) delivering to it's full target - a chained zone has been pulled in and eaten all the inventory!!! In theory, a chained zone should not pull through until there are no directly linked banners left to deliver.

I am getting very frustrated and I note others are having similar experiences.

-----------

Now - the consistent advise on zones and the priority engine is as follows;

1) You should have a single low priority, uncapped/targeted banner so that all impressions are logged and the priority engine knows how many imps/hour are truly available

Each time I do this, the low priority, uncapped banner constantly overdelivers and prevents my HPC's from delivering any more than 40% or 50% of targets.

Whilst I understand that the priority engine needs to know and calculate what inventory is truly available, the above advise begs the following questions;

1) What is the point of zone chaining if there is always one unlimited/uncapped linked banner to a zone (e.g an adsense banner)? In theory there will never be a situation where; "If no banners from this zone can be delivered, try to... Display the selected zone instead"

2) Why does the priority engine underdeliver LINKED HPC campaigns and instead pull through a chained zone when in theory it NEEDS to know what is truly available? What's the difference between linked HPC's and linked low priority uncapped banners? Why on earth does it stop delivering linked HPC's when they are not even close to their targets when enough inventory is available for it to keep delivering HPC's and in the process continue calculating true zone inventory?

-------------

Whilst I have about five years experience using this ad server, I did not use Max Media Manager during the period that the software had forked. Perhaps I am doing something fundamental in my implementation of zones (evolved over the years of using phpAdsNew) that does not map across to OpenX 2.4. There is nothing in the documentation that would suggest my implementation is not ideal - but I put it out here for discussion.

-- My website delivers approximately 10 million page impressions per month. Most pages have two ad placements, some only one and some three, but on average I deliver 20 million banners per month.

-- I am not running a network - so I have created only one Publisher in OpenX.

-- I have eight different content sections that I need to separately target. Firstly because that is how the advertiser wants to buy and secondly so that I can calculate the banner revenue generated by the content v's the cost of producing that content. (To that extent I absolutely love all the eCPM revenue, media cost etc. details that 2.4 generates)

-- I run four different banner sizes - but not every content page runs every banner size, most only run two banner sizes.

-- I have created a zone for each banner size in each content section. Therefore I have approximately 20 different zones including four zones that do not have any invocation code on the site and are only used to deliver adsense banners as a chain linked zone (i.e with the expectation that no banners will deliver from these zones until AFTER every paid campaign has totally delivered it's target - this is what used to happen perfectly with 2.0).

-- All invocation code uses the cachbusting unique variable (a random number to 8 digits) and all invocation code has a Source tag. All invocation is done using the iFrame method.

-- I have not made much use of channels (yet), but as far as I can tell, channels are just preset delivery limitations making it easy to quickly set up complex delivery limitations that are frequently used. (Is this understanding correct or flawed? Please advise.)

That is how my implementation of zones is set - Is my implementation flawed or within the realms of what is considered normal? Please advise.

--------------

Lastly, documentation for 2.4 still has not been fleshed out adequately. Whilst I appreciate that documentation requires a lot of effort, much of it already existed years ago with phpAdsNew - there should be no reason now (six months after OpenX geting significant investor sponsorship) for there to be so many gaps between what's on a screen and the corresponding documentation. A basic reconcilliation of software screens v's documentation is all that is required.

As an example:
Zones > Advance Zone Properties

Where is the information about Zone Forecasting settings? What is the purpose of this setting and what is the consequence of ticking the Channel option? Someone has gone to a lot of trouble to code this feature - 20 minutes is all it would take to document what it's all about.

--------------

Currently the problems with the priority engine is costing me, I'm not delivering full priced HPC's and I'm dissapointing advertisers who's campaigns complete too quickly.

After three months of battling with the priority engine (and pending finding out exactly what Zone Forecasting is all about) I'm going to attempt one last reoganisation of my adserving. I plan to reduce all my zones to just the absolute bare essential and use channels as the main method of targetting to see if that makes a difference.

--------------

I have been offered a beta account for Google's AdManager - so far it remains unused.

I have a lot of time and energy invested in OpenX after five years of use. I get the fact that it's free open source software and I'm not a paying customer - but OpenX also has real investors now and they will require lots of users for any successful exit and/or ROI play.

If I can not resolve what the hell is doing on with the priority engine I'll be left with no alternative other than to put my energy into implementing Google's product throughout my CMS - no small task, but I need to do something.

I've posted this rant because I believe OpenX needs to know where publishers like msyself are at - the last three months trolling these forums leads me to believe that beneath the surface there are there more disgruntled users on this issue than first meets the eye - especially amoungst those of us who have been long time users of 2.0 and previous.

--------------

[end rant /]

[Arlen Coupland]

Hi,

I completely appreciate your situation and I don't feel at all that it was a rant, but simply some frustration and largely some good description of your scenario which does help us help you

Here is a suggestion which you might not like: try an upgrade
I usually avoid this suggestion, but 2.6 has some changes in the priority engine which sound like they might help you.

The main question I have on your setup -- what targeting are you using for the high priority campaigns? It helps to analyze the target wanting to be reached vs the # of imps available for that audience (geo, source, etc) vs the # of imps for the general audience
It may be that you do want to use zone chaining to get the best results. Its hard to give a suggestion that works for everyone, but the 'use a low priority' banner is a general suggestion which works in most cases.
Using zone chaining works well for something like:

ZoneA has 1 HPC banner targeted at canadians. The zone is on homepage. 50k imps er day, 10k are canadian. HPC wants only 4k imps a day.
Then set zone chaining to load a 2nd zone with low priority banner if no banner is available in ZoneA.
This way, ZoneA is balancing the HPC against 4k available imps a day, rather than 50k imps. So it can more easily adjust limitations on the specific audience, rather than the general traffic.

Hope that helps

[erictbear]

Ok,

Thanks for your reply Arlen.

I have just upgraded to 2.6 - The new interface is very smart, very web 2.0!

It's nearly as big an upgarde as 2.0 to 2.4 was.

I will let you know how the priority trafficking goes - I'll have useful numbers in about 72 hours to compare.

[afrigator]

Hi Arlen - I too have been trolling the forum for days and it looks like this scenario is still existent even in 2.6.3 which we're running. You can check my posts to the forum but my problem is identical to erictbear. In a last attempt to try and resolve this I am now trying to setup zone chaining.

I currently have 1000 zones and now due to this fundamental flaw I have to create an additional 1000 zones so that I can create a zone chain for each website that's in my openx.

Surely there is a better way but OpenX refuse to confirm or deny that a bug exists in the MPE and to be honest I think there are a few bugs.

I guess this is turning into a rant (more so than erictbear) but I've got a lot riding on a successful implementation and OpenX is failing to deliver on a critical factor...

[nedroy]

Hi There - we are also experiencing the same problem with HPC settings. We were very surprised to not see more posts about this issue (glad we found this one). This appears to be a bug and a major one at that. We run a small advertising network with six sites and zone-chaining isn't appropriate because each zone is set by a specific ad-size and web site (each site has a very different audience). We have always had issues with some HPCs not starting at their due dates and since we upgraded to 2.6.3 we have experienced exactly the same issues as described above (on top of the existing bug). We're now having to consider a paid ad-server after spending considerable time and effort getting OpenX hosted and working.

This is a serious issue that needs addressing ASAP.

[afrigator]

Hi There - we are also experiencing the same problem with HPC settings. We were very surprised to not see more posts about this issue (glad we found this one). This appears to be a bug and a major one at that. We run a small advertising network with six sites and zone-chaining isn't appropriate because each zone is set by a specific ad-size and web site (each site has a very different audience). We have always had issues with some HPCs not starting at their due dates and since we upgraded to 2.6.3 we have experienced exactly the same issues as described above (on top of the existing bug). We're now having to consider a paid ad-server after spending considerable time and effort getting OpenX hosted and working.

This is a serious issue that needs addressing ASAP.

Yip this is a bug. The work-around for us was to modify the core source code for OpenX so that we could bypass the MPE completely. Basically we have now added the ability to limit banner views to XX amount per hour on a low-priority campaign.

The only way to bypass the MPE is to use a low-priority campaign as this doesn't rely on probability checking. So all our paid campaigns are now running on low priority with our modified X impressions per hour hack and everything is working perfectly for us.

It is less than ideal and really frustrating - I just hope openx actually resolve this problem in the near future before we have to move away from it.

[gracie]

Hi, we just migrated over our entire network (100 million impressions a month) to OpenX 2.6.3. After launching the campaigns, we have exactly the same issue. and this is MAJOR!

Can OpenX at least admit it's a problem, so we know someone's working on it? I haven't found any openX comment about this is being an issue. The scary thing is that they think it's working fine and no one's fixing this. When can we expect MPE to function?

At the moment I need to find an alternative to serve up our HPC. Seems setting it to low priority is the only solution. However, it's not letting me to cap the campaign.

Yip this is a bug. The work-around for us was to modify the core source code for OpenX so that we could bypass the MPE completely. Basically we have now added the ability to limit banner views to XX amount per hour on a low-priority campaign.

Afrigator, can you let me know how your mod works? Can you still cap your HPC when it's setting as low priority with an hourly cap? IE. campaign with a booking of 2000 clicks ending at the end of the month. Is it possible to set ending date with an hourly or daily cap to manage the deliveries?

Thanks.

[afrigator]

Afrigator, can you let me know how your mod works? Can you still cap your HPC when it's setting as low priority with an hourly cap? IE. campaign with a booking of 2000 clicks ending at the end of the month. Is it possible to set ending date with an hourly or daily cap to manage the deliveries?

No you can't use date-based capping. We have to manually check how many impressions to serve per hour and then continuously check each day if it's meeting the quote. Once it's reached its limit you need to deactivate it.

This is really not ideal at all but alas, not sure what else to do.

[acyd_burn]

Just want to state that i have the exact same problem too. HPC's are always underdelivered, there is no longer a way to instantly see how your campaigns are progressing (forecasting screen known from phpadsnew)... it is very very frustrating - i consider switching back to the good old phpAdsNew code - there everything worked absolutely fine.

[Xav]

I'm looking for the old phpAdsNew code, do you know where could I get it? I have recently "upgraded" to OpenX and I'm completely frustrated.

[Mentor]

Hi,

We are going to release an OpenX 2.8.1 soon. This version contains of priority engine fixes which would work similarly to the old PhpAdsNew.

[Xav]

I have been reading issues like these posted a year ago; in this scenario, "soon" and "similarly to the old PhpAdsNew" seems too undefined for my needs. This topic was opened on july 2008!

I'm receiving complaints from several advertisers; they are noticing the situation. So, sadly, after 7 years with phpAdsNew (and now OpenX), I'm migrating to Google Ad Manager; good luck for all of you anyway.

[dacumen]

I'm looking for the old phpAdsNew code, do you know where could I get it? I have recently "upgraded" to OpenX and I'm completely frustrated.

http://download.openads.org/Openads-2.0.11-pr1.zip

I've just re-installed version 2.0.11 of OpenX on my server. It's the most recent version of 2.0.x I could find. I'm not sure yet how I'm going to configure the two implementations, but I'm thinking a swap-zone with 2.8 handling overall serving and 2.0 handling my high-priority, paying ads.

I've retained a consultant to review my implementation and code. He's a a former core developer on OpenAds, so I'm hopeful that if I've done something wrong between my religious upgrades, server moves, and idiotic idea to change my maintenance interval, he'll find the issue.

I'm eagerly awaiting the release of 2.8.1 so that the underlying core of OpenX returns to intelligent functionality. I'm confident that we'll be able to get OpenX working on my server by the end of May.

...if not, then maybe I'll join you at Google.

[dacumen]

After the upgrade, this issue still remains. I have a campaign scheduled to deliver 10,000 impressions by Sunday that's puking out 200 impressions per day, and another that's supposed to be capped at 1,000 impressions per day that delivered 3,000 yesterday. The MPE is completely wacked!

I'd be more upset if my consultant wasn't working on a pluggin to fix the MPE, since you guys obviously can't.

[Jonny Bergström]

Sorry to be bumping this heated thread, but I'm seeing similar results. The HPC's don't deliver at all, or very very minimal amount of impressions. I'm using OpenX 2.6.1 however, and it's kind of discouraging to read the above comment that the MPE still doesn't (seem to) work in 2.8.1.

Which gets me thinking - in forums like these I guess 95% of the people who post, post about problems they have with a certain feature. But when do you hear the success stories? Never?

Can someone stand up and say something like "Yes, I served successfully 15000 ads for a HPC during a month's campaign in a zone delivering 100000 imps during the month. The imps of the HPC was around 500 every day!"?

What would be even better is a minute tutorial - complete with screenshots - on how to set up a HPC and successfully have OpenX serve a set amount of imps over a specified time - and that's guaranteed to work.