Campaign Massively Under-Delivering

[maoxiong]

As suggested, I've put this in a new thread. Perhaps I'm missing something on how Rev works, but I don't see what I can be doing wrong.

 

Ok - so I have one zone, in which one remnant campaign is running, with 4 banners underneath it:

 

B1 - Shown to UK Users OR Worldwide Signed in Users

B2 - Shown to Worldwide users on the homepage only

B3 - Shown to Worldwide Guests who don't meet B2 criteria

B4 - Same as B1

 

Running just this - fine, no problems - as expected.

 

However - adding in a contract campaign to the mix:

 

C1 - UK Users Only.

 

And things start to go very wrong.

 

The contract campaign currently is under  delivering by about 20%. I did the calculation that it should be sending about 21,000 impressions a day, but it's not even hitting 17,000. B1 or B4 are shown instead, when they shouldn't be getting any action.

 

On some days, the total UK impressions available for that zone might only be 19,000 - in which case the under delivery drops too.

 

I think I remember reading somewhere that contracts didn't work properly if the total available impressions were close to what is needed, but I'm not sure where that was. To get around this for now, i'm using an exclusive contract, but obviously, thats not ideal as on some days, it sends too much and I need to end the campaign a few days early.

 

Probability for C1 banner is always around 66%.

 

Any ideas - this is pretty frustrating!

 

[maoxiong]

Hmm, I've just realised something else too. Even though B1 has a weight of 2 and B4 has a weight of 1 - B4 is being shown about 3 times as much as B1!!

 

So it seems that isn't working right either (though it is fine if the Contract Campaign isn't running)

[Eliza0406]

Hi,

In general, Ad is not shown frequently in contract campaign.

Ad is displayed based on three aspects,

1. Start/End Date
2. Daily Impression count
3. Priority value

Also campaign weight need not to be provided in contract campaign.

[Matteo Beccati]

The main problem here is that UK traffic is (I believe) a small portion of the total available traffic for that zone and the prioritization algorithm lacks the feature to take into account targeting (time/date limitations aside). It surely tries to compensate for the "missed" impressions, but it falls a bit short.

 

This is a known limitation. I have some ideas on how to improve it, but it would be a major undertaking that I can't afford without proper sponsoring.

 

My suggestion would be to add other banners to the contract campaign in order to cover as close as possible 100% of the zone traffic.

 

As for B1 vs B4, that's pretty weird. What does the zone priority screen tell you?

[maoxiong]

You mean just duplicate the current banner a few times under the campaign.

 

As for priority, it's a bit odd (unless I'm not fully understanding how it works) - as if we take yesterday for example:

 

Zone Z1 had 33,132 impressions - of which 21,877 were UK traffic.

 

Throughout the whole day, the probability on the contract campaign was around 67-70% - which would have covered the needed 21,877 impressions, however only 18,809 were served.

 

The only other possibly spanner in the works, if that campaign is also sent to a second zone, which has incredibly low traffic - maximum traffic about 1,200 a day, with some hours being zero (usually early morning) - could this be affecting things?

 

And B1 vs B4 - the probability is shown correctly - 40% for B1 and 20% for B4. although taking yesterdays stats, were 1868 for B1 and 3089 for B4!

[Matteo Beccati]

You mean just duplicate the current banner a few times under the campaign.

 

As for priority, it's a bit odd (unless I'm not fully understanding how it works) - as if we take yesterday for example:

 

Zone Z1 had 33,132 impressions - of which 21,877 were UK traffic.

 

Throughout the whole day, the probability on the contract campaign was around 67-70% - which would have covered the needed 21,877 impressions, however only 18,809 were served.

 

The only other possibly spanner in the works, if that campaign is also sent to a second zone, which has incredibly low traffic - maximum traffic about 1,200 a day, with some hours being zero (usually early morning) - could this be affecting things?

 

And B1 vs B4 - the probability is shown correctly - 40% for B1 and 20% for B4. although taking yesterdays stats, were 1868 for B1 and 3089 for B4!

 

What I meant, is that the same campaign should also have a banner targeted outside the UK, but that's hardly the problem with the numbers you post. I believe the engine should be able to accomodate. The other zone shouldn't be much of a problem with Revive (as oppesed to older OpenX releases), but please try to unlink it and see if things improve.

 

As for B1 vs B4, there must be something else getting in the way (different targetging, delivery capping?)

[maoxiong]

Well that campaign is a UK company who only want UK visitors.

 

The remnants are a mix of Adsense and Tribal Fusion - which is the usual mix, I only throw a contract in when someone specifically wants to take out direct advertising.

 

B1 and B4 - both have exactly the same limitations and no capping.

 

Let me try the unlinking of that zone - and report back in a day, I'll see if it (somehow) fixes B1 and B4 too

[maoxiong]

Ok, that campaign was linked to only that one zone - here at the statistics:

 

Zone total impressions: 31,900

Zone UK impressions: 20,310

 

Campaign shown:17,328

 

So again - way off. It needs to hit 633,000 a month to make target - which even at full count, would miss. By then being 20% under, it's even worse

 

What I don't get though is that the probability is still showing 66% - which would be correct (assuming it's the 66% of 31,900)

 

B1 and B4 are also still delivering wrongly too - (1849 to 2946)

 

For now, I've switched it back to exclusive.

[maoxiong]

Ok, as expected - switching back to Exclusive (no other changes) and the campaign is now taking up all the UK hits.

 

And B1 and B4 have now started to display in the correct 2/1 fashion.

 

Again - no other changes made, except changing the C1 type from Contract to Exclusive.

[Matteo Beccati]

Exclusive probably suits better what you're trying to achive.

 

If set to contract you should make sure you don't set a daily target manually, but book a certain numbers of impressions and set an expiration date instead: this might help compensating underdelivery.

 

As for B1 and B4, tbh I have no idea. I don't see how contract/exclusive could affect the ratio. There must be something else we're missing here.

[maoxiong]

HI Matteo,

 

Well no, Exclusive is no good - as on some months it sends the impressions too fast, so they run out a few days early.

 

On the contract, I'm not setting a daily target - I simply have it set for 1,800,000 impressions, with a 3 month expiration. The only way it could hit that is by using up at least 21,000 a day - but as you've seen, it's no where near it - despite the probability percentage being correct.

 

Hmm, I've just looked at all those stats again, and this could be a red herring - but - if we take the stats I sent you above for example, where the ratio should be 2/1.

 

But each day (when the contract is running) B4 is sending at least 2000 too many impressions, whereas C1 is down by around 2000 impressions.

 

Could this be the logic somewhere?

 

If the campaign was just coming up short, we'd expect all the other banners to raise their impressions equally - however, only B4 does - could this be where the missing impressions are going... (a breakdown by country on those stats show B1 at 392 and B4 at 2394 for UK impressions)

[Matteo Beccati]

breakdown by country?

[maoxiong]

Sorry, I thought I had replied to this, but it didn't go through.

 

Yes, I have the oxCountryStats plugin installed, so I can see the reports down to country level.