Using DFP to pit ad networks against each other

[jpenns]

I've seen people talking about setting up DFP with multiple ad networks like Lijit and having DFP basically auction your ad space off to the highest bidder. This sounds great to me. The problem is I can't figure out how to do it. I'm not familiar with DFP at all, but even as I've spent hours familiarizing myself with it, I haven't been able to figure out this specific task. And there's not a lot of info out there about how to do it.

I saw an older thread on this forum where people were discussing it, which is what drew me here. Does anybody here do this, and would you be able to post a guide or something?

Thanks so much in advance!

[diberry]

Re: targeting. Some of the ad networks do their own geo-targeting, which is why I don't feel like I even know where to start on that one. Do you target, say, certain networks for US and certain networks for "international"?

As for device, what do you do there? I use a responsive theme with the same ads showing on desktop and mobile, so I'm not sure there's much point in my case. Watching how CPMs have plummeted across the board this year makes me think I'm better off focusing on growing my traffic and other income streams than trying to figure out how to wring $0.00075 more out of each impression.

[child please]

Runfun - ok so you have 4 AdSense line items - I don't understand why you have the 970x250 and 300x250 AdSense line items. Do you also have competition with AdSense enabled under "Inventory"? For the above ad units, do you have f-caps or any targeting in place?

I would be enabling competition with AdSense in the Inventory/ad unit settings and getting rid of those line items altogether.

[child please]

diberry - yes, some networks do geo target, but why would you want to send them those impressions in the first place. Look at this example below, in two scenarios and you might see the difference.

Network A - targets only US visitors

1. Impression from Canada, DFP sends it to Network A (because you have no geo targetting in place). Network A can't fill that impression and either defaults to a pass back code or worse, serves up a blank/default/PSA ad. If it goes to a passback, say it's your AdSense code, AdSense fills it with an ad, say $0.05 CPM

2. Impression from Canada. DFP doesn't even send it to Network A because you have specified that geo target. It will now send it to Network B (who does monetize CDN impressions) and you have AdSense dynamically competing against that impression. You avoid a pass back and now depending on the bid from AdSense, the competition is won by either Network B or AdSense, at a much higher CPM than in scenario #1.

I've posted this before - the goal is to minimize the # of impressions going to your pass back code - you are losing $ this way and yes, it does add up over the course of a day, week, month x multiple ad units. :)

Your second question, if you're using strictly AdSense then yeah, no real point. But if you are using other networks - some of whom may not be monetizing mobile traffic - my scenarios above would apply in that case too. Why send them mobile traffic if they are going to ignore it?

[child please]

Finally, not sure how to take your last sentence about wringing every last fraction of a cent from an ad. All I'll say is it works for me and while I don't want to get into specifics or sound like I am bragging, it is the best decision I have made. Not only has my AdSense income stayed flat or even gone up a little bit, I have added $6k per month from only 2 other networks. Why? Because AdSense is getting less impressions, but they are forced to pay MORE for those impressions.

Competition is a beautiful thing!

[Runfun]

Well advertisers can target different dimensions so if advertiser A only targets a 336x280 dimension and advertiser B targets a 728x90 dimension it can compete with each other.

I'm just testing it and it sounds reasonable it will work. I still don't get the frequency cap because what if you capped the dimension with the best CPM? Other problem is finding another network like OpenX but here in Europe there aren't much alternatives.

[child please]

Not understanding the first paragraph. I meant why enter AdSense as a line item and have it compete in a static sense when you could enable it for each ad unit and have it compete dynamically. This is proven to yield higher CPMs.

To explain f-caps, here's the best example I can think of -

Network A
Impression 1 - $5.00 CPM
Impression 2 - $5.00 CPM
Impression 3 - $2.00 CPM
Impression 4 - $0.50 CPM
Impression 5 - $0.00 CPM
TOTAL CPM - $2.50 CPM (what they report in the console and what you'd probably put into your DFP)

Network B
Impression 1 - $3.00 CPM
Impression 2 - $2.50 CPM
Impression 3 - $2.50 CPM
Impression 4 - $2.50 CPM
Impression 5 - $1.50 CPM
TOTAL - $2.40 CPM

You enter 2.50 in for Network A and 2.40 into Network B without any f-caps. Guess what happens? As per DFP allocation rules, Network A will get ALL the impressions. Even if you put 2.50 for both and the impressions are split 50/50, why are you giving Impressions 3, 4 and 5 to Network A when Network B is going to pay you higher for those!

Best setup: F-Cap Network A at 2/24, then it rolls over to Network B for impressions 3, 4, 5. You are now allocating your inventory to gain the best possible CPMs.

I'm not here to convince anyone, but these are the facts.

[diberry]

I think I've been unclear to the point of putting us at cross-purposes.

Currently, NONE of my traditional banner ad networks do their own targeting, so far as I know. Years ago, I worked with a network that demanded I use DFP and only target US impressions for them, but I couldn't figure out DFP, they didn't help me, and we parted ways. The networks I have *now* tell me there are no countries or devices they won't fill, if they have inventory, which they do sometimes, not other times. So, I wouldn't know where to start on targeting.

I'm guessing either you know how to tell from your end what countries are getting impressions via which networks (and I definitely don't), or you're working with bigger, better networks - like the one that asked me to geo-target - than I'm able to get with at the moment, and those networks tell you what they want so you can provide it.

That's why I'm thinking I need to focus on traffic and then get with better networks, and maybe then they can make useful suggestions about targeting.

I do 100% get what you're saying about avoiding passbacks. Using frequency caps, I have finally got my top network to a 50% fill rate (their reporting). Other networks, I can't get beyond about 30%. I don't doubt targeting would help, but the problem is that I don't know what impressions they're going to fill from one week to the next. I definitely see the occasional blank space where an ad should be, especially on mobile, so I know the networks are being more selective than the account reps tell me. I just don't know how to figure it out for myself or get them to tell me anything more useful.

Does that make more sense out of what I was saying before?

[child please]

Yeah it does. Ask them for a breakdown by geo. Even if they do fill every impression from every country (doubtful), the cpms will vary wildly but what you might see is one overall site wide cpm. It could be that US impressions are performing well but some other geos are pulling it down. Saying all this without knowing your sites geo breakdown. If that is the case, it makes sense to enter in multiple line items of the same ad size...one at a lower cpm for euro and one for a higher cpm for US.

[Runfun]

Now I switched to only OpenX competing with Adsense so no more line items filled with Adsense dimensions.

Before I made the switch I recognised just a few impressions for OpenX but it had a higher CPM... weird. But surprisingly Adsense has a higher CPM than without DFP. Next step is finding another network like OpenX.

[Runfun]

It seems to work well after some minor adjustments. Only thing I'm wondering is where I can see in my Adsense Dashboard what revenue it brings as I get through DFP income?

[diberry]

Runfun, I haven't found a way to separate that out.

[diberry]

Well, I've come across something interesting.

I have been running several sites with RhythmOne ads for years (used to be BurstMedia). After I switched one of those sites to DFP, the CPMs fell to 20 cents and below. My other sites, which don't have DFP, are still getting the same higher CPMs (50 cents at the least to well over $1) they'd been getting for a couple of years - in fact, they've improved some this year.

I've written to my contact at RhythmOne for her thoughts. It could just be something wrong with my implementation of DFP. On the other hand, two ad networks have recently told me their ads work better if they are hard coded to the pages, which is new to me. It's always been the case that putting iframes around ads could go badly, and maybe that's all they're talking about. I'm not sure what DFP puts on them - there are no iframe tags in my source code, but it sure works like an iframe, allowing stuff to load around the ads.

Another ad network told me I'd get higher CPMs if I let the ads load at the same time as the content. Which is precisely what DFP does NOT do, because it's always been a best practice to have the content load first so people don't bounce out while waiting for an ad call to stop hanging up the pageload. But apparently advertisers have other ideas.

Of course, my various ad networks ALL have different, conflicting ideas about what's best, so it's hard to figure out what's working now, let alone what will work 6 months from now. Web advertising just isn't mature yet - it's still looking for that established system that will work consistently and for the long term. It took decades for TV to move from in-show endorsements to commercials to infomercials. Many web ad networks test more ideas than that inside a year, and most of those ideas fail for one reason or another - often, because the advertisers can't keep up any better than the publishers can.

I'll let you know what my RhythmOne contact says when I hear back from her.

[Runfun]

Runfun, I haven't found a way to separate that out.

Thanks and I've the solution (maybe) in my report/stats at my Adsense dashboard I see the performance of a single day. At the bottom of all the ad units it's for example totally $30 but the day was a total of $35. I think the difference is what I earned with DFP.

I've another question, is it possible to see reports in DFP of separated line items performing against Adsense? While using one line item it was easy to recognise but when using more line items I only see the total amount.

[diberry]

I have no idea. I don't think DFP has any idea what you're earning from any of the ad networks.

[diberry]

RhythmOne never got back to me, because that's how more and more networks are treating publishers. Just like Google.

So I did some experimenting. I added some plugins to create a new mobile design for phones, because a lot of ads just don't play well with responsive design. This design doesn't create a new page URL, it just shows this design on devices that identify as phones instead of showing my responsive design for other devices. Bottom line: now the ads look good on every device. But more importantly, I'm not using DFP on the mobile site. I just hard coded the ads to test it first.

Results? CPMs for mobile are higher than for desktop now. Huh.

So I hard coded ads from a couple of networks into a couple of ad slots into my desktop pages while leaving the rest to be served via DFP. I specifically left out RhythmOne, so their ads are still only being served by DFP. Results? Desktop CPMs rose on the networks whose ads are hard coded into the site. CPMs with RhythmOne are still awful on the DFP site, and actually rising significantly on my other, hard-coded sites.

That may sound like a clear indication that ad networks other than Adsense don't like being run via DFP tags. But there's other stuff going on: very low or volatile fill rates, Labor Day US Holiday throwing things off, etc.

I've thought of one other thing to test. I set up DFP so I had ads from the same network competing for the slots. For example, in the same sidebar slot, I'll have Amazon's 300x600 tag and their 160x600 tag, each set up to run when DFP gets down to their CPM level, each with a frequency cap. That doesn't seem to be a problem with some ad networks, but others may register it as an issue. I'm going to change that up so each network only has one shot at each ad slot, see if that changes RhythmOne's terrible CPMs.

I'm also going to go back to using DFP on all my desktop slots, see if that affects CPMs.

[Runfun]

Thanks for all the messages here because I was looking for answers for weeks and I finally got DFP running thanks to this thread. Now I've a question about generating the tags. Do I've to generate all the tags over again when I add a new placement? It looks like it because there's another number in the tags when I generate new ones.

[diberry]

Runfun, YES. Isn't the annoying? If they number stayed the same, you could just add in the line item from the header section, and code in the actual new placement, and that'd be it. Instead, you have to replace everything every time.

I did a couple of tests this week.

--I set most ads in DFP to only compete for one slot. On some networks, that makes no difference to CPM, but on others it helps. I guess some networks don't care, but others are monitoring for the possibility of an ad showing up twice on one page and similar viewability concerns.Fair enough, and it has improved my income a little.

--I ran the ads without DFP for almost a full day. I had intended to go 24 hours, but I couldn't take it, because DFP makes a HUGE difference on load time. About 30 points in Pagespeed, in fact. That alone makes me determined to stick with it.

--The network who told me I'd get higher CPMs if I hard coded their ads to the page? That was true. But the higher CPM wasn't so fantastic that it was worth the extra load time.

I'm feeling like I have a rudimentary handle on DFP now. Since none of my ad networks are giving me any helpful tips on how they target, I'm at a loss on taking it any further. But I'll keep reading.

[Runfun]

Do you mean DFP improves the pagespeed?

In the meanwhile I'm making some improvements and I'm serving my first campaign for a company who's advertising at my website. I've only one problem. I'm serving a 728x90 banner for desktop and 300 250 for mobile. I see exactly the same views but a difference in clicks. Some how I believe it's counting the placement for each size at the same amount.

Maybe you already know but sending someone else a report is very easy. Add a contact, click at the order of the company (for example) than run report and save it with a name. A button with 'share' shows up and you can share it with the contact. The contact has to be with a Gmail account. Only problem left is the amount of views that's the same but they where paying me for a week and not an amount of views or clicks.

[diberry]

Yes, I meant that DFP improved my speed in Google's Pagespeed tool by about 30 points. That's one of the great things about DFP - even if you're running lots of ads with lots of scripts, your website loads quickly, and that can help improve bounce rates and (to some extent) rankings.

No, I haven't tried to do much with reports yet. Good info, thanks!

As for that issue with the desktop v. the mobile ad... that's odd. How are you coding it? Do you have a different theme for mobile, or are you using a script to swap out the ads according to device?

And in DFP, did you set them up as one line item, or two separate ones?

[Runfun]

I made two different line items because DFP didn't recognized any smartphone or tablet even if I exclude or include certain devices. So I used a script and they strange thing is that the impressions aren't exactly the same but now almost a week running it's:

Desktop: 175,115
Mobile: 173,306

Normally I've +/- 78% desktop visitors so it doesn't make sense.

[Runfun]

An update from my experiences. I was surprised by the bad score in Pagespeed Insights so now I turned off single request and of course no synchronised loading. Score for desktop jumped from 66 to 79.