WTF is a ‘tagless’ solution, and why are there so many tags? (Part 2)

In case you missed part I of this post, you should check that out before reading further. If you’ve already been down that road, here’s the next installment in our series about tagless solutions.

OK, so what’s the downside for a publisher?

Assuming you’re setting these providers up as price priority line items, there isn’t any competition downside. If you were setting line items up as sponsorship that could limit AdX’s effectiveness which may not be wise depending on your yield strategy.  There are also concerns that adding partners ahead of your invocation code increases page latency and I strongly suggest publishers monitor this closely if they choose to experiment with tagless.

Also, depending on how sensitive you are to protecting your 3rd party data, one could argue that having a variety of ‘tagless’ solutions looking at every impression that flows through your property is problematic. For the SBM sites this isn’t an issue as we sell less than 5% of our inventory directly so a majority of what we offer is already on the open exchange but it’s something to consider and decide for yourself what’s best.

How much of your inventory do these ‘tagless’ partners take?

This will totally depend on your site, your content, the value of your audience and the ‘tagless’ partner(s) you select to work with. That said, if you tend to perform well with your current programmatic partners that would indicate your audience is valued highly in an exchange environment and this should be a solid boost.

Why do you keep calling these solutions ‘tagless’ if they require so many tags?

Fair point and I’m honestly not sure. I’ve heard people suggest ‘heavy iron’ as an alternative title (referencing the amendments to the source code) but that doesn’t really work either. I’m open to better naming suggestions if you have them.

We understand you didn’t personally create the name, but any theories on what it might refer to?

Not personally but my friend (who also happens to be a great developer) Peter Bernheim of Slader.com suggests the following: I think the idea is that maybe they are ‘tag-second’ in that the decision on how and whether to fill an ad slot is more or less decided before the ad tag is invoked on the page. The script that is added to allow for the first look at impressions allows the tagless partner to evaluate the impression but not return creative.  The eventual call to the partner, should their tagless tag win, does return a creative but perhaps does less logic (?) on the partner’s side before returning a creative.

What are some other theoretical benefits of this technology?

Assuming proper setup, this will allow a publisher to set more accurate price floors for top level bidding against AdX. By selling a higher percentage of inventory at ‘first look’ levels before a passback waterfall you’re increasing revenue and avoiding the latency/discrepancy associated with passbacks. As mentioned previously, you’re also passing a direct value into AdX which starts the bidding process at a rate higher than it would have otherwise which should result in incremental revenue.

Have you measured the discrepancy percentage between these technologies and your ad server? 

Yes, thus far partner vs. DFP impression data ranged from 0 – 4% difference. An improvement vs. most 3rd party javascript setups and a vast improvement vs. passback setups.

What in gods name is a KVP and should I be afraid of them?

A KVP (key-value pair) is just a way to pass a specific value into an ad unit you’ve already setup (in DFP) to allow for custom partner targeting. In this case, that custom targeting would be an identified fixed bid from one of your ‘tagless’ partners for an incoming impression. The KVP is how you let DFP know which of your ‘tagless’ partners has signaled yes (and at what CPM) for an incoming impression. The partners you decide to test with will provide the KVP’s you’ll need and should offer plenty of support around building them (or do it for you).

You mentioned changes to the invocation code, that scares me. What does it mean?

Each of these solutions have a few required changes to the head tag as well as the invocation code you use to run DFP. I don’t handle the changes or the deploys of the changes personally but I can say after a few initial bumps the first time we tried it, it’s been very smooth since then. What helped us break through on the first attempt was to ask for an example of how the source code should look on the page, and we worked from there. That was actually much easier to follow than the supplied documentation. This seems obvious but it’s very important to lean on the technical contact provided to you for questions, concerns and QA. We also make sure to test in a staging environment before taking anything live.

Do these solutions work with both sync and async GPT tags?

So far all but one of the technologies we’ve looked at work with both sync and async tags.

What companies offer ‘tagless’ solutions?

I’m positive this isn’t a complete list (feel free to leave your companies name in the comments and I’ll add you) but I’m aware of tagless solutions from: Amazon (A9), Casale, Criteo, YieldBot, Sonobi, OpenX, Rubicon Project, and Pubmatic.

How many tagless partners should you work with? 

I’m honestly not sure what the correct answer to this question is. On the one hand, the more partners you work with the better chance you have of one of those partners placing a high bid for an impression (in theory). The flip side of that is stacking partner after partner could add latency to page load and/or result in lost impressions. This is precisely what I’m attempting to figure out at the moment so I will update this as I have a better sense of what works for us.

9 Comments to WTF is a ‘tagless’ solution, and why are there so many tags? (Part 2)

  1. Hi,
    I was wondering if you might be willing to go through the step by step process of setting this up in DFP using key values and two primary partners plus Adx and Adsense.
    Thanks. Much appreciated.
    C

    • Hi Colin, thanks for reading. Our developer (Jon) is going to examine this in his blog series. He’s posted part I and part II and will be coming out with part III shortly. The tricky thing is each setup is slightly different but he’ll certainly cover the general basics of setup.

  2. Emry, I am enjoying your posts and appreciate you bringing this info into the light. I am curious to know if you arrived at a conclusion on the optimal number of tagless partners? Or how many have you run before you started to see meaningful negative impact?
    Many thanks,
    Mike

    • Hi Mike. Thanks for reading and thanks for the kind words. I think that answer (like everything else in this space) depends. On one hand you could easily argue it doesn’t make sense to work with a variety of exchanges due to DSP conflict (since they are competing for the same impression in a tagless setup). On the other hand, you’d need to A/B/C test to see if that actually creates conflict and I’m guessing scale has a lot to do with it. In general the strategy we’re going with is working with 1-2 re-targeters, and 2-3 exchanges. We’re exploring strict timeout options as to not delay the page load and try to watch individual performance closely of each partner. I’d be open to testing more but I think I’d sub out a poor performer for a new partner for testing. Another report worth looking at would be % of times KVPs are properly passed to DFP, you want a partner who is efficient at bidding on your impressions since they are exposed to 100% of it. Hope this helps!

  3. Hi

    Thanks for your posts. I am wondering what is your “strict timeout policy” (e.g. <500mS?) and how the vendors you are testing are performing? Would you be able to share stats on each vendor's average response times?

    Thanks

    Nicole

    • Timeout policy should be tested to see what works best for both your partners (What time allows them to push the maximum number of KVPs) and your page (What time is too long and starts costing you impressions served per pageview). There really isn’t an ideal number, you need to test and see what’s best for yourself. You could also just run them all async and when the call hits GPT the ad server is fired.

  4. Hi Emry, Thanks for the detailed explanation, great post!
    As you mentioned briefly, these tagless solutions are mostly a way for many 3rd party to collect Publisher data and fuel other impressions on other websites with these data. It’s exactly what DFP is doing for Google but not sure Publishers want everyone to access their 1st party data this way.
    This being said, we see all these 3rd providers now offering server to server integration thus allowing a real tagless implementation: integration is seamless for the publisher and he controls everything on an impression level from a central UI. Data collection through cookie sync pixels remains questionable but it’s much more transparent (and controllable) than an in-page tag implementation!

    • Thanks Romain. Would you mind expanding a bit on this part of your comment, “This being said, we see all these 3rd providers now offering server to server integration thus allowing a real tagless implementation: integration is seamless for the publisher and he controls everything on an impression level from a central UI. Data collection through cookie sync pixels remains questionable but it’s much more transparent (and controllable) than an in-page tag implementation!” Thanks for reading!

Leave a Reply

Your email address will not be published. Required fields are marked *