Thanks for the questions, Jan. Haha, no worries, happy to answer whatever I can. My answers are a little all over the place as well =)
- How do you feel about the number of header bidders versus latency versus revenue?
It's a great question and as you indicated, there really isn't one specific answer since every pubs setup, site and goals are going to be much different. We're nearly 100% driven by programmatic revenue so header has been huge for us.
In terms of thinking about latency, you first have to asses what your capabilities are in assessing performance. Something every publisher can measure no matter their team size or analytical capabilities is how header (or adding header partners) is affecting your pageview to impression ration. Track your PV growth YoY, establish that baseline. Track your impression growth YoY, establish that baseline. Mark the days you went live with a specific header bidder or header bidding if it's new, what sort of impact are you seeing? Ideally your PV and impression growth is closely tied but if you add a new partner and all of the sudden you're seeing impression growth no longer keeping up with PV growth, you may have an issue.
(Keep in mind things like ad blocking will impact this which is why finding your current course and speed baseline is so important for accurate measurement)
When we first went live with header we relied on things like this and while it's obviously far from perfect, it's much better than nothing.
Today we're able to roll a new header partner out to a specific segment of traffic, say 5%. From there we're able to look at how that partner is impacting both revenue (what lift are they providing) and GPT fire time (latency) and with that information we make a determination about whether the increase in lift makes up for the increase (or hopefully there isn't an increase) in GPT fire time.
Keep in mind that GPT fire time can be impacted even with a fixed timeout.
So that's a long way to say, I don't think there's a right answer but understanding what's right for your property is really all about understanding how you're going to measure. If the first version of what I described doesn't suite your business needs then I think the conversation becomes, how do we improve to the point where we can get closer to the second method I described. Or I suppose this is a good time to challenge the industry as a whole to figure out how publishers can get this sort of data in a way that's clearer to understand.
Simpler ways to think about it would be talking to other pubs who might have tips on implementation or partners they've seen and fixed latency spikes from. How did they achieve that? Take those best practices and put them to work.
- Question: Or have you gone full S2S or single-request?
I love the idea of S2S but we're still on client side. I don't think there's a clear winner in the space yet and while I'd love to explore it, I'm not overly eager to be an early adopter until I have a clearer sense of what will shake out. We continue to look at options and will be testing sometime down the road.
- What analytics tools do you use to evaluate your header bidders?
We use a combination of Google Analytics with imported DFP data as well as Kibana for more realtime/performance related metrics. Kibana handles the visualizations but a ton of effort and engineering went into getting us to a place where that could be evaluated as a realtime optimization tool. We'd need to do an AMA with engineering for that complete download!
- Question: Also, you say that it was 88% ad focussed at SBM. What other revenue streams have you added?
Our sites offer a subscription tool as well which allows users to sign up for extended features and functionality.
Lastly, you say you spend a lot of time researching trends in the space. What are your predictions for the big trends in 2017?
Server to server will continue to emerge and it will matter but client side header bidding will remain relevant.
More and more ad budget will shift into PMP as agencies and brands are realizing the controls and efficiency increases it provides.
EBDA will roll out and it will matter to publishers in the space but it will be a compliment vs. replacement for code on page for advanced publishers and an excellent tool for publishers with high direct sell through rates. It will also raise the bar and improve the conversation around latency, pushing S2S to be adopted (both through EBDA and on page) to more and more pubs.