On February 21st, 2018, we hosted an AMA with Sovrn’s Support Engineering team. We received many questions around Ads.txt, header bidding, & DFP.
Below is a transcribed version of our live AMA.
Q: What’s the deal with ads.txt?
A: Ads.txt is a simple verification method used by DSPs/buy-side platforms to ensure that the SSPs/Exchanges and pubs on those platforms have permission from site owners to present their inventory for sale.
Q: How do I know that header bidding is a good idea for my site?
A: We have learned that having a stable setup that is working through DFP, that you understand & that is effective, is a win. Start with waterfall & then move towards header bidding. You want to make sure you have a few partners queued up prior to moving towards header bidding; one partner will not yield the best results.
Q: What are the benefits of using the Header Complete product Sovrn offers vs using other auction in the header options?
A: One of the biggest positives, is that it is all handled & manipulated in the Sovrn user interface. There is only one piece of code to place in the header. Any changes you make, are handled within Meridian. This helps with latency & we can collect more data to help you troubleshoot different aspects of your setup. The entire header complete solution is built off of prebid. The more implementations we do, the more we realize: the simpler the better. All inputs for server side config, have been tested to work, producing less room for error.
Q: At what point do you find publishers having diminished returns (partner wise) in their wrapper? For a scaling publisher, how many demand sources would you recommend?
A: The prebid.js documentation recommends 1-5 demand partners. However, latency is the big question – how long are you willing to have the ad server wait for all of the bids to return from the header auction? Although, 1-5 is prebid’s recommendation, they also state you can have up to 10 partners depending on the balance between latency & revenue. We recommend you add around 5, and experiment from there. Our most successful partners have between 6 & 8 partners. All publishers & partners are different. The prebid docs provide a great way to test for your optimal number of partners.
Q: For scaling publishers, are you seeing a shift towards server side auctions or do you think client side will remain the most popular in the next year? Or Hybrid models?
A: We think server side will gain popularity in the next year, but client side might still be the norm. There are less connections to be made if S2S is done right. We would suggest hybrid models as you make a transition. Do some A/B testing & then you can move over to one or the other 100%. The important thing to note about hybrid, is some publishers may work better on either side. It is better to keep each partner on the side that they perform best.
Q: If you are a small ad network, Is it possible to set up multiple sites within one DFP account?
A: It is preferable to use one DFP account rather than multiple. This will eliminate the complication excess and competing multiple line items. There is a lot you can do within DFP to fit specific needs.
Q: What would you say sets Sovrn apart from other ad tech vendors, exchanges, and ad networks?
A: First, we view our publishers as business partners as opposed to clients. Our main focus is publisher success, not just slinging ads. We have an online Community where you can communicate with Sovrn employees as well as other industry experts & publishers. There are several resources in the Community to help keep up to date with industry trends. We are also working to build out solutions outside of just transactional ad tech, for example, heightened reporting capabilities, hashed email & VET. We provide solutions that help publisher grow, even if it means Sovrn isn’t the main benefactor of what we implement.
Q: What is Sovrn’s view on publishers who use sovrn demand directly integrated through the header and also via tag-based 3rd party vendors who provide publishers with Sovrn re-seller ads.txt entries? Does using both sovrn directly and as a reseller put the publishers direct seat at a disadvantage as the buy-side could potentially have access to the same impressions through a cheaper supply path via the re-seller?
A: This has become very common in the industry, therefore, it is impossible to answer generally. Re-sellers can oftentimes have very functional tech! Some of these vendors’ products are widgets or otherwise additive inventory that would not normally be monetized. Rule of thumb: if it is working, don’t fix it. If something is working well for you, makes sense to you, the page is loading, why go and try to fix it? Pick one or the other if you are not seeing success.
Q: Does the ads.txt file actually do anything? Does it block unauthorized resellers from selling a publishers inventory? How will this impact fill rates?
A: Yes, DSP’s adoption rates are steadily increasing, so it will be important for pubs to keep accurate ads.txt files hosted on their sites because, as adoption rates by DSPs increase, fill for publishers not utilizing or incorrectly implementing ads.txt files, will see a decrease. And yes, for DSPs who are utilizing ads.txt, it will block unauthorized sellers and resellers from selling on their platform. Important to note that not all DSPs are currently using ads.txt, and others have varying degrees of strictness.
Q: Are there any tips on how we should target Sovrn tags in DFP, specifically for Geo targeting, Device/Browser targeting & Frequency capping. In Meridian, I don’t see a breakdown by geo, only by US and “Other”. Is there any data or report that I can pull to show performance by geo? I was thinking that if the CPM or fill rate differs significantly by region, then it would make sense to create separate line items by geo in DFP rather than than a blanket line item that covers everything. Same applies to browser or device.
A: Specifically for geo, through DFP, you can use their built in queries that are available in the reports section. The ‘network geography’ & ‘device & browser‘ reports are a good place to start. If you need more granularity, you can write your own custom query. Frequency capping can be used as a tool do some A/B testing to determine which partners are performing better. Breaking out tags by Geo may improve performance if you have a significant portion of international traffic.
Q: Should the Sovrn tags be frequency capped and if so, at what number? Is it more of test-and-see approach dependent on the site? If that is the case, what metric would be most helpful to determine a frequency cap? Fill rate or yield?
A: Frequency capping was originally intended for direct and guaranteed line items to ensure that guaranteed and repetitive creatives were being properly distributed across a site’s readership. With that said, playing around with frequency caps at the top of your stack can be a good way to grab at the highest possible CPM’s for your best and most unique traffic. The easiest metric that we would suggest focusing on is the eCPM and overall yield per partner and ad unit. These metrics will change depending on location within the stack and price floors but testing them against each other will help to get a better understanding of bid distribution. If you do decide to frequency cap at the top, we would suggest implementing secondary tags of the same partner further down the stack to test performance at that mid-quality inventory as well.
Q: Why (or why not) is it advantageous to run Sovrn (or any partner) on a client side wrapper and via a 3rd Party S2S (Google EB) simultaneously? Many publishers report seeing incremental lift from combined implementations. What is the reason for the lift? Is the Google EB auction being primed by the results of the client side auction?
A: The combined implementation is probably not the reason for the lift. If you switch half your tags to EB, and that is working better for you, you will see a lift because one half is working better. The EB auction is not being primed by the results of the client side auction. We suggest you test to find out what works best for you. For EB especially, there are timeouts & issues – depending on where your traffic is coming from, this may add latency. We suggest combined implementations only to find what works best, and then move full to one or the other. Tech will vary significantly from site to site. We are here to help publishers with testing by providing reports & troubleshooting.
Q: Is it normal to see differing performance by geo?
A: Traffic will differ drastically by location & type of inventory. Sovrn has a demand team focused on increasing performance across all geos. The more data we have in this area, the better we can serve our publishers.
Q: In your opinion what is the site performance impact of using header bidding verse using standard sovrn tags?
A: It depends on your site, the amount of traffic you are getting, whether or not you have an ad server, etc. There are a lot of considerations when moving towards header bidding. It depends on how exactly you are implementing the stand alone tags & header bidding.
Q: How does win rate affect your header bidding algorithm? If bids are requested on a sizemapped or lazy-loaded unit that may never get sent to the adserver, does your algorithm penalize this behavior? What about bid caching?
A: We have observed that the sites with 1:1 ratio between ads requested & slots to fit into, perform better. Though we don’t know to what extent, demand does partially throttle requests from sites that send unfillable tags. Publisher bid caching can cause impression timeout issues. Sovrn does not engage in bid caching in our auction and we strongly discourage publishers from this practice, for that & other reasons. If the user takes too long to scroll down to that ad unit, there is a chance that the impression has timed out by then. Your ad server might record that as an impression because it went through the whole system, but it has timed out on the SSP side. This causes discrepancies. The impression timeout varies from 30-60 seconds. If utilizing lazy loading, use separate Prebid.js auctions for those units when they are actually on the page.
Q: In the header bidding and EB world, have optimizations now come to an end? Do publishers now simply have to focus on the best SSPs/demand partners or do CPM floors still help increase revenue?
A: This is something that is always changing. Buyers have their own algorithms for determining how much they want to spend on a site & it can change week to week. Setting floors is a big focus for us currently from a data perspective. Now that header bidding has become more widespread, we can start to dig in deeper into our transaction flow to ensure we are optimizing as best as possible. Whatever you have set as CPMs in DFP as direct deals or price priority waterfall will dictate what the floor needs to be to be effective.
Q: What are some of the biggest pitfalls that publishers experience when implementing header bidding?
A: The more straight forward your website is built, the easier it will be to implement header bidding. Simple is better. Sticking to what is in the spec & a wrapper that works, is very important. Competing line items in DFP (waterfall & wrapper) with inflated rates, can override one another and could end up serving a lower CPM than the header bidding line item. You have to be open to testing.
Q: From Sovrn’s point of view, what are the benefits and drawbacks of using DFP over other popular ad servers such as SMART ad server?
A: DFP has become so universal that there is so much documentation & references. On the other hand, it is hard to get ahold of someone at DFP. A few publishers have told us they are moving towards a different ad server for more support. DFP does have so many different options & overcomplicating your setup can mess with your performance.
Q: What are a few things successful publishers do to increase fill & yield.
A: Optimizing the bid request is important – making the right calls. On the other hand, serving a few different sizes into one placement can help achieve the highest CPM. A publisher that is willing to go with the spec & do things in a way that is cohesive with the wrapper is beneficial. We are currently looking for best practices that we can share with our publishers. Stay tuned!
Q: Do certain placements (higher up on the page) perform better than others?
A: Yes. Making sure to split out tags above the fold & below the fold, is really important. A leaderboard will perform better at the top of the page rather than at the bottom of the page. 728X90s, 300X250s, 160X600s & 320X50s will be some of your best performing sizes. Depending on your traffic, it is smart to split out by geo and device type as well.