Gavin Deadman Discusses DSP Algorithms, In-House Exchange-Buying, And The PV Impression

I did a post recently on why I thought SOME advertisers would bring exchange-buying in-house. One big advertisers to recently put in place such a strategy was King.com – a leading social gaming company based in London. ExchangeWire recently spoke to Gavin Deadman, DSP Manager at King.com, about the in-housing of the DR display buying function, and how employing such an approach can benefit a direct advertiser like King.

Can you provide an overview of the King.com exchange-trading strategy and your DSP manager role?

GD: When using a DSP, the exchange/SSP you buy across is determined by the algorithm similar to all of the other criteria such as max. bid price, publisher you run on etc. I’m a big fan of putting in a CPA goal and then letting the algorithm decide where to deliver it in order to hit target. Our strategy requires a DSP with an algorithm – and we only work with vendors that have that kind of capability. After much testing we have settled on one partner.

Our goal is to market King.com products across DSPs on a self-service basis – bidding and optimising on both prospecting and remarketing strategies globally. In a nutshell I do the same job as SEM managers but for display and more.

What was the motivation form bringing the exchange-buying in-house?

GD: To save money by cutting out the middle man – and run display DR as it should be done. We have time to go granular with bid strategy setup with learnings attached, making it easily accessible to the remainder of the business – marketing, product design and media all under one roof.

Do you work with any ad networks? How does internal buys compare with ad nets in terms of performance?

GD: We do work with some ad nets. Certain ad networks still have very niche targeting features, which certain exchanges still have to come out with and then link via DSPs. I am currently working on a project to integrate ad networks into the DSP we are using to give us access to additional inventory. It has never been an issue anyway but it’s still nice and for the algorithm to decide whether there are sections of ad networks which perform better than exchanges .

I have built up a variety of case studies now for DSPs vs ad networks – and DSPs have come out on top every time. DSPs have a 47% higher CTR, deliver ads 90% cheaper in terms of CPM costs, have a 93% cheaper eCPC and most importantly delivers acquisitions 93% cheaper. Also the conversion rate on a POST CLICK basis is 1% higher.

Are you doing mostly doing re-targeting – or is prospecting the main focus for your DSP buys?

GD: Prospecting is the main focus as remarketing is something which has an ‘always on’ strategy applied similar to brand search. In terms of budget allocation, roughly 70% prospecting and 30% remarketing as there are only a certain amount of users you can remarket (with a respectable freq. cap applied of course). We are now looking at what effect remarketing has on the algorithm when optimising prospecting campaigns. Re-marketing has goals further down the sales funnel (targeting lookalikes), because all display tech optimises on a last view / click basis. This results in re-marketing stealing the original click / view from the prospecting campaign – and makes it harder for the prospecting campaign to pick up enough data to optimise on.

Obviously King.com can leverage a lot of proprietary data. Do you use third party data vendors as well? How has it performed? Is the data market still underdeveloped here in Europe?

We have used BlueKai and Exelate for US campaigns but are waiting until we have enough back end data to determine whether the extra cost is worth it. I still certainly think that data integration with DSPs is not there yet. There is still confusion about pricing models and the volumes are not available in Europe where the data exchanges have integrated with DSPs.

Has automated buying made it easier for King.com to run display campaigns?

GD: Certainly. I have time to traffick, setup bid strategies, trading, display marketing strategy, optimise, develop creative concept and reporting. Historically you’d need at least 4 to 5 people to do all of this. Without a DSP we simply wouldn’t be able to warrant bringing display in house

Do you think DSPs and other buying platforms being used to its fullest potential at the minute? Or is the industry too focused on chasing the last impression? Is the post-view impression window distorting the real picture?

GD: DSPs are not being used to their fullest extent. A lot of marketers are still demanding agencies ‘get the numbers in’ even though 100% of PV conversions are not genuine. I think that until marketers are brave enough to start challenging and adapt to change then media attribution models (and not just digital) will be at the top of everyone’s agenda over ‘spending’.

Six months in and we are only touching the surface when it comes to DSPs. The next step will be more data tests, integrating Facebook, mobile, testing expandables, video, integrating SEM and then the adserver – all of which will be coming over the course of this year.

You would normally give an ad network some creative tags and they’ll deliver it. There’s a lot more thought that needs to be put into DSP campaign setups- like bid strategy testing, optimising on different ROI tags and towards different effective buy structures. We also use DSPs as an analytical tool. A lot of time gets put into our SEM keyword / campaign structures and DSPs are no different. I currently have over 150 bid strategies on the DSP so it’s a lot different than the usual 5 IAB formats, budget with goal, and go live process

Do you think display buying could be as effective as search post-click?

GD: Looking at our post click performance, we are so close to Facebook and SEM – but DSP still has more budget than both. Exchanges will ultimately work even closer with publishers, passing all data back via an API to DSPs. This will allow the algorithm to learn more and become even more targeted. We should expect everything to increase from delivery. The big development I’m excited about is exchanges passing back custom data from publisher registration fields.

I see a lot of UK CTR data – and the numbers clicking ads is decreasing all the time. Ads nowadays are not targeted correctly or the ads looking pretty bland. I think there will be more targeting options available (like dynamic targeting for prospecting), marketers will be managing the creative, media and marketing process more closely and that will bring a wealth of post click conversions across a variety of audiences.

Is King.com the start of a bigger trend of in-housing of media buying within the display space?

GD: You currently have one buying tool for Facebook, one for SEM and one for display and mobile. A typical business looking to make more margin would now see this as an opportunity to in-house media buying. Something which will hold this back in the short term is specialist resource. But the new breed of marketer will be a media and marketing manager doing both. There will be an influx of new talent from universities going straight into the advertiser directly. Tech providers will train them on how to use their tools. Everyone will have to adapt.

On certain DSPs, volume and reach, quality, algorithm / optimisation tech, targeting and self-service are all there and I can’t see why any display advertiser wouldn’t use them to drive incremental volume cost efficiently.