In its second piece on the rise of the ad tech app ecosystem, ExchangeWire speaks to another European player building a buy-side app that can be run across multiple platform partners. Here CEO John Snyder discusses the Grapeshot app in more detail and its approach to building on the pipes of the new programmatic eco-system.
What is the app? How does it work?
Grapeshot Keywords was launched in April 2012 on AppNexus App Marketplace and quickly had 50+ buy-side customers using positive and negative keywords to target and avoid certain pages in the AppNexus bid stream. Over 12 billion ads a day have the actual pages analysed by Grapeshot, with each page profiled at the raw keyword level. Grapeshot channels are also defined by keywords, giving the power for customers to craft their own keyword sets for each brand – and opening up full transparency of how segments are defined. The app also shows matching inventory – the type of pages in which your campaign will run – for each channel.
We believe the transparency and safety of using keywords is the tipping point for getting more brand spend on RTB – as brands have always had keywords defining their needs. Showing the keywords which define standard or custom channels, as well as the matching inventory wins customer confidence. It is what RTB needs to get those brand dollars coming over to RTB.
Who would be a typical user?
Trading desks and RTB exchange specialists who want to create proprietary segments tuned to their Brand. Toyota may need “hybrid, environment, climate change” and other keyword contexts, but also to avoid pages about “crash, battery recall, accident”. Banks need to find credit cards and mortgages but avoid contexts around bankers’ bonuses, fraud etc. Each brand has specific keyword needs – to target, as well as to avoid toxic contexts.
What does the product solve?
Programmatic buying is beset with manual processes, like white-lists. Serving a bank brand into white list content like New York Times where the article is about excessive bank bonuses and libor-rate fixing is just not adequate. Brands need to access the long-tail in a safe and targeted way. Context has always been the motherhood of traditional advertising, and RTB should be no different. Grapeshot makes it easy to isolate brand relevant contexts that are safe and performing.
What are the opportunities around scale and distribution by leveraging a platform like AppNexus/ Google/ Turn/ MediaMath?
Immense. We found 50+ customers without trying. It is immensely satisfying to meet your customer in person for the first time many months after they have been using your app and services on AppNexus. Grapeshot Keywords has quickly become the most used app on Appnexus – which is not about downloads, but regular day-in, day-out usage. We are proud to work closely with Xaxis EMEA, one of the largest buyers on AppNexus. Keyword contextual gives big brands the confidence to buy RTB.
Turn and MediaMath have other benefits as a platform: centralised billing and different reporting tools. Each DSP has different interfaces and features. What is special about AppNexus is the strong technology and API interfaces which underpin a fantastic “innovation platform”, as I call it.
AppNexus seem to have embraced the app model. Do you think the app strategy will be adapted by other platform vendors – like some of the DSPs or maybe MediaOcean?
Apps sit on top of APIs. Anyone in software knows that producing a formal set of APIs is harder than simply writing your own software and getting it out the door – API delivery takes more design and discipline. To retro-fit an elegant API infrastructure is a hard job for any software team – so I think AppNexus have an inbuilt advantage here by starting with open APIs at the outset.
Will this actually make it easier to build ad tech companies. No need to raise ridiculous sums of money to build out the infrastructure and resource to distribute the product – when the pipes have already been built by a number of large infrastructure players?
The pipes are important – but all that means is that you now have a large waterhose of data hitting your own servers, as many billions of bids need to be analysed for contextual relevance, not just a few billion. Pipes make it easier to integrate, but when the waterhose is switched on, you still need the infrastructure to cope. We are fortunate that our codebase is so compact and scalable, we only run a handful of servers in three datacentres in Los Angeles, New York and Amsterdam to cover the whole daily AppNexus traffic – which we analyse raw and full-on. With other platforms there are often caching layers between us and the waterhose of bids, but on AppNexus there is no such filter: we take it all, full-on.
Often point solutions in ad tech find it very difficult to build sustainable models in the space. If a product is successful it will be aped by a stack offering thus reducing its competitiveness in the market. Does an App strategy give ad tech companies the focus to not only innovate but also build a sustainable business?
If your app can be aped, then you do not have a differentiated product capability. As co-founders at Grapeshot, we have been operating in search software since 1980 and delivering enterprise scale systems since 1992. Grapeshot is a new stack of intellectual property built between 2001 and 2008 before reaching the ad market in 2009 – so I just do not think anyone can copy what we do. They do not have the same 30 years of deep search experience and the research bedrock we have had at Cambridge University. The app interface and workflow is what can be copied, and product positioning – but not the core technology underneath.
What platforms are you currently connected to? Is there risk involved in being reliant on one platform?
Grapeshot has been serving sell side premium publishers since 2009. We count Thomson Reuters, Mail Online, Microsoft’s NineMSN, Glam Media and many more as customers – but we are new to buy-side. The Appnexus app has given us 50+ buy-side customers almost overnight and the fact that Grapeshot Keywords is one of the most used apps on AppNexus, and we charge three times the price of our contextual peers, shows us keywords are as important as Google has always made them . It is only natural when different agencies use different DSPs, that we integrate with other platforms. We are integrated with TURN, Ignition One and a few other platforms, with MediaMath forthcoming. Instead of Google holding keyword contextual within its core product offerings, we aim to open up keyword control to a variety of platforms and exchange partners.
Are there technical challenges in being a connected API partner? In terms of data pipes / flows, does being a separate layer to the ad server represent missed opportunities in terms of fully maximising the potential of the product?
Good question. We deliver keyword contextual insights pre-bid – so we know the words on the page in which the ad is about to be bid, and served. There is a huge opportunity to switch the creative in real-time based on that context. A TV brand can change the creative ad served, based on whether page context is around TV, smartphones, gaming, music or movie interests – so the opportunities for Dynamic Creative Optimisation is huge. Data providers still find the pipes and platforms inefficient in passing pre-bid data through to the actual adserver for creative switching – but this is set to improve as DCO takes hold. For example we equip a client where 200 new creatives refresh every 24 hours, and the creatives each have a keyword contextual segment – running across AppNexus. But the platforms are not really designed for this degree of real-time creative flexibility quite yet. For a yellow pages company we switch between 40,000 creatives based on contextual decisioning – most platforms cannot cope with that diversity of DCO right now.
Is there a risk to the number of apps emerging that operate on a pre bid basis? How many points can realistically be pinged, before submitting a bid request, before you actually time out?
Platforms tend to require answers back in 10ms. We try and do it in 1ms – so I see no problem here. The platforms send out requests, and if the responses are not in time, then they just carry on without that data segment. It is up to the app providers to get their responses in. The real issue is a crowded app marketplace – just like iTunes perhaps. Users may get confused about which app does what. However once people know certain apps are good, word gets around.
Does Grapeshot offer a solution in terms of making premium inventory more relevant (and thus helping to increase monetisation opportunities) or is it a way of cleaning cheaper, easier to access, long tail inventory? Or both?
Grapeshot traditionally serves sell side. In RTB we see publishers wanting to mimick their direct sales of publisher editorial channels into virtual Grapeshot channels on RTB defined in total alignment with their own editorial sections/segments. Then, on top, crafting the “Christmas” channel or “pet care” channel, or whatever the advertiser needs, allows them to charge higher CPMs – and this can be done in the private exchanges. However on open auction RTB traffic, it is usually the buyer with the upper-hand, as they can use keyword contextual to make sure they are buying users inside relevant content.
As in the above example, bank brands need to be in context about mortgages or credit cards, but NOT bank bonuses and fraud. Secondly with contextual they buy the cheaper URLs with no cookie competition and get relevant CPMs at $0.50 rather than $3.00. This is without the effort of whitelist, and with full keyword control to cherry-pick amongst the longtail the exact pages that deliver performance because the impressions are genuinely contextually relevant to the brand. The concept of while list needs to switch from manual assessment of domains, down to the keyword contextual relevance of actual pages. We know negative stories about Bank brands in our premium national newspapers are very bad for those advertisers. Same as airlines which crash, or chocolate that makes people obese. Whether seller or buyer – contextual relevance still matters a lot more than just the cookie retarget.