A Q&A with Chris Hogg, Technical Account Director, Grapeshot

A Q&A with Chris Hogg, Grapeshot

Digital advertising pros who are also Spaghetti Western fans thoroughly enjoyed Grapeshot Chief Executive John Snyder’s wonderfully titled OPS Markets session on Thursday, Feb. 9: “The Good, the Bad and the Ugly of Digital Advertising.”

While we’re still trying to talk Snyder into wearing a poncho and chewing a cigarillo during his presentation, he did kindly agree to have his Technical Account Director Chris Hogg mosey on down and answer a couple of questions about Grapeshot’s offerings, real-time categorization and how behavioural and contextual targeting can operate in harmony.

 

Read our Q & A with Chris below.


Interestingly, Grapeshot takes its name from an artillery term – what’s the story there?


Yes, you’re right it’s more than just a company name, it describes the way that we analyse words contextually on a page. Instead of looking at single keywords that can give many false positives, we look for a cloud of words together to clarify the meaning of a page. In the same way, a cannon can either fire a single cannonball or fire out many smaller types of ammunition like a grapeshot.


Although behavioural/audience targeting is all the rage in the digital space at the moment, what factors do you see driving contextual targeting back into the limelight?


Although, it doesn’t sound long in terms of other industries, I’ve been working in online display for 14 years. Formats and techniques come and go – like the pop-up, remember that? As soon as people realise that a format or technique is negative for a brand, they stop using them.

 

Who hasn’t these days had the experience of being cyber-stalked? I get around eight Audible.com ads a day. I’m already a customer, I don’t want to buy any more. In fact, I am getting really annoyed by it.

 

A lot of behavioural targeting isn’t very clever, it is in fact just cookie re-targeting. That’s been around for years. And the truth is that after six months, retargeting results drop off the end of a cliff, and you have managed to wind your audience up.


There is research that suggests that people don’t mind ads being contextual to what they are reading. However, there is also research that shows that people do not like their privacy being infringed. Contextual advertising falls into the first category. It doesn’t use cookies in any way.

 

I think that there is a slow realisation that re-targeting needs to be done more skillfully if it is to find its right place in the mix of online marketing methodologies. Contextual advertising in comparison has a long, established role in advertising effectiveness. Not surprisingly Google has an important role to play here, they are constantly publishing research showing contextual advertising out-performing re-targeting – however, with the focus on re-targeting and behavioural, it isn’t getting the attention it perhaps should.


How is Grapeshot’s contextual analysis similar and different to semantic targeting techniques?


Okay, so here is a very quick journey to the geek side of contextual advertising. There are two main methods: probabilistic and semantic. They both try and understand what a page is about, what is its DNA?

 

Grapeshot uses a probabilistic model. This means that we look for patterns of words together, to understand what a page is about. We don’t have complicated dictionaries that understand the meaning of the words on it. In the same way that a navigator can use the patterns of the stars to navigate where he is without actually knowing anything about the stars that he is using, Grapeshot uses probabilistic maths to understand the relationship between words on the page without knowing what the words mean.

 

The probabilistic model has some really important pay-offs in terms of flexibility and speed. It means that new contextual channels can be built by ad ops execs without much prior knowledge. It means that we can work in any language and finally it can be deployed in very fast environments like publisher ad calls and RTB situations.


How does real-time categorization work? What are the advantages of such flexibility?


With publishers, Grapeshot places a small JavaScript tag on the page. This fires every time a page loads. It calls Grapeshot, which then sends its contextual analysis back as a key-value into the ad server ad tag as it loads. If a contextual ad has been targeted in the ad server to that key-value, then a contextual campaign is served.


With agencies and advertisers on the RTB side, Grapeshot is passed the URL on an API level by the SSP or DSP Platform. We then analyse those URLs and store the keywords next to the URL. If we see that page again and the advertiser or agency has created a channel that is relevant to that page we pass that information back to the DSP and if the bid price is sufficient then an ad is served. This all happens in 6 to 10 milliseconds.


The advantage of Grapeshot’s flexibility for both publishers and agencies is that it allows them to create custom contextual channels, and to change these whenever they like. This means that advertiser’s briefs are answered to their fullest extent.


How does Grapeshot blend contextual targeting and behavioural targeting? Do you see a hybrid model prevailing in the display space?


I believe that there is a way of blending contextual advertising and Behavioural targeting. It isn’t done much at the moment. If you behaviourally target IN CONTEXT, then your ad will seen to be relevant to the user and not negative to the brand. If you re-target OUT OF CONTEXT, then after a while you will effect your brand negatively. In some ways there is no choice but for the two disciplines to blend.


Publishers on the whole are wising up to the value of their first-party data. Any decent ad server will be helping them by stopping any kind of data leakage. However, the next stage is to take control of their first-party data. Context is one of the most vital pieces of data – knowing what articles have been read by a user in the last three months is much more detailed and valuable than knowing simply what sections of a site a person might spend most time on. That is the real power of the publisher’s data, the detail that allows really powerful campaigns to be placed. Grapeshot will be there helping publishers when that time comes.


How does Grapeshot ensure brand safety – e.g., nixing airline ads on a news story about a plane crash?


Advertisers help fund all types of wonderful online content. I think the least publishers can offer is the chance to appear in the best light possible. Even on premium websites, there is still content that an advertiser might not want to have his brand associated with. How many times, do campaigns get pulled because press is deemed to be too negative and pervasive to support the risk? A lot.

 

Take the Toyota re-call. That story went world-wide, disrupting hundreds of different online initiatives. In this situation Grapeshot allowed its publishers to create a custom negative channel for Toyota. These included words like ‘re-call’ or ‘crash’ and the campaigns continued. If we deemed a page to be in that category, the Toyota ad was not served.


On the RTB side of the business, Grapeshot allows the same flexibility in terms of creating custom negative channels, but it also automatically blocks bid on pages that contain any hint of illegal downloads, race-hate, sexual content, violent content or unmoderated forums. There is a lot of muck out there in RTB world and it useful to be able to block bids. Most ad verification companies allow you to send blank ads for dodgy content, but Grapeshot doesn’t even bid.


How can publishers best take advantage of Grapeshot’s sell-side services?


Well there are three ways. First, respond to advertiser’s briefs in the most forensic way. This usually results in winning more business, because in this competitve market, it’s ‘he who cares wins.’ Grapeshot allows you to carve your content in the best possible way for the advertiser.


Second, if you have lots of unsegmented traffic, say in news, why not let Grapeshot see if it can help turn this into more valuable and sellable categories? It really unlocks the potential of their valuable content. We have found customers who didn’t realise that they had enough content about food to create a distinct channel or, more valuable for one network, a health channel.


Finally, why not experiment with setting up higher price floors on the inventory you give to networks or SSPs by giving it a contextual flavour?

 

OPS Markets

Up to date on the latest happenings in the automated trading space? OPS Markets brought digital advertising leaders and ops professionals together to discuss and develop best practices for operational excellence in the evolving automated landscape.