A dive into all things contextual with Gourmet Ads’ Benjamin Christie 

With Chrome’s identifier deprecation approaching in the year up ahead, many advertisers have wisely turned their efforts to contextual solutions. We spoke with Benjamin Christie, founder and president of Gourmet Ads & Healthy Ads, who shed light on how advertisers are using contextual in the run up to cookie deprecation, how the challenges of creating a contextual targeting product can be overcome, as well as how contextual products are set to evolve over 2024. 

Why build your contextual targeting solution when there are many on the market?

There are a variety of reasons we built our own Contextual Targeting Solution. In the past, we tried using several contextual targeting services, but we found that they were not suitable for both our advertisers' needs and our operational requirements. They either lacked accuracy, scale, transparency or in some cases recency. 

Today, with our own contextual solution, we offer activation via both Managed Services and Programmatic. The vast majority of our advertisers are running Programmatic Advertising and we activate via Deals and running via their own DSP. When talking to both prospective and existing advertisers, they emphasised transparency as their top priority. They wanted to know which pages their ads would run on (or at a minimum, a sample list). Therefore, we developed our architecture starting with URLs and allowing for sample lists to be shared. By building our own contextual targeting solution, we have complete control over the URLs that are included — or more importantly, not included — in a specific segment.

What use cases are you solving with your contextual offering? 

The most important use case is replacing third party cookie based segments with comparable content and scalable contextual segments. This has come through deep consultation with our brand direct advertisers as well as agency trading desks. They are demanding a solution that is privacy-compliant, cookieless, highly relevant, scalable, and cost effective.

For food campaigns, most advertisers use third party segments best categorised as “in market”. They are built using past grocery transactional data of consumers that have bought specific products. Using steak as an example, our steak segment is built around content such as steak guides, steak crusts, steak brands, how to cook steak, and steak recipes.  

On the health side of things, it's a little different. Privacy is the greatest issue. We are replacing very broad health segments (some very hit or miss) with highly specific medical conditions which are privacy-compliant. For example, to target content around a condition like asthma, our segment contains highly relevant keywords and phrases together with the names of medication brands and their active ingredients. We store no user data, just visitation of content in real time.

How do these vary in comparison to contextual products in other verticals?

We built our solution from the ground up to help our food and health clients. To do that, we’ve indexed over 250k domains and over 7 billion URLs — quite the significant dataset! Our focus is on food and health brands, and our off-the-shelf targeting segments reflect that. But we can also use that vast dataset to create non-endemic segments for any possible vertical.

The funny thing we are finding with advertisers is that after we provide a couple of contextual based deals, we are getting asked if we can build non-endemic segments. Only last week, we had a request for a broad-reaching Electric Vehicle segment as well as a segment to reach cattle and sheep farmers looking to vaccinate their herds.

What are the key challenges associated with developing a contextual targeting product, and how can these be overcome

When we started the process of building a contextual targeting engine back in 2020 (in the middle of COVID and lockdowns), we had a number of challenges to solve. Some of these included being cost-effective, storing website URLs, extracting the context of a URL, indicating negative or positive sentiment, scaling segments and optimising segments to scale — all which we solved. 

One of the largest challenges we had initially was ensuring we were cost-effective. We never wanted to be optimised off a plan or deal because our costs were too high. Our initial feedback from buyers was that they needed to do “more with less budget.” As such, our plan was to price contextual into a floor price. Today, there are two main challenges we are looking to solve. The primary issue is forecasting of newly created contextual segments. Right now, the challenge is we have to rely on historical data to predict the size of a newly created segment, and it can take seven to ten days of bid requests to see if there is enough volume. Whilst our activation of contextual targeting is 100% cookieless, we still need a cookie in the initial stages of forecasting when a new segment is created. 

The second is recency of actual content. We all know that newly published content is highlighted on homepages, included in newsletters, shared on socials, etc. This means that newly created pages typically have significant volumes of users compared to older content. Our team is constantly working hard on ingesting content in multiple ways to make sure we are adding the latest relevant content to individual segments on a daily basis. Doing this will help scale segments and help campaign delivery. We’ve started working with large publishers to get direct feeds of URLs as they create new content. This has cut the time down from published to included in segments significantly. 

How are both contextual and non-contextual products set to evolve over the course of 2024?

For contextual, I would say that AI is going to be a massive game changer. There are just so many applications that can be implemented. When OpenAI first launched the ChatGPT API, we built it into our contextual engine. We started with building very basic yet specific prompts to either scale a segment or improve the accuracy of segments. 

One thing we’ve found specifically for Health segments is that Contextual seems to work better in smaller, more focused segments or clusters compared to one large segment. For example, instead of having one Diabetes segment, we’ve created 8 segments under the umbrella of Diabetes, which gives the clients using this much more scale on deals. Type 2 Diabetes Medications, Type 1 Diabetes Medication, Diabetes Diet, Diabetes Continuous Glucose monitoring, Diabetes Devices, etc. 

For non-contextual products, I think there are paths. The first I think many will go back to is the good old category targeting that we as an industry were doing 15 years ago. Food campaigns targeting food sites: that’s a little like contextual, but it’s broader and less focused. 

Then there is the User ID front. I’ve said many times that the User ID race is like my favourite sport of Formula One. Each Sunday there are 20 starters, but there will only be a few on the podium come the end of the race. I think User IDs will be the same. There are lots of providers, but only a few will be left within a couple years. And I think the winners will be somewhat dictated by advertisers. 

In the run-up to identifier deprecation next year, how are advertisers using contextual products in desktop and mobile environments? What concerns need to be addressed? 

Both our two brands, Gourmet Ads and Healthy Ads, are seeing an influx of requests for contextual targeting since the announcement of the deprecation of the cookie. There is definitely an increasing cohort of advertisers out there testing alternatives of ID targeting right now. We’ve been asked to layer a contextual targeting segment across an identifier, particularly for building more scalable segments. 

From talking to advertisers and their advertising agencies, we are hearing two concerns. 

The first is the general question of “what does cookie deprecation mean for us?” Whilst the transition to deprecation of the 3rd-party cookie has been talked about for some years now, some media buyers and traders don’t really understand what it means for them and their advertisers. We recently analysed 100 calls with our accounts team, and the vast majority had admitted they were not ready for the transition to going cookieless. 

The second question is simply: “will contextual actually work? Can it replace our existing buying strategy?” And that’s not just from a performance perspective but a scale perspective. Brands are advertising to engage and influence new consumers, so the question they are asking is will they see the same level of scale (compared to cookie segments) and will they engage the same from a performance perspective?