Contextual ad targeting paves the way for new opportunities
Advertisers and marketers are always looking for ways to remain competitive in the current digital landscape. The challenge of signal loss continues to prompt marketers to rethink their current and future strategies. With many major browsers phasing out support for third-party cookies due to privacy and data security concerns, marketers will need to find new ways to identify and reach their target audience. Contextual ad targeting offers an innovative solution; a way to combine contextual signals with machine learning to engage with your consumers more deeply through highly targeted accuracy. Contextual advertising can help you reach your desired audiences amidst signal loss – but what exactly is contextual advertising, and how can it help optimize digital ad success?
In a Q&A with our experts, Jason Andersen, Senior Director of Strategic Initiatives and Partner Solutions with Experian, and Alex Johnston, Principal Product Manager with Yieldmo, they explore:
- The challenges causing marketers to rethink their current strategies
- How contextual advertising addresses signal loss
- Why addressability is more important than ever
- Why good creative is still integral in digital marketing
- Tips for digital ad success
By understanding what contextual advertising can offer, you’ll be on the path toward creating powerful, effective campaigns that will engage your target audiences.
Check out Jason and Alex’s full conversation from our webinar, “Making the Most of Your Digital Ad Budget With Contextual Advertising and Audience Insights” by reading below. Or watch the full webinar recording now!
Macro impacts affecting marketers
How important is it for digital marketers to stay informed about the changes coming to third-party cookies, and what challenges do you see signal loss creating?
Jason: Marketers must stay informed to succeed as the digital marketing landscape continuously evolves. Third-party cookies have already been eliminated from Firefox, Safari, and other browsers, while Chrome has held out. It’s just a matter of time before Chrome eliminates them too. Being proactive now by predicting potential impacts will be essential for maintaining growth when the third-party cookie finally disappears.
Alex: Jason, I think you nailed it. Third-party cookie loss is already a reality. As regulations like the General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA) take effect, more than 50% of exchange traffic lacks associated identifiers. This means that marketers have to think differently about how they reach their audiences in an environment with fewer data points available for targeting purposes. It’s no longer something to consider at some point down the line – it’s here now!
Also, as third-party cookies become more limited, reaching users online is becoming increasingly complex and competitive. Without access to as much data, the CPMs (cost per thousand impressions) that advertisers must pay are skyrocketing because everyone is trying to bid on those same valuable consumers. It’s essential for businesses desiring success in digital advertising now more than ever before.
Contextual ad targeting: A solution for signal loss
How does contextual ad targeting help digital marketers find new ways to reach and engage with consumers? What can you share about some new strategies that have modernized marketing, such as machine learning and Artificial Intelligence (AI)?
Jason: We’re taking contextual marketing to the next level with advanced machine learning. We are unlocking new insights from data beyond what a single page can tell us about users. As third-party cookies go away, alternative identifiers are coming to market, like RampID and UID2. These are going to be particularly important for marketers to be able to utilize.
As cookie syncing becomes outdated, marketers will have to look for alternative methods to reach their target audiences. It’s essential to look beyond cookie-reliant solutions and use other options available regarding advertising.
Alex: I think, as Jason alluded to, there’s a renaissance in contextual advertising over the last couple of years. If I were to break this down, there are three core drivers:
- The loss of identity signals. It’s forcing us to change, and we must look elsewhere and figure out how to reach our audiences differently.
- There have been considerable advances in our ability to store and operate across a set of contextual signals far more extensive than anything we’ve ever worked with in the past and in far more granular ways. That’s a huge deal because when it comes to machine learning, the power and the impact of those machine learning models are entirely based on how extensive and granular the data set is that you can collect. Machine learning can pull together critical contextual signals and figure out which constellations, or which combinations of those signals, are most predictive and valuable to a given advertiser.
- We can tailor machine learning models to individual advertisers using all those signals and find patterns across those in ways that were previously impractical or unfeasible. The transformation is occurring because of our ability to capture much more granular data, operate across it, and then build models that work for advertisers.
Addressability: Connect your campaigns to consumers
How does advanced contextual targeting help marketers reach non-addressable audiences?
Jason: Advanced contextual targeting allows us to take a set of known data (identity) and draw inferences from it with all the other signals we see across the bitstream. It’s taking that small seed set of either, customers that transacted with you before that you have an identity for, or customers that match whom you’re looking for. We can use that as a seed set to train these new contextual models. We can now look at making the unknown known or the unaddressable addressable. So, it’s not addressable in an identity sense, it is addressable in a contextual or an advanced contextual sense that’s made available to us, and we can derive great insight from it.
One of the terms I like to use is contextual indexing. This is where we take a set of users we know something about. So, I may know the identity of a particular group of households, and I can look at how those households index against any of the rich data sets available to us in any data marketplace, for example, the data Yieldmo has. We can look at how that data indexes to those known users to find patterns in that data and then extrapolate from that. Now we can go out and find users surfing on any of the other sites that traditionally don’t have that identifier for that user or don’t at that moment in time and start to be able to advertise to them based on the contextually indexed data.
Historically, we’ve done some contextual ad targeting based on geo-contextual, and this is when people wanted to do one to one marketing, and geo-contextual outperformed the one to one. But marketers weren’t ready for alternatives to one to one yet. We want marketers to start testing these solutions. Advertisers must start trying them, learning how they work, and learn how to optimize them because they are based on a feedback loop, and they’re only going to get better with feedback.
Alex: Jason, you described that perfectly. I think the exciting opportunity for many people in the industry is figuring out how to reach your known audience in a non-addressable space, that is based on environmental and non-identity based signals, that helps your campaign perform. Your known audience are people that are already converting – those who like your products and services and are engaged with your ads. Machine learning advancements allow you to take your small sample audience and uncover those patterns in the non-addressable space.
It’s also worth noting that in this world in which we are using seed audiences, or you are using your performing audiences to build non-addressable counterpart targeting campaigns, having high-quality, privacy-resilient data sets becomes incredibly important. In many cases, companies like Experian, who have high quality, deep rich training data, are well positioned to support advertisers in building those extension audiences. As we see the industry evolve, we’re going to see some significant changes in terms of the types of, and ways in which, companies offer data, and make that available to advertisers for training their models or supporting validation and measurement of those models.
Jason: Addressable users, the new identity-based users, are critical to marketers’ performance initiatives. They’re essential to training the models we’re building with contextual advertising. Together, addressable users and contextual advertising are a powerful combination. It’s not just one in isolation. It’s not just using advanced contextual, and it’s not just using the new identifiers. It’s using a combination to meet your performance needs.
It’s imperative to start thinking about how you can begin building your seed audiences. What can you start learning from, and how do you put contextual into play today? You are looking to build off a known set and build a more advanced model. These can be specialized models based on your data. You can hone in and create a customized model for your customer type, their profile, and how they transact. It’s a greenfield opportunity, and we’re super excited about the future of advanced contextual targeting.
Turn great creative into measurable data points
Why does good creative still play an integral part in digital advertising success?
Jason: Good creative has always been meaningful. It’s vital in getting people to click on your ad and transact. But it’s becoming increasingly important in this new world that we’re talking about, this advanced contextual world. The more signal that we can get coming into these models, the better. Good creative in the proper ad format that you can test and learn from is paramount. It comes back to that feedback loop. We can use that as another signal in this equation to develop and refine the right set of audiences for your targeting needs.
Alex: If you imagine within the broader context of identity and signal loss, creative and ad format becomes incredibly powerful signals in understanding how different audiences interact with and engage with different creative. In the case of the formats that serve on the Yieldmo exchange, we’re collecting data every 200 milliseconds around how individual users are engaging with those ads. Interaction data like the user scrolling back or the number of pixel seconds they stay on the screen, fills this critical gap between video completes and clicks. Clicks are sparse and down the funnel, and views and completes are up the funnel. All those attention and creative engagement type metrics occupy the sweet spot where they’re super prevalent, and you can collect them and understand how different audiences engage with your ads. That data lets you build powerful models because they predict all kinds of other downstream actions.
Throughout my career, I learned that designing or tailoring your creative to different audience groups is one of the best ways to improve performance. We ran many lift studies with analysis to understand how you can tailor creative customized for individual audiences. That capability and the ability to do that on an identity basis is starting to deteriorate. The ability to do that using a sample of data or using a smaller set of users, either where you’re inferring characteristics or you’re looking at the identity that does exist in a smaller group, becomes powerful for being able to customize your creative to tell the right story to the right audience. When you layer together all the interaction data collected at the creative level on top of all the contextual and environmental signals, you can build powerful models. Whether those are driving proxy metrics, or downstream outcomes, puts us in a powerful position to respond to the broader loss of identity that we’ve relied on for so many years.
Our recommendations for marketers for 2023 and beyond
Do you have recommendations for marketers building out their yearly strategies or a campaign strategy?
Jason: Be proactive and start testing and learning these new solutions. I mentioned addressability and being in the right place at the right time. That’s easier in today’s third-party cookie world. But as traditional identity is further constricted, you will have these first-party solutions that will not be at scale, so you’re less likely to find your user at the scale you want. It would be best if you thought about how to reach that user at the right place at the right time. They may not be seen from an identity basis. They might not be at the right place at the right time when you were delivering or trying to deliver an ad. But you increase your chance of reaching them by building these advanced contextual targeting audiences using this privacy-safe seed ‘opted-in’ user set; this is a way to cast that wider net and achieve targeted scale.
Alex: Build your seed lists, test your formats with different audiences, and understand what’s resonating with whom. Take advantage of some of the pretty remarkable advances in machine learning that are allowing us, really, for the first time to fully uncork the potential and the opportunity with contextual in a way that we’ve never done before.
Jason: At the end of the day, it’s making the unaddressable addressable. So, it’s a complementary strategy; having that addressable piece will feed the models. But also, that addressable piece still needs to be identity-based, addressable still needs to be part of your overall marketing strategy, and you need to complement it with other strategies like advanced contextual targeting. The two of them together are super complimentary. They learn from each other, and it’s a cyclical loop. Now is the time to take advantage and start testing and understanding how these solutions work.
We can help you get started with contextual ad targeting
Contextual advertising can help you stay ahead of the curve, identify your target audience, and continue to drive conversions despite signal loss. We’ve partnered with Yieldmo to help make sure that your marketing campaigns are reaching the right target audiences on the platforms that are most relevant. To get started with contextual ad targeting to reach the right audience at the right time and drive conversions, contact our marketing professionals. Let’s get to work, together.
Find the right marketing mix in 2023
Check out our webinar, “Find the right marketing mix with rising consumer expectations.” Guest speaker, Nikhil Lai, Senior Analyst from Forrester Research, joins Experian experts Erin Haselkorn, and Eden Wilbur. We discuss:
- New data on the complexity and uncertainty facing marketers
- Consumer trends for 2023
- Recommendations on finding the right channel mix and the right consumers
About our experts
Jason Andersen, Senior Director, Strategic Initiatives and Partner Solutions, Experian
Jason Andersen heads Strategic Initiatives and Partner Enablement for Experian Marketing Services. He focuses on addressability and activation in digital marketing and working with partners to solve signal loss. Jason has worked in digital advertising for 15+ years, spanning roles from operations and product to strategy and partnerships.
Alex Johnston, Principal Product Manager, Yieldmo
Alex Johnston is the Principal Product Manager at Yieldmo, overseeing the Machine Learning and Optimization products. Before joining Yieldmo, Alex spent 13 years at Google, where he led the Reach & Audience Planning and Measurement products, overseeing a 10X increase in revenue. During his time, he launched numerous ad products, including YouTube’s Google Preferred offering. To learn more about Yieldmo, visit www.Yieldmo.com.
Latest posts
Data may not be the most glamorous aspect of marketing but it is at the heart creating meaningful consumer interactions in today’s data-driven world. In our award-winning, annual Digital Marketer Report we asked senior leaders about their top challenges and priorities. They said things like standing out against competitors, creating relevant interactions and customer acquisition. To my surprise, they didn’t say data. However, all the top challenges and priorities are predicated on having accurate, enriched data that is linked together in a central location for a complete customer view. The sobering fact is that most brands are not there yet. Most are not fully utilizing their data assets and maximizing their marketing intelligence. Take a look at these stats from the Digital Marketer Report to get an idea of where most brands are with data management practices. Overall, 45% of companies collect data via mobile – be it a mobile website, app or both TWEET THIS! 97% of companies suffer from common data errors. 61% of companies named human error as a top reason for data inaccuracies. TWEET THIS! On average globally, companies believe that 23% of their budget is wasted annually due to poor data quality. TWEET THIS! Today, only 35% of companies manage their data centrally through a single director. TWEET THIS! 99% of companies believe achieving a single customer view is important to their business. Only 24% of companies say they have a single customer view today. TWEET THIS! 29% of marketers who enrich their data with third-party data only add one type of data. TWEET THIS! One-quarter of marketers don’t enrich their data with any kind of third-party data. TWEET THIS! It’s important for marketing leaders to understand that they first need to focus on data and creating a customer-centric organization to support good data-management practices. Only then will they be able to reach their goals.
To say that Amazon Prime Day was a raging success is a considerable understatement Prime Day, the manufactured holiday by Amazon.com to mark the site’s 20th anniversary on July 15, 2015, was the biggest day of the past year for Amazon.com…by far. Experian Marketing Services’ Hitwise® online intelligence tool reports that Amazon.com garnered over 83.3 million visits from mobile and desktop browsers on Prime Day. By comparison, Prime Day topped Cyber Monday — the previous record holder — by a staggering 51.5 percent and Black Friday by 77.2 percent. It’s actually a bit ironic that Amazon promoted Prime Day as having more deals that Black Friday. That’s because visits to Amazon.com have actually been growing so steadily that last year’s Black Friday isn’t even among the site’s 10 biggest traffic days of the last 12 months. Nearly half of the days in July this year, in fact, have had visit counts that surpassed Black Friday. Compared to the previous Wednesday (July 8), the number of visits to Amazon.com on Prime Day rose 68 percent. But Amazon wasn’t the only retailer celebrating. Prime Day was definitely more of a “a rising tide lifts all boats” kind of event. For instance, 57 of the top 100 retail sites in the Hitwise 500 also saw increased visits on July 15 compared to the Wednesday prior. Walmart and Best Buy, which offered competing deals, had a an especially strong showing on Prime Day with visits to their sites up 97 percent and 44 percent, respectively, over the previous Wednesday. Search on Amazon Prime Day So what deals were consumers flocking to Amazon.com to take advantage of? To find out, I used Hitwise to compare the search terms leading to Amazon on Prime Day compared to the previous Wednesday. No surprise, many of the products that had the greatest increase in search click share week-over-week were Amazon’s own, including the “Amazon Fire Stick” (+417 percent) and “echo” (+357 percent). But also on the list were other products that were part of Amazon’s celebratory deals, such as “Nexus 6” (+352 percent), “Instant Pot” (+271 percent) and “PS4” (+250 percent). Overall, searches for Amazon Prime Day and variations thereof were common on July 15 and Amazon.com received 59.8 percent of search clicks from “Amazon Prime Day” searches, 61.2 percent of which were paid, meaning Amazon invested heavily in making sure that it successfully captured the traffic of interested consumers. But obviously not all Prime Day searches lead to the retail giant. Media sites Wired, CNNMoney.com and NBC News each captured at least three percent of traffic following a Prime Day search. In fact, seven of the top 10 winners of “Amazon Prime Day” searches were news and media outlets, which largely did not invest in paid search for that term. Walmart, which captured 1.5 percent of “Amazon Prime Day” searches, however, did pay. Among the “Amazon Prime Day” search clicks that lead to Walmart.com, 63.6 percent were paid. Of course two can play at that game. Amazon was also busy buying search terms of competitors on Prime Day. For instance, 23.9 percent of the clicks to Amazon from searches for “Target” were paid as were 15.9 percent of “Newegg” search clicks, eight percent of search clicks from “Best Buy” searches and 2.5 percent of search clicks for “Walmart.” Sources of traffic In terms of sources of traffic to Amazon.com, News and Media sites accounted for a 56 percent greater share of upstream traffic on Prime Day (6.7 percent) than the previous Wednesday (4.3 percent). But media weren’t the only ones talking about Prime Day, consumers, too, were taking to social media to chat up the event. And while not all online chatter was positive, Social Media sites delivered 15.2 percent of all referred traffic to Amazon.com on Prime Day, up from 11.3 percent of traffic the Wednesday prior, a relative increase of 35 percent. Despite the fact that Prime Day was meant to celebrate a milestone anniversary for Amazon, the overwhelming success for Amazon and other retailers will likely make it an annual event. If that is indeed the case, consumers will come to anticipate the day much like they do Black Friday. Marketers shouldn’t discount the potential for an annual Prime Day event to disrupt normal consumer spending patterns as well as drive even more dollars to be spent through online channels. For more information on how you can leverage Hitwise, the world’s largest sample of online consumer behavior, to improve the effectiveness of your search, display, affiliate, mobile, email and social media marketing campaigns, click here.
Pat Pellegrini is General Manager for Experian Marketing Services’ Consumer Insights group in North America. He also serves as Chief Research Officer and strongly supports the important role that high quality measurement science plays in driving deeper, more actionable consumer understanding. Every year, the American Association for Public Opinion Research (AAPOR) holds its annual conference where research professionals gather to discuss the latest trends, innovations and research. As a leading provider of consumer insights, Experian Marketing Services has an important role to play within AAPOR, the top professional organization of public opinion and survey research professionals. Since joining Experian as Chief Research Officer, and now as General Manager of Consumer Insights, I have driven our organization — already known for our trusted consumer insights — to move even more aggressively on innovations in measurement science, whether that is through self-reported consumer surveys like our Simmons National Consumer Study or passive data collection like our Hitwise online intelligence data. Consumers are being profoundly affected by rapid changes in technology and we are committed to being at the forefront in developing new research approaches to ensure that our data accurately and reliably reflects consumer decisions, behaviors, attitudes, lifestyles and media preferences. Experian Marketing Services was well represented at this year’s AAPOR conference both among attendees and presenters. In fact, six exciting topics from Experian were presented at the conference based on the work we do here every single day. Having this level of recognition from our research peers and the AAPOR organization is an outstanding accomplishment that speaks to our commitment toward research quality and innovation. It’s these types of insights that will help propel the industry forward. At the conference, I was delighted to present, along with my colleague Gerry Dirksz, findings from a successful initiative that Experian Marketing Services recently undertook to produce higher response rates to our Simmons National Consumer Study. Declining response rates are an obstacle facing nearly every player in our field and one that has real implications on data quality. Because certain key segments of the population (e.g., Millennials, Hispanics, etc.) are disproportionately less likely to respond to or participate in research studies, marketers and others who rely on such research may be handicapped by potential blind spots in the resulting data which prevent them from fully understanding important consumer audiences. To prevent such blind spots from occurring in our data, we tested a new two-phased dynamic incentive program that took into consideration an individual household or specific respondent’s propensity to fall into certain “hard-to-reach” segments. With this information in hand, we were able to offer different (often higher) incentives to “hard-to-reach” respondents from the start in order to improve response rates of those segments. Compared to control groups, response rates among those whose incentives were determined by the dynamic model were significantly higher. Having now rolled out this successful process for our entire sample frame, we have seen overall response rates improve markedly. While Experian is not alone in testing new initiatives like these, we are most definitely pioneers in the space and are setting an example for others to follow now that we have demonstrated results. In this regard, I look forward to blazing new trails based on well-thought hypotheses from our expert team. This will ensure that our data continues to be the gold-standard for helping our clients better understand consumers so that they can create more relevant and engaging interactions. Other topics presented by Experian at AAPOR include: The effects of total navigational burden, length of instrument and page complexity on item non-response This paper examined item non-response for long surveys which is a major topic of concern for survey researchers, particularly those looking to collect a large number of measurements from a single source. Specifically, binary (yes/no) questions were studied and potential factors that might contribute to item non-response were identified for further exploration. Patterns of response and non-response to sexual orientation measures In this presentation, relationships and patterns of item non-response among non-Hispanics to a measure of sexual orientation contained within the Simmons National Consumer Survey were explored. Since 2009, item non-response to the sexual orientation question has declined by a rate of 26 percent with a decline among 18-34 year-old respondents of 43 percent. U.S. Hispanic receptivity to self-reported measure of sexual orientation This paper examined the effects of adding a question on sexual orientation with respect to mail survey instrument return rates and item non-response rates among the U.S. Hispanic population. A key finding from this research was the pronounced, significant difference in item non-response rates which were disproportionately higher for Hispanics compared to non-Hispanics and significantly higher for Spanish-language Hispanics versus English-language Hispanics. Does providing an email address in an initial contact study indicate respondents will be cooperators in a subsequent online panel study? This study examines how the cooperation rate for a future study differs between individuals who are willing to provide an email address in the initial phase and those who are not. Results indicate that respondents who provide an email address in the initial contact study are more cooperative in joining a subsequent online panel than those who do not provide an email address. Patterns of non-response to health, diet, nutrition and apparel measures conditioned on body mass index This study examines the relative response of overweight or obese individuals to questions concerning diet, nutrition, health and even apparel in comparison to those individuals classified in the normal range for BMI. Based on our findings, a BMI measurement, outside of the normal range, would not negatively impact a study’s non-response rate to other survey questions related to one’s self-image. For more information about these presentations, please contact us at consumerinsight@experian.com.