Matching Ads with Emotional Context to Programming. An AI Approach with Wurl

Wurl, a connected TV adtech company owned by marketing platform AppLovin, has embarked on an interesting approach to ad monetization. Partnering with BrandDiscovery with its GenAI-powered contextual targeting, Wurl is said to be able to, “precisely match ads with the emotion and context of programming in real time,” according to Wurl’s VP of Business Development for Agencies and Brands, Peter Crofut.

He explained that by aligning an ad’s emotion to the content’s emotion, brands can improve viewer memorability and campaign engagement which he says, “ultimately drives stronger outcomes and performance. Our own data has shown that advertisers can experience a significant uplift in conversion of 2-3x when they find moments of emotional resonance.”

Charlene Weisler: How does the AI matching component work?

Peter Crofut: Wurl uses GenAI to analyze the emotion of the content in the scene closest to the ad break. That analysis is based on Robert Plutchik’s wheel of emotions which includes eight primary emotions at three different intensities (high, medium, and low) that can be combined to represent the full spectrum of human emotional states. Once the emotional score for the scene is determined, it is then matched to an ad with a similar emotional score. This prevents viewers from having a jarring experience where, say, they’re crying from watching a sad scene and are then disrupted with an overly upbeat ad. When scenes and ads are misaligned that’s where negative attention can be created for a brand.

Weisler: What data is collected – how is the platform measured?

Crofut: To determine the emotional segments of the content for BrandDiscovery, we leverage GenAI to view and analyze content from a variety of different sources including images, sound, and text. From there, BrandDiscovery assigns, at a scene level, the corresponding emotional value as determined by the AI. Notably with GenAI, we offer scene-level targeting through BrandDiscovery as opposed to categorizing an entire program, series or channel under any one contextual category. Equally, BrandDiscovery’s targeting is free – we don’t charge any data fees for emotional segments. In terms of measurement, we work with independent partners like EDO and Kantar to look at things like campaign effectiveness, brand awareness, brand favorability, and purchase intent for BrandDiscovery campaigns.

Weisler: How does scene-level analysis work?

Crofut: BrandDiscovery was born out of a curiosity to consider whether or not ads that were emotionally aligned with content perform better. During initial testing, we discovered that the most significant correlation with ad engagement took place when the emotions depicted in the scene just before the ad break aligned with those portrayed in the ad creative. As a result, Wurl developed its own GenAI models trained against millions of assets to analyze content in real time. As noted earlier, we use Plutchik’s wheel of emotions to determine the emotional score of the scene, and then match it with an ad that scored similarly. This analysis is calculated in real time across FAST (free ad-supported streaming TV) channels just before the ad break for highest relevance.

Weisler: Can you give an example of what a campaign looks like?

Crofut: We recently partnered with Media.Monks, who was tasked with executing a CTV campaign for a financial services client to expand its audience, improve awareness, and ultimately boost purchase intent. Using BrandDiscovery’s emotion-based targeting to run CTV ads across Wurl’s network of premium content publishers, Media.Monks was able to bolster the resonance of their client’s brand messaging and optimize targeting for more impactful brand engagements. The client saw a 7x lift in aided brand awareness and 2x lift in purchase intent compared to five-year industry benchmarks for Lending brand impact studies.

Weisler: What are the opportunities here?

Crofut: CTV offers massive potential for brands to connect with audiences in meaningful ways, especially as the biggest screen in the home. And, while contextual targeting on CTV certainly isn’t a new concept, we’re confident in the idea that emotions-based targeting is truly reshaping how we think about this approach today. By aligning ads based on the emotional context of the scene right before the ad break, brands can meet audiences where they already are and not just increase attention, but increase positive attention among viewers.

Weisler: What are the challenges?

Crofut: Media buyers today face a fragmented CTV ecosystem with multiple walled gardens and no standardized metadata – one movie or TV show could be categorized as multiple different genres depending on where you look. And, while CTV offers a significant step above linear in that it offers the ability to control who sees an ad, it still lacks control over the placement of the ad. All of this has resulted in limited transparency into where ads are displayed and in what contexts, ultimately contributing to the fact that advertising dollars have not kept pace with the shift in viewership from linear to CTV. By aligning ads’ emotions with the emotions of the programming, brands can have more confidence that their ads will show up in the appropriate context and better resonate with audiences.

Weisler: What are your next steps?

Crofut: In advertising, we hear a lot of talk about finding the right user at the right time with the right ad – though, in reality, we don’t really know what “right” is. As with all AI-driven innovation, the technology only gets stronger over time as it learns from more inputs. Right now, we’re in the early stages of seeing brands utilize emotions-based contextual targeting on CTV – as more advertisers adopt this approach, we’ll have even greater knowledge to learn and grow from as an industry and help us define what the “right” context truly is.

 

 This article first appeared in Mediapost.com 

Artwork by Charlene Weisler

Next Gen Audience Based Targeting Comes of Age Activating Emotionally Attached Audiences Changes the Playing Field. Interview with Gary Reisman.

The media industry’s transition into more data driven solutions enables us to get closer to a highly valuable consumer target.  Data ubiquity and modeling capabilities have definitely changed the landscape.  That said, Gary Reisman believes a new approach that adds consumer Emotional Attachment into the process can significantly improve Audience Based targeting – both in the quality of the audiences targeted and ROI resulting from media investments toward those audiences.

Reisman’s company, LEAP Media Investments, is a new type of Audience Development Company.  The company has conceived a way to activate audiences who are intrinsically connected with specific brands and who can be activated in online, social and mobile media and, in advanced-TV planning and buying.

 A patented process, 10 years in the making, LEAP develops “right-brain” data that quantifies the level of emotional attachment a consumer has to a brand, service or product.  The data, when modeled with big-data sets creates scaled, highly responsive audiences that enable more focused upper-funnel brand consideration and purchase.

“It’s widely known that over 75% of our purchase decisions are made based on emotion – meaning the bonds and relationships we have with brands and products yields relevance and meaning that drive us to engage and purchase brands and products.  Today’s demo, behavioral and transactional data simply do not develop audiences “connected” to the brand” Reisman explains. The next generation of audience-based targeting must also add emotion.

Reisman has dedicated his efforts to activating emotional attachment in media. “As a marketer, I saw a huge gap between the brand insight used for strategic positioning and creative development and the less insightful data used for media activation,” he explained. “For example, Walmart may know what drives attachment to their consumers. But when they then go to plan and buy media, they are stuck using ‘left-brain’ behavior and transaction data with some demo overlays  – attributes that do not leverage the brand equity and connections they have with their consumers,” he concluded.

Charlene Weisler: What do you mean by Emotional Attachment?

Gary Reisman: We define emotional attachment as the quantified value of the visceral brand connection a person has to a particular brand. Essentially it is their unwillingness to give up the brand. When quantified, this attachment translates to a person’s level of desire to engage, use or buy the brand or product and it correlates or equates to their brand share, revenue contribution and profitability. It is also a quotient that defines a consumer’s lifetime value to a brand or product.

Charlene Weisler: Why is emotional attachment so important?

Gary Reisman: It’s not just important, it is essential.  We have found that building audiences with increased levels of emotional attachment provide many benefits. Based on case studies, we have found that activating media based on LEAP’s emotional attachment leads to increased ad attentiveness/engagement (consumers are +50% more attentive to the ad message), increased brand advocacy, sharing and potential viral activity (consumers are +43% more likely to share brand messaging across their social channels) plus an increases in sales (consumers are up to 2.5 times more likely to buy) and increased media ROI. 

Charlene Weisler: How does LEAP develop their Audiences?

Gary Reisman: LEAP uses a patented data collection technique that tags thousands of individuals with the level of emotional attachment they have to hundreds of brands and media properties.  Through proper data sharing techniques, tagged individuals are matched with big data houses (DMP’s/DSP’s), then modeled and scaled for insightful strategic planning and audience activation

Our data creates three unique audiences that brands can target – Brand Enthusiasts Audiences – who closely represent those most attached to the brand; Brand Conquest Audiences, those who are moderately attached to your brand or competitive brands, and Brand Expansion Audiences who are the least attached and who offer the least amount of return.

We advise advertisers to reallocate their investment dollars to their two most emotionally attached groups – the Enthusiasts and Conquests.  When this is done, we have seen media ROI increases can be as high as 50-75%. Brands we have directly worked with have seen significant year on year sales growth.

We are also enabling a fusion with Nielsen data so the results can be stewarded through sales contract systems. Our data can also be incorporated into DMPs for planning and DSPs for advanced television online, social and mobile targeting.

Charlene Weisler: So give me an example.

Gary Reisman: Every brand has its own business plan and strategic needs. So a brand like Pepsi can develop and activate media that targets LEAP’s Pepsi Enthusiasts to drive low-hanging fruit sales. Alternatively they can plan and activate the LEAP Pepsi Conquests Audiences to gain more revenue from Pepsi consumers that are in need of a marketing push or Conquest against competitor’s LEAP Conquest audiences to steal share. Of course Pepsi is just one example; LEAP has developed audiences for over 400 brands across 25 categories

Charlene Weisler: Where does LEAP fit in the world of big data?

Gary Reisman: There is simply too much data and it is creating complexity, chaos and additional expense for brands. Also, everyone generally has the same data and is doing the same thing with data. To have a competitive advantage, you need a new approach; a new, more strategic, way to derive enhanced value and insight from that data and most importantly, activate true Audience-Based Targeting in today’s media marketplace.   Overlaying LEAP data provides a powerful, proven lens on the data – determining which data is more important that other data and modeling audiences predicted to engage with the brand.

This article first appeared in Media Village.

This is No Fleak. Q&A Interview with Jared Feldman

As the Founder and CEO of Canvs, a technology platform created to measure and interpret emotions, Jared Feldman understands and champions how emotional investment is valuable within the marketplace. New York City-based Canvs, which launched in public beta in April 2014 and officially into the marketplace in December 2014, has, according to Feldman, “become the industry standard for measuring audience reactions at scale through analysis of real-time Twitter data.” 

He got his start in the industry as a NYU student and while there met his now-partner Dr. Sam Hui. Upon graduation, the two formed Mashwork, which culled actionable insights for media clients. Canvs which grew out of Mashwork, has been experiencing great strides in recent months, including announcing their Series A funding and their recent partnership with Viacom. 

In this interview, Feldman talks about the impact of certain emotions on content success and by genre and the future of traditional research in an age of emotion tech.

Charlene Weisler: What social media data do you use in Canvs?

Jared Feldman:  Canvs receives its Twitter data from Nielsen, which captures relevant Tweets about original video programming from TV and over-the-top streaming providers for linear airtimes as well as on a 24-hour-a-day, seven-day-a-week basis. Analyzing those Twitter feeds from Nielsen, Canvs maps the emotional resonance of audience reactions online to first-run TV shows. Canvs then displays unique qualitative views showing the emotionality surrounding specific characters, plotlines and moments — mapping the emotionally charged reactions to 56 unique emotion categories including “love,” “dislike,” “annoying,” “beautiful,” “boring” and more. We just launched our own Facebook analysis capabilities where we summarized on an emotional basis all comments on video and the reaction to content. It is not a matter of measuring, for example, the number of tweets. We function to show what those tweets are all about – why emotions behind the social activity are so important. An emotional audience has greater recall and engagement. We can measure, by the use of emotional measurement, which ads were most engaging. 

Charlene: This is different from traditional research methodologies.

Jared: Yes. The only way to measure engagement and attention historically has been through focus groups. But it is time consuming and expensive and difficult to measure the emotional tenor of the audience.

Charlene: How do you measure an emotion like irony?

Jared: Irony falls into a bucket of emotions that require humans to review and measure. The methodology we use to measure emotions requires both science and judgement. You can’t rely on computers alone. But having a room of people judging doesn’t work either. We do not measure on sentiment – there are no positive or negative metrics. We have identified 56 common ways to express how people feel including love, shocked and crazy. But there is a lot of unstructured text, text that is not grammatically correct. Sometimes people do not use real words — words like ‘on fleak’ or ‘bae’ are commonly used by Millennials. These words are not picked or properly categorized by sentiment research. We built words and phrases that denote emotions. And we have recently included amplifiers and emojis. In the case of irony or sarcasm, a post like, “Man, I looove that show” with a thumbs down emoji can trip up algorithms. This is way we use a combination of efforts including machine learning with state-of-the-art language processing. We use humans to calibrate irony, sarcasm, jargon and misspelling, etc.

Charlene: Collecting audiences by emotions instead of demo groups is a departure from standard media measurement. 

Jared: Yes. We organize the world by common feeling – if we feel the same, we are the same beyond the demographics. How we feel affects what we do.

Charlene: What are the most predictive or important emotions?

Jared: It varies by genre. Any emotion in comedy means that the audience will come back next week. In Reality shows, it is ‘excitement.’ Our methodology was used by MAGNA to predict programs that would be renewed or picked up for a full season. We had an 85% accuracy rate in predicting shows that would be renewed or picked up. This is extremely exciting to us because it means that the reactions on Canvs are representative of the overall audience. 

Charlene: You mentioned the emotion “crazy”. What is that?

Jared: It is part of the colorful language that exists on social media and is neither positive nor negative. It includes exasperation and implicit excitement – OMG or That Is So Nuts. What’s interesting is that within drama programs, ‘crazy’ and ‘love’ are the two emotions that had the strongest correlation within loyal Tweeting behaviors. More specifically, a 10% increase in the share of ‘crazy’ among the total numbers of reactions which could reflect a jump from low to high ‘crazy’ for a  given program. It’s important to understand why this is. In scripted drama, “crazy” is an indication of ratings because there is a heightened sense of investment. It is a strong indicator for us. 

Charlene: Does emotional measurement replace the standard demographic measures?

Jared: There is still a place for demographics, but demographics as we know them will evolve. They’ll become more actionable over time. For example, rather than prioritizing where a person lives or what income bracket they’re in, researchers will ask: What do they find funny or crazy? Aligning viewers to core emotions will take precedence.

Charlene: Will a service like Canvs replace traditional research?

Jared: Traditional research will continue to bleed but will never go away. Soliciting responses will always have a place. However, we’ve always been bound by what people are willing to speak about and speak freely about. That’s one of the main reasons why traditional research can’t continue to compete with a qualitative technology platform like Canvs. Because, we’re capturing how people feel when they’re less inclined to be as guarded as they would in focus groups. If there’s one thing that’s clear, it’s that innovation has never been more important. Social media will become more of a social science. It has never been actionable before on the qualitative side but as deeper and more sophisticated technologies in this space continue to evolve Canvs will definitely be at the forefront of this space.

This article first appeared in www.MediaBizBloggers.com

Q&A with Tal Schwartz, CEO of ClickTale

Tal Schwartz, founder of ClickTale, started his data oriented career at an investment fund. He explains, “In a sense I was a scientist of financial data, because I was building models and trying to understand how financial assets are related to one another, and predicting their future movement based on historical patterns.” 

Schwartz put his data analysis background into his work at Click Tale as CEO. He says, “Today, people have become faceless data streams. When you go online and do something, you’re not treated or observed as a real human being. We enable businesses and marketers to understand what people are seeking, what their needs and desires are, and what their intent is.”

In this interview Schwartz talks about how ClickTale tracks the consumer experience, data, privacy without opt-in, measuring ROI, the impact of connected TVs and how the media landscape will look in the next three to five years.

CW: How does ClickTale measure intent? 

TS: ClickTale uses unique session replay technology, precision heat maps and customer experience consultants who work to help fix friction points on our customers’ websites. We gain insights into online customer behavior through various methods. One of these is anonymous session recordings of website visitors, which we are then able to show to our customers so they can see exactly where and why their visitors are experiencing frustration. These session replays are videos that depict exactly how anonymous visitors are experiencing a website -- they can see what their visitors do from the moment they enter the site until the moment they leave. The goal is not to see what individual customers are doing, but rather to understand why many customers are experiencing the same problem or having the same complaint. Another way is through heatmaps, which show an aggregate view - the most visited points on each page as well as the least visited. This enables them to see, for instance, that a specific item they’ve introduced is not being used due to its placement below the fold, where users are less likely to see it. Customer experience consultants offer another layer of assistance, as we have experts to give key recommendations that improve the users’ experience. We also have a web psychologist on our team, who contributes a deeper level of consumer behavior analysis as a way to provide more insight and help our customers optimize the customer journey for their users.

CW: What data do you use?

TS: We do not collect any personally identifiable information but we record billions of in-screen behaviors, such as mouse moves, mouse clicks, hovers (for desktops); touches, tilts, zooms (for mobile users); as well as a host of other gestures, no matter the device (desktop, tablet, mobile).

CW: What do each of these gestures mean – how do you parse out the insights from this type of data?

TS: Each gesture provides a unique view of the data. Being able to look at each gesture individually and then aggregating all the data in several different types of heatmaps enables us to drill down into the root cause of any challenge users are facing on a website. For example, looking at mouse clicks on a particular element: comparing the total number of clicks with the number of unique clicks, we can understand if there’s an issue with a broken link or loading times and so on. If we see many clicks on a single call-to-action button by specific users, we can understand where they’re coming from (geographically), which browser they were using, and then uncover the exact group of visitors who may have been impacted by the problem.

CW: How do you access the collection of the data without getting to the individual level? 

TS: The aggregation is done through any one of our high-fidelity, data-rich desktop or mobile heatmaps, which enable our customers to see the precise way users interact with the page and compare segments side-by-side. To view the collective behavior of desktop users, our mouse move, mouse click, attention, and scroll reach heatmaps help to show which areas on the page are the most interesting and engaging: where users are looking, which promotions most affect their behavior, and how their entire journey takes shape. The same is true to understand the behavior of mobile users – we have tap, attention and exposure heatmaps. For every heatmap, we provide a data-rich, graphical overlay with link analytics to display how every link on the page is performing. For example, you can understand if a particular link is being noticed enough, or intuitive enough.

CW: Is this opt-in?

TS: No. However we block the recording and collection of any Personally Identifiable Information (PII) entered by keystroke, as well as any PII as defined by our customers. We take a number of measures to ensure we never record, transmit, save, or display such information. To elaborate on this point, we only keep track of when keys are clicked, but not which keys are clicked. As a failsafe, even if any PII unintentionally reaches ClickTale’s servers, it is removed by the server-side rewrite rules before it can be stored. Furthermore, we use only 1st party cookies to guarantee the anonymity of site visitors.

CW: Give me an example of how you use the data to gain insights.

TS: A great example is looking at a search bar. There is a lot of debate in the industry as to whether customers using the search function are having a good or a bad experience. We use the data to answer this question, per company. For example, if the time to click on the search button is 40 seconds or more, we’ll drill down to the individual session recording to understand why a visitor couldn't find what he/she was looking for. But if time to click on the search button is 5 seconds, this might be the way the visitor prefers to use the site, which means this is not a bad experience. In addition, we can look at the time it takes a visitor who hovers over the Search bar to actually click on it, as well as what percentage of hovering visitors ultimately click. Taking all of this together, we can answer the question of whether the search function is delivering a good user experience for that website.

CW: How do you prove ROI?

TS: I’ll give you a couple of examples.

Before Christmas, Walmart was gearing up for its busiest season and launched a new Gift and Toy Finder tool as a means to boost sales. But there were a few glitches that were troubling the team at Walmart, so they turned to ClickTale to help optimize the buying journey for their users. ClickTale’s heatmaps revealed that 20 percent of the visitors were not seeing the ‘Finders’ tool. Furthermore, some of the visitors weren’t intuitively clicking on the call-to-action ‘Go’ button, and were getting confused when they didn’t receiving search results; both findings were revealed by watching ClickTale’s session replays. Visitors faced other points of friction including a JavaScript error when clicking on the call-to-action ‘Go’ button. As a result of fixing these errors, Walmart experienced an increase of 24 percent in their sales while improving their customer service.

Another example is Lenovo – when Lenovo launched its Yoga laptop last year, conversions weren’t what the company expected. Our session replays showed Lenovo exactly how customers were moving through the site, and our heatmaps aggregated the mouse movements of thousands of visitors. We discovered that certain browser types weren’t rendering their web site experience well, making it difficult to find the buy button on some web browsers. Lenovo made the required adjustments, and conversions went up by a factor of three to four times in terms of sales.

CW: Will you be getting into the area of connected TVs for your business and if so how will you get to the individual usage?

TS: Our goal is to reveal insights into consumers’ true digital experience on every device and platform which reaches significant market adoption. If it's technologically feasible and makes business sense, we will show the experience on Connected TVs. Right now, we have not seen much demand for insights into the cTV experiences, but we are keeping close tabs on the market.

CW:  Where do you see the media landscape going in the next 3-5 years?

TS: We see the era of customer engagement with automatically personalized website experiences taking off within the coming few years. Most of the big enterprises we speak with admit that they aren’t really there yet today but are actively working towards it. However, their goal here is more than the simplistic and often clumsy website or ad personalization that we see today, where visitors are retargeted based on product pages they’ve visited, even if they just browsed a page for a split second. That kind of targeting will likely not be acceptable to consumers in years to come, and might even be characterized as ad spam. Instead, I think we’ll see analytics behind the scenes that evolves to measure actual customer engagement, behavior, and intent, enabling better predictions on personalized experience. This will be perceived as a service rather than an interruption. It’s a win-win for the consumer and the brand. I call this new experience the "Predictive Web" and I believe it will become a reality in the next 3-5 years.

In other words, this means sites will adapt in real time based on consumers’ real-time behavior, predicting what a visitor is looking for and helping them achieve their goal with much less effort. It will feel a bit like a magical browsing experience, in which consumers both enjoy the browsing experience more, and get more done in less time.

This article first appeared in www.Mediapost.com