Preview Q&A


About PreView

What is PreView?

PreView is a SaaS system to measure human response to digital content at scale.

Preview is built on the most advanced facial coding technology and machine learning (ML) to provide marketers a true measure of consumer attention and emotional response so they can optimize advertising creativity and increase quality exposures.

Chief marketing officers, brand managers, media buyers, and creative strategists can use Realeyes PreView to identify and test the ads and creative elements that are the most compelling to consumers and deliver the best attention outcomes. PreView is purpose-built for speed and scale across large portfolios of video content globally.

Using front-facing cameras to analyze facial coding data of opt-in viewers as they watch video content online, Realeyes PreView allows advertisers to optimize their content and targeting.


What problems does PreView solve? For whom?

For advertisers, Realeyes reduces media waste and maximizes impact by revealing the attention-earning power of advertising creativity at both a unit and portfolio level.  The system provides predictive in-market performance metrics like second-by-second audience retention, the ability to capture and hold attention in the first seconds, and the presence of emotional peaks that lead to memorability. PreView combines these scores into an overall Quality Score, a power ranking that quickly reveals a video’s propensity to deliver attention outcomes. All scores are delivered in the context of competitive benchmarks and geographic norms.  PreView empowers CMOs to align creative and media functions within their organizations behind a single, critical KPI: predicted attention performance.

For publishers, Realeyes provides an objective, third-party measure of creative quality that is useful in incentivizing advertisers to create and submit higher-performing creative assets. Higher quality creative leads to better experiences for consumers, more attentive and loyal audiences for publishers, and better business outcomes for advertisers. PreView aligns publishers and advertisers around real outcomes (I.e., attention) necessary for advertising impact, and provides an ideal measurement alternative to brand-lift studies, which rarely have any correlation to real-world sales.


How much does PreView cost?

Preview is priced to incentivize scale and maximize ROI for clients through volume testing and the improvement of their advertising creatives’ Quality Scores. Volume packs and subscriptions, with cost-scaling efficiencies.

Example pricing range: A premium “Tier 1” account, which runs hundreds of millions of dollars annually on video advertising, could represent a $2M annual contract value within three years. With that opportunity, a 20-study PreView pilot would be proposed at $60K at the low end, or up to $120K – $240K for a larger pilot that includes a validation study and data science consulting. Upon pilot validation, we would target $300K – $600K for a first-year subscription. For a smaller “Tier 2” account, with a potential annual contract value between $500K and $1M, a more palatable path might be a 10-study pilot at $40K.

See Realeyes internal pricing strategy guide.


Can I apply PreView against specific demographics?

Yes. PreView can be deployed against any audience segment available within our partner network of research participant providers. Demographic segmentation within individual PreView studies is not available now, though this feature will be made available in the future.


What types of advertising use cases is PreView great for?

PreView is transformational for advertisers seeking to minimize media waste and maximize the impact of their video advertising creative across all varieties of digital formats and linear TV. That includes video advertising on online digital channels like Google/YouTube, Facebook, and TikTok; OTT services like Roku; and streaming video including subscription services like Netflix, Disney+ and Amazon Prime, and ad-supported ones like Hulu and Peacock. Traditional TV formats include over-the-air and basic and premium cable channels.

PreView is purpose-built for the most common standard video ad units across digital and conventional channels and formats – typically spots:60 or less, and even memes (i.e., animated eye-catching gif files).

Other use cases, such as static images and long-form content (i.e., movie trailers or branded/sponsored content vignettes), are measured with Realeyes Custom.


In What Markets Does PreView Work?

The preView was developed and launched in the U.S. and Europe, however, PreView is designed to work in most other markets (80+ countries) including in Japan where Realeyes has a key sales and client success office. If a client demands, we’re equipped to support specific validation and adapt to localization requirements. Additionally, PreView algorithms are always getting smarter thanks to machine learning, along with validation against in-market performance data provided by partners and clients.


Are PreView studies secure and confidential?

Realeyes upholds industry standards for ensuring client confidentiality, and that includes the PreView product. Clients frequently turn to Realeyes to test confidential messaging for major product launches, as well as for competitive conquest.


How are PreView analytics delivered?

The primary delivery vehicle for PreView data is a static report that includes a summary of Quality Scores included for each ad creative in a given study. The report also includes for each creative sub-scores that reveal the different components that drive a high or low-Quality Score. It also includes attentive viewing predictions and key scene moments that comprise the ad and should be considered for optimization.

All PreView reports include a .csv with all granular metrics so that you can import them into your own internal analytics and reporting systems.

Preview is built for better selection and optimization for small to very large groups of ad creatives, in order to improve attention outcomes and overall campaign impact. For situations where an individual ad creative requires deeper diagnosis, PreView includes our back-end dashboard with frame-by-frame attention and emotion tracer lines. The dashboard is primarily used internally by Realeyes staff, though your Customer Success representative can introduce it to you.


How is benchmarking applied to PreView? Can I select benchmarks?

In our attention-scarce word, competitive benchmarks are a critical component of PreView. Relative competitive performance influences the PreView’s Quality Score. Competitive benchmarks also are included against most other granular metrics in the report, so you can know where your creativity is strong or weak in the quest to earn attention and drive impact.

Benchmarks are industry and category-based and are applied according to your brand’s industry classification. In cases where benchmark sample sizes are insufficient, PreView defaults to the overall market benchmark in which a given study was executed. Realeyes can apply alternative industry benchmarks per client request.

There are not benchmarks specifically for animatics, though PreView is still useful with animatics for early readings into creative quality. PreView is designed to help you work toward better attention outcomes throughout the advertising cycle, from creative development, production, creative selection, and portfolio optimization.


Why does the PreView focus on the:15 mark?

Digital video platforms like Facebook and TikTok have surfaced an unspoken truth in advertising: it is difficult to hold viewer attention for very long. The granular and transparent viewing completion metrics of digital publishers have led to a common best practice: keep ads less than 15 seconds in duration. That is even a best practice bestowed by TikTok, one of the most popular and engaging video platforms in the world.


Does PreView include the ability to deploy survey questions?

PreView is a pure software-enabled facial coding measurement tool designed for managing better attention and advertising outcomes. The predictive analytics and diagnostic features enable scale, precision, and economics that are unprecedented in the advertising industry – attributes which surveys are incapable of achieving. For cases where a survey is required, Realeyes offers a survey solution as part of Realeyes Custom, with additional fees.


What happened to the old “Realeyes Score”? How does that compare to the “Quality Score”?

Realeyes has discontinued the Realeyes score, which was a composite score that heavily weighted various emotional characteristics of video creativity. Driven by client input and market validation, Realeyes, through PreView, has adopted the Quality Score for its ability to deliver more predictive and actionable analytics of large and varied advertising portfolios. PreView is designed for fast, precise decisioning of large ad portfolios, and provides an automated diagnosis of key elements of video ads that drive attention outcomes.


How does PreView generate its prediction algorithms?

PreView develops its predictive algorithms by decoding the relationship between 1) emotional and attentive scores in forced-exposure facial coding tests and 2) historical ad performance and other data provided by clients and partners.

For example, we discovered that attention and emotion scores in a forced-exposure video test can predict in-market video view-through completion rates with an 85% second-by-second correlation on average, with ads 30 seconds in duration or under. This prediction algorithm, initially based on over 200 ads across seven markets, surfaces as the “Audience Retention” sub-metric within PreView’s Quality Score. The chart below illustrates the average Realeyes predicted audience decay versus actual view-through rate data.



Additionally, the Attention Potential sub-score has predictive power to anticipate viewers’ opinion of a video creative under real-world conditions. Specifically, we found a connection between our Attention Potential sub-score and opinion-based survey performance indicators like “interesting versus boring.” A positive correlation of ~0.3 suggests that in the case of a substantial volume of creativity we have tested, a high Attention Potential (engaged, attentive audience) is associated with a large proportion of the audience finding the ad interesting, and vice-versa. A negative correlation suggests a link in the opposite direction: if people score high in Attention Potential to a video, they will be less likely to report it boring. But if their attention and emotional engagement decline, they will find the video boring.

Links With PreView’s Attention Potential Sub-Score

Survey Performance Indicators


“I find this ad interesting”


“I find this ad boring”


*All results are significant


Clients and partners who share ad performance data are eligible to receive key perks, including advanced custom analytics and early access to new Realeyes products and features. It also enables custom solutions such as prediction algorithms tailored to a specific brand.



What is the ROI of PreView?

Attention is an advertising outcome that precedes any effect. Realeyes developed a sophisticated ROI model using attention data from Realeyes’ PreView product, applied to an actual portfolio of video ads from consumer brands. Our case study scenario shows that optimizing creatives for attention within media activations can easily achieve 30% performance improvements (or waste reductions). It also can achieve a 10X ROI based on the cost of applying attention prediction data. Therefore, our conclusion is that to advertise effectively, an advertiser must embrace attention through reliable, valid, and quantitative measures that are comparable across media and time.

In every execution of the model thus far with real client data, we’ve been able to reveal significant opportunities to reduce media waste by managing the quality of creativity with PreView intelligence. The model has been reviewed by some of the most experienced advertising measurement leaders in the world, including Bill Harvey, CEO of Research Measurement Technologies (one of the fathers of modern media planning and advertising research); Ted McConnell, founder and former director of P&G’s marketing innovation center; Barry Fischer, former EVP of market strategy at NBC Universal and more.

We encourage you to download the PreView ROI white paper, “The High Price of Low-Quality Video Advertising.


PreView Methodology

How does PreView test video ads?

PreView tests video creative assets in a forced-exposure setting. After selecting target audience attributes, the PreView system recruits incentivized participants from our network of research panel providers. There are typically 150-350 included in a study, depending on client objectives. Participants then visit our consent website from any device, then immediately enter the study, where they will watch up to five ads in a session duration of five minutes or less – from anywhere on their own device, be it a phone, tablet, or laptop/PC. The participants’ own devices and front-facing cameras measure facial cues and send attention and emotion response data for each video frame back to Realeyes’ cloud servers for processing. The raw attention and emotion data then are processed through PreView’s prediction and intelligence algorithms and delivered to customers in the PreView report, along with granular metrics in an accompanying .csv file. The end-to-end cycle is 24 hours or less.


How does attention and emotion tracking work?

Long-standing academic research has established that there six basic emotions that are universally recognized by humans - happiness, surprise, anger, sadness, disgust, and fear. This phenomenon has evolved into a formally accepted branch of psychology called facial coding, pioneered by Dr. Paul Ekman in the 1970s and 1980s. The basic emotions are generally consistent in their expression, so we can accurately train computer algorithms to recognize them using computer vision and machine learning techniques.  In addition to universal emotions, Realeyes has pioneered the measurement of attention, which happens in parallel, second by second, frame by frame.

The computer is fed a series of training instances – sets of images representing each emotion. Using these data sets, the computer learns to associate the image with the emotion label and is thus able to assign these labels to unseen data - images it hasn't encountered before, such as new recordings of people watching video content. Realeyes has the most accurate emotion classification system, enabled by the largest culturally sensitive AI training set including 620 million emotion labels across 28,000+ videos.  Precise emotion detection powers marketing intelligence that drives better business analytics.



Viewers' responses are recorded through their webcams with their consent while they watch video content. These are streamed to our cloud servers, where they're securely processed - low-quality recordings are filtered out, facial expressions are analyzed using our algorithms, and the results are aggregated and reported on our online dashboard in near-real-time.



To see an attention and emotion tracking technology in action, download our Emotion Booth App here.

For full details on the ins-and-outs of the science behind our technology, download our emotion measurement whitepaper here.

Which audience research panel providers do you work with?

We work with the best-in-class panel companies to ensure worldwide access, high quality, and speedy deployment of PreView analytics. Popular panel partners include Cint, Toluna, Fulcrum/Lucid.

We can integrate nearly any panel partner instantly and are continuously growing our panel partnerships. If you have a preferred panel partner that you are interested in using, let us know.


What methodology is used for recruitment?

Participant recruitment and incentivization are fully handled by our panel partners. We purchase samples only from panel providers who adhere to ESOMAR standards for recruitment. Incentives may range from monetary compensation to gift cards and prizes. Sample sizes and composition are determined according to client objectives.

The full library of ESOMAR Codes & Guidelines can be found here.


How are Realaeyes data validated?

Foremost, the psychological framework that Realeyes applies in its technology and measurement approach has been validated through decades of academic and clinical research. The accuracy of our facial coding algorithms is always improving and validated through periodic testing. In fact, they are more accurate than all other major emotion detection software providers, according to a large-scale internal test. In fact, the accuracy is approaching that of psychologists and human annotators in video-viewing contexts.

Second, we validate the reliability and reveal trends within our core attention and emotion measures, benchmarks, and norms by conducting a large-scale, bi-weekly test of videos repeatedly in key markets. We apply the same selection of universally appealing videos over time to understand how humans’ emotive response to video content evolves. This technique and study duration also allow us to identify fluctuations cause by key events (i.e., a pandemic’s correlation with reduced frequency of positive response).



Does your technology compromise consumer’s privacy?

Realeyes is consumer friendly and privacy safe. The company, its products and services, and its suppliers abide by Europe’s GDPR and the United States’ COPPA regulations.

We measure participants only if we have a lawful legal basis to do so.
Importantly, Realeyes does not engage in facial recognition or identify recognition.

Have more questions? Submit a request


Please sign in to leave a comment.