AI helps marketers maximise creative ROI at scale
AI has been the game changer of the decade in advertising and marketing, bringing both excitement and scrutiny. While its ability to streamline processes is widely appreciated, its effectiveness—especially in creative decision-making—has often been questioned. This applies not just to content creation but also to ad testing.
Smart marketers know that the right creative can drive higher ROI, and that validated ad testing provides the insights needed to optimise ads and select the best ones to air. AI-powered ad testing takes this a step further by delivering scale and speed, ensuring all ads can be evaluated, not just hero assets in major markets. But does this come at the cost of accuracy? How can marketers differentiate between AI solutions which are developed quickly based on a few simple rules, and those which use much more sophisticated models based on extensive training datasets?
Latest comprehensive LINK AI validations prove accuracy without compromising on speed
Kantar’s LINK AI provides a data-driven, efficient way to assess an ad’s potential—in just 15 minutes. AI prowess is determined by the quality and size of the training dataset it learned from and LINK AI builds on the success of Kantar’s extensively validated LINK database which has over 260k+ tests, and 35M human interactions across global markets. As a result, LINK AI delivers predictive insights that closely align with human-based survey results. The latest validations confirm its accuracy across various channels, markets, categories, and brands.
LINK AI has a rigorous modelling process which leverages the latest machine learning (ML) technologies in neural networks with state-of-the-art embedding methods for feature extraction. Our ML model is an ensemble of algorithms trained to bring together all these features to predict the creative effectiveness KPIs. This is proven through extensive use of modern validation methods, including hold-out validation on over 10,000 ads, 10-fold cross validation, and in-field validation for many clients across diverse categories, markets and brands.
LINK AI consistently predicts ad performance across key metrics. On average 8 out of 10 TV and digital ads that rank in the top 30% for branding, enjoyment, or persuasion are also predicted to be high performing by LINK AI for the given channel. This strong alignment with human-based survey results reinforces its accuracy in assessing creative effectiveness.
Latest LINK AI video validations, Global
Validation Definition: Percentile matches are the % of ads scoring within the same percentile range (accounting for survey margin of error). Matching is based on LINK+ standard ranges of 0-30, 30-70 and 70-100.
This strong alignment holds true across different markets as well. In the example of U.S., Brazil, and Germany, LINK AI consistently predicts top-performing TV ads with over 80% accuracy across key metrics. This global reliability reinforces its value as a powerful, scalable tool for marketers looking to optimise creative effectiveness with confidence.
Latest LINK AI validations, TV percentile match per market
We also see similarly strong match rates across a wide variety of very different categories ranging from food & drink, personal care and household products to automotive, retail and services.
Beyond LINK AI validations with LINK survey results, some clients also validate LINK AI through to other external effectiveness indicators. One FMCG client validated 224 ads for 10 brands in 11 markets using MMM (Market Mix Modelling) sales modelling. They found that the top third of ads on LINK AI’s Short-term Sales Likelihood metric delivered 15x more sales than the bottom third of ads.
Some common AI validation pitfalls
While most marketers are increasingly comfortable with the idea that AI creative predictions work well at scale, they become less certain whether it will work for their ad, in their market, in their unique situation. We sometimes see advertisers attempting to “validate” AI creative testing by running a handful of ads through AI testing and attempting to compare those scores with survey or in-market results. This is invariably a mistake. Attention in these exercises tends to focus disproportionately on the one or two ads out of 10 which matches less well, or the one or two metrics which match less well among this small sample of executions, rather than celebrating the 8 or 9 successes. We often see people failing to account for the margin of error, which is present in any form of survey testing, treating the survey results as gospel, rather the representative estimate they really are.
All of this does AI testing a disservice. AI predictions do not think in the same way humans do. While they can be trained on the outcomes from human observations and opinions, they are arriving at their effectiveness conclusions and predictions in a fundamentally different way. When there are differences between AI predictions and human survey sampling, this does not mean the AI predictions are automatically “wrong”. Far more likely is that the ultimate truth about the strength of the execution lies somewhere between the two data points.
AI can understand and assess creativity
Can AI decode human creativity and predict its impact accurately? The answer is, yes it can. And the best way to demonstrate this is via real-world examples.
In a recent partnership with EFFIES Europe, we ran 655 winning entrants through LINK AI. Even within this set of successful ads, LINK AI was clearly able to differentiate between good and great ads. The gold winners scored 9 percentile points higher than bronze winners on Impact and 12 points higher on our key long-term Power metric.
Average percentiles of Effie winning entries predicted by LINK AI
A standout example of gold-winning creativity is Magnum’s ‘Stick to the Original’ multi-market campaign. The ad hilariously tackles private-label competition, as a woman’s discovery of a fake Magnum spirals into a full-blown relationship crisis. LINK AI accurately predicted the ad’s border-defying success, which you can read more about in detail in Kantar and Effie Europe’s booklet, along with other award-winning creativity cases.
Not only can LINK AI identify award-winning ads, it can also avert potential disaster. In early 2024, Apple released ‘Crush!’ an ad that sparked huge criticism. It featured a massive hydraulic press compacting symbols of creativity and culture into the latest iPad. While the tagline 'The most powerful iPad ever is also the thinnest' was meant to highlight how much the device offers in a sleek design, many saw it as a metaphor for machines overpowering culture and creativity.
We tested the ad with LINK AI, and the results reflected the real-life reaction it received indicating underperformance. In hindsight, Apple's approach with this ad may not have landed as intended, limiting its impact despite the creative effort behind it. Interestingly, the foresight provided by LINK AI could have offered valuable insights to help assess the ad’s reception before launch, enabling a more informed decision.
Unlocking new ways to access valuable insights
Marketers now manage more creative assets than ever in their campaigns and AI enables marketers to explore new possibilities to access insights at scale. AI-powered testing provides insight across channels before committing resources. This is especially useful for testing digital across formats which have historically been tested less via survey approaches. The approach is not limited to finished films, LINK AI can be applied to animatic videos as well as early-stage ad summaries. LINK AI can also predict success for static images (digital display, outdoor and print ads).
With LINK AI’s ability to conduct meta-analyses at scale, brands can now also uncover broader creative trends and optimise their strategies more effectively. Creative tags on LINK AI identify key diagnostic features, ranging from tangible elements like 'brand presence' and 'ad pace' to more complex ideas like sustainability, provides powerful insights at scale. This facilitates large-scale meta-analyses that uncover brand-specific creative strategies and platform-specific trends.
Additionally, AI-powered testing makes it possible to assess ad performance across multiple markets, even before translation, helping marketers refine their approach early in the process. Want to know if your UK ad would perform just as well in Brazil or China? LINK AI has you covered with a market transference feature you can deploy before you invest in adaptations or media spend. This approach allows you to use survey testing in lead markets as seed data for even smarter predictions than is possible via standalone AI testing.
Harness the power of AI-powered testing to make smarter choices
AI-powered ad testing is transforming the way marketers make creative decisions, delivering speed, scale, and accuracy without compromising on effectiveness. LINK AI’s latest validations prove its reliability across key metrics, markets, categories and channels, helping brands optimise their ads with confidence. From predicting ad performance to enabling large-scale meta-analyses and market transference, AI is unlocking smarter, data-driven choices for advertisers. Discover how you can unleash the power of LINK AI for your brand, whatever channel, market or category.
