AI at Kantar

Using AI and machine learning, Kantar continues to forge new innovative paths helping advertisers drive stronger marketing impact through creative.
It tests creative assets quickly, iteratively and at scale, decreasing time to market and increasing ROI for our clients. Link AI on Kantar Marketplace is the fastest, fully automated, AI-powered solution to guide creative and media optimisation available today.  It is built on a database of over 250,000 ads and 35 million human interactions.

Find out more with LinkAI

Get highly scalable AI-powered cookie-less unified measurement and understand the relative contribution of all marketing mix elements. UMMO (Unified Marketing Measurement and Optimisation) puts you in control of your marketing budget and calculate your unified marketing (ROI) and their short and long-term impact on sales and brand equity. 

Find out more with UMMO

NeedScope has helped drive brand growth in more than 15,000 studies across 115 markets.

Understand the functional, identity and emotional needs in your market and how best to access them, using Kantar's validated psychological framework and AI capability. Analyse a brand’s imagery and video to understand the degree to which they are in alignment with the brand’s targeted emotive positioning across touchpoints.

Find out more with NeedscopeAI

Kantar BrandNow embeds Kantar’s proprietary AI technology Trend AI, to removes noise from survey data, making data lag-free and truly real-time. Understand the key drivers propelling your brand growth, so you can invest your resources in the right place at the right time, with the confidence of Kantar's globally accredited brand equity framework.  

Find out more with Kantar BrandNow

Generative AI is a new and exciting type of artificial intelligence system capable of generating new content such as text, images or videos. Generative AI systems are based on "foundation models", that are capable of ingesting text, images, audio or videos and generating data of any or all of these types, in response to human prompting. Large Language Models (LLMs) are a special class of foundation models that work with text data.

What makes foundation models very powerful is that they are "task agnostic", i.e., they can be highly proficient on seemingly unrelated tasks that they were not explicitly trained for. For example, LLMs are typically only trained to predict the next word in a sequence of words. However, this training when done at a massive scale, apparently allows the models to write poetry, do math, write code, solve puzzles, excel in exams and much much more.

Experts agree that we are far from arriving at Artificial General Intelligence – the time when a machine will be able to understand or learn intellectual tasks as a human would. However, foundation models could be a possible first step.

Both Traditional AI and Generative AI are based on Machine Learning - algorithms that enable machines to learn from data without being explicitly programmed. However they differ in important ways.

Traditional AI models are mostly focused on advanced analytic tasks such as prediction or clustering. They typically work with just one form of data and typically are trained for one specific task that they become very proficient at. They need to be trained from scratch or "fine-tuned" to perform well on a different task. Generative AI models on the other hand are based on foundation models that can ingest data of various types (text, image, video) and output any/all of these types. Generative AI models require orders of magnitude more data than traditional AI models but when provided with such data are more versatile - they can be good at activities they were not explicitly trained for.

Neither form of AI is necessarily "better" or "worse" than the other. Each has its own benefits depending on the intended use case and in fact in many practical situations, the best solution involves a combination of both. For example, Link AI uses both Generative AI and Traditional AI as part of the same overall framework to issue its predictions.

There are quite a few known risks that we see in the market research industry. One is that the model starts making things up that are factually incorrect or fantastical - a phenomenon referred to as "model hallucination". A second is that the models might not be entirely up to date - we’ve seen this with time series examples where the previous version of ChatGPT gave wrong answers because it was only updated to 2021. More fundamentally, LLMs have no ‘knowledge of knowledge’, so there’s no such thing as a confidence level. And LLMs have no notion of time or temporality, or maths, which is rule-based, so they are currently limited in their interpretation of data to what they can discern through generic associations.

Aside from technical issues, there are also potential legal and ethical issues that arise. Intellectual property, for example: is this a creative act by the LLM, or is it re-hashing someone else’s IP? Does sharing your own data on the open web mean you give permission for it to be used by LLMs? And finally, the quality of the datasets the models use could easily reinforce biases and stereotypes without ‘knowing’.

Use cases for Generative AI fall into 3 broad categories:

1. Making things more efficient: Foundation models can free up a researcher to focus on what is important.

- Summarisation: market research collects a lot of data in the form of words – survey verbatims, qualitative interviews, and focus groups. LLMs could summarise, order and prioritise responses expediting the work of the researcher when creating a narrative for the client. LLMs and foundation models can even summarize videos and images!

- Automated reporting: market research also produces large volumes of quantitative data that need sorting, summarising, and presenting. LLMs could quickly organise and create draft headlines based on charts, tables, models, as well as executive summaries.

2. Doing things at higher quality: Foundation models can do what earlier AI models did at much higher quality, sometimes surpassing a human

- "Attribute" identification: LLMs can identify themes, assess sentiment, brand affinity, brand perceptions, identify emotions, fix for the researcher to refine.

- Prediction: Foundation models allow us to extract embeddings (mathematical representations) that other machine learning models can use to predict some outcome of interest. For instance, does the dialogue in a TV ad help predict its performance? How can we relate people’s qualitative experience interacting with a service representative to their brand loyalty or churn? The quality of such embeddings is often significantly superior to the previous generation of AI models

3. Doing entirely new things: Generative AI opens the door to many new applications

- Intelligent interviewing: already in use by the industry, conversational AI will come on in leaps and bounds, responding to previous answers and routing questions accordingly. And designing quant questionnaires will never be the same again, the machine can help with automating and standardising the process!

- Creative Writing: this could be anything from creating discussion guides, initial drafts of presentations, marketing copy, concept statements to video ads

- Conversational search queries: think of ’an intelligent agent’ that sits on top of data platforms you can query in natural human language. The agent then analyses potentially massive databases ’underneath’ and fetches back the results in natural language.