Expert Speak: Supercharging Brand Marketing with AI

 Expert Speak: Supercharging Brand Marketing with AI

Anshuk Gandhi, VP of Product Development at Swayable, writes about impact of AI on brand marketing — how AI is enabling brand marketers with deeper insights, creative freedom, and effortless personalizationin this edition of the ‘Expert Speak’ column.

The last six months have been like the “Apple II” moment for AI. Large Language Models (LLMs) have turned AI from an esoteric, convoluted tool, only accessible to wizards at large tech companies (or speculated about by sci-fi novelists), into a friendly sidekick that even a ten-year-old can play with.

LLMs have an ability to understand context and generate long-form text and complex images. They’re not just pattern matching people with ads or predict click-throughs; they’re actually participating in the creative process! They’re giving brand marketers superpowers they only daydreamed about a few years ago, with the ability to speed-read consumer research, spark creativity, test concepts in a flash, and personalize assets like never before. This article explores how AI is transforming the way brand leaders and creative directors can leverage these technological advancements to supercharge their effectiveness.

1. Making consumer insights accessible and more insightful

Brand Marketing begins with understanding your audience. Traditionally, analysts would spend days poring over sizeable consumer research data to draw insights and present them in a digestible form to brand leadership. Follow-up questions, such as alternative demographic cuts, would trigger another cycle of analysis, taking several more hours or days. Digesting qualitative insights is even more challenging, often leading analysts to identify a few verbatims among thousands that confirm existing hypotheses.

AI is now bestowing superpowers on brand teams to put their arms around all of their data and extract insights in a matter of minutes. It’s like having your very own data analyst by your side. It’s making data analysis directly accessible to the brand teams. Marketers can now synthesize large volumes of information – structured and unstructured, and combine a variety of disconnected data sources. For most types of analysis, we’re no longer constrained by the availability of data analysts to parse through large quantitative research, or by a platoon of associates reading through thousands of customer reviews or trying to reformat spreadsheets. Today, a marketing associate can upload a hefty volume of quantitative research data (like a Swayable RCT ad test, often over 30K rows of data) into ChatGPT’s Code Interpreter, ask questions in plain English, and synthesize insights from it in an afternoon.

These insights are not limited to simple aggregations. Swayable’s Advanced Services team members who are domain experts in brand marketing, now use AI tools to create reports with advanced statistical analyses and whip up custom data visualizations for clients, tasks which used to require involvement of a data scientist.

Consumer research also gathers volumes of qualitative verbatims through surveys, focus groups, social posts, customer reviews, etc. We’re no longer limited to simply classifying qualitative data into ‘positive’ / ‘negative’ sentiment categories, or staring at vague word clouds. LLMs are really effective at deciphering meaning from informal verbatims, clustering them into semantic groups and drawing specific takeaways. Swayable is prototyping an innovative approach of marrying this qualitative data with quantitative research to produce uniquely powerful insights. More on this later in this article.

Also read: Expert Speak: The impact of AI on the beauty industry

2. A magical paintbrush for creativity in brand marketing

AI is equipping every individual with a magical paintbrush for creativity. Creative teams can experiment a lot more, test, and get feedback on many more concepts. It’s much easier to elicit constructive feedback from brand leaders and from consumers to a sketched out idea vs. just describing it. And now, producing these sketches is as easy as pie for anyone on the team. A creative designer can draw a sketch in Stable Doodle, add context and instructions in a prompt and generate a pretty good rendering within minutes instead of hours. They can bring their storylines to life by turning these images into scenes with Runway and stitching these concept scenes together with tools like Genmo. They can add speech from text without looking for a voiceover artist, and create background music for the concept using Beatoven

In my opinion, the most significant impact will be a) opening up the creative process beyond the creative team, and b) an explosion of creative ideas that turn into concepts. Anyone can now be a designer, at least for rough cut sketches, storyboarding and prototypes. One doesn’t need a design background to “render” ideas using Midjourney, Dall-E, Adobe Firefly, Canva AI and a host of other tools. Team members that lacked the tools to express their ideas can now participate more richly in the creative process. Teams will now have the time to explore far-fetched ideas that in the past they would deprioritize in the face of looming deadlines. They will take more risks and unlock new brand expressions. And AI can also provide a safety harness around this experimentation, by helping align copy and images with the brand voice and design guidelines. 

As James Slezak, CEO of Swayable puts it:

“If you’re going to make one thing and it takes a long time, you’re not going to take as many risks. But we’ve all had crazy ideas and wondered, ‘Would they fly? Would they lift consumer intent?’ Now there’s a chance to do that more easily.”

3. Rapid concept testing

The flip side of taking more risk and generating more concepts is to be able to test them for persuasive impact, and understand which concepts motivate people to think positively about the brand (brand lift), reinforce the right brand attributes, and drive purchase consideration and intent. One reason why teams don’t explore many concepts is that testing these concepts has historically been expensive and taken weeks. However, this is no longer the case: for example, Swayable’s RCT tests can measure persuasion impact accurately within 24 hours!

Kathryn Rathje, Partner at McKinsey & Company emphasizes the importance of this speed:

“Now that we have the ability to get information real-time, we can do a lot more pre-testing, creatives who embrace data and make it part of their DNA will be successful.”

AI is now bringing scale benefits to concept generation, by analyzing cumulative learnings from the entire history of creative tests for a given brand, also known as “meta-studies”. AI trained on test results for all prior concepts can identify features that lead to higher impact. And this is now possible even if the concepts are not exhaustively labeled and tagged. Multi-modal models (example: Twelve Labs) that combine video, audio, speech, images, text and numerical data sets are quickly becoming adept at recognizing not just objects from an image but the semantic concepts conveyed by the entirety of an image or video. They can now analyze hundreds of ads to decipher the impact of using a narrator with a southern accent, the complexity of the storyline, or the tempo of the music, the amount of product exposure, etc. on the persuasive impact of an ad. Swayable is already prototyping this technique to produce the next version of its groundbreaking meta-study of political campaign messaging.

AI also helps execute a combination of quant and qual research with hundreds of participants – which is more efficient and gathers much more useful information than traditional focus groups or simply listening to social posts or customer reviews. At Swayable we’ve developed an approach, in collaboration with Outset.ai, to combine quantitative survey response data with an AI chatbot – this chatbot is able to engage respondents in a substantive conversation and leverages the quantitative response data to add insightful emotional context that explain the “why” behind statistical RCT ad testing results.

Researchers in academia and at Swayable are also working on creating synthetic audiences to approximate how a real person would react to an ad or asset. We’re not there yet, but it’s within the horizon that we can train AI models to give responses that are fairly representative of a large population sample. That might make it possible to get a rough analysis of a concept video within minutes.

Also read: Recursion drug discovery tech gets new US$50M; partners NVIDIA

4. Personalized ads at scale

Marketers can no longer rely on a single ad film to motivate all of the audiences that a brand needs to connect with. Brands still have to tell a consistent story, but the interpretation of that story needs to be tailored to audience segment, delivery channel, and channel context. To execute a personalized campaign for a new product, even modest-sized brands often need hundreds of assets. This has always been a speed bump on the road to personalization.

As Stevie Archer, Executive Creative Director at SS&K noted recently at Cannes Lions:

“From a creative perspective, the challenge has always been the at-scale part, the more things you have to make for different people, the [more expensive and time consuming it is].” 

But with the latest AI tools, however, creative teams can scale up creative development, and personalize them across audience segments. They can produce dozens of versions of product imagery in minutes. Tools like Descript help can edit narration in videos without having to re-record. They can reformat an asset to fit native formats for multiple digital channels, with just a few clicks. And auto-generate copy variants to test across channels. Within the next year, I anticipate tools that will systematically create variants of imagery, narrator, storyline, music and even dominant emotion in video ads, that speak the motivators for a given audience. Freed from the task of resizing and re-editing assets, it means more time for creators to spend on conceptual thinking and brand story development.

— 

Now let’s take a step back and look at the larger picture. Over the last ~15 years, the focus of marketers has been pulled away from brand advertising to hyper-optimized bottom of the funnel direct response ads. Why? Because that’s where the data was available, you could A/B test dozens of creatives together, and the previous generation of AI/ML models worked better with simple features in text/images and optimizing based on high frequency click-throughs and add-to-cart actions as the objective function. The efficiency from these models made testing and executing direct-response ads really efficient, but they couldn’t help develop or refine the brand story or be trained on the audience’s emotional response. Generative AI is now changing that – It’s making it possible to get much richer and granular consumer insights, help produce creatives at a much faster speed, and test & experiment at the top of the funnel. With this efficiency, the focus of marketing is about to swing back towards building the brand!


About the author:

Anshuk Gandhi leads product development at Swayable, a technology company that helps brands pre-test the impact of campaigns and messages on consumer perception and persuasion. He is a data-oriented product builder and is currently focused on applications of AI to draw deeper understanding of why certain campaigns are more effective than others.

Previously, Anshuk helped build recruiting and job seeking experiences at LinkedIn and Triplebyte. Prior to his work in technology, he was a management consultant at McKinsey & Company. After hours, Anshuk helps promote live jazz in the SF Bay Area.

Note from the author: This article links to or references certain AI tools as examples. However, it’s not intended to be an exhaustive list. There are numerous teams working on AI products that I may inadvertently overlook.


Also read: GeologicAI’s robot geologists get US$20M to scale its fleet

Editor

The Tech Factor is a new-age media platform focused on Canada. Our mission is to provide insightful and comprehensive coverage of the Canadian startup and technology ecosystem. To connect with us, report any inaccuracy in our writing, or to share news about the ecosystem, please reach out at editor [at] thetechfactor.ca or fill the 'Submit Tips' form in the top menu.

Related post