Synthetic data specialist Ideally wants research to be used at the start of the creative process

There’s significant debate about synthetic audiences. Sure, they save agencies and brands considerably on time and money, but there are inherent risks as well. Probably the biggest is that most synthetic data models are regressive — they trend toward the mean. They smooth out the edges, compress the variance, and pull everything within a standard deviation or two of average. 

But another risk is that too much research is commissioned but seldom used, and within the creative world, rarely used as a springboard to informing creative ideas at their outset.

Ideally, a New Zealand-based market research firm, is looking to reinvent research when serving the creative side of the industry by integrating consumer insights earlier in the creative process, enhancing creativity and reducing risk. The platform connects data from various teams, enabling compounding knowledge and synthetic data usage.

“We wanted to pressure test whether, if we brought the consumer voice in earlier in the creative process, we could help creatives and marketers and agencies push the bounds a little more, take a little bit more risk, and feel more confident in their approach,” said Joshua Nu’u- Steele, co-founder and chief revenue officer of Ideally, which started in 2023 right around the time generative AI started to bloom. “How do we add a few more voices into the [creative] mix. It’s not a yes or no, it’s not a pass or fail — it’s ‘I hit a fork in the road, which way should I go?’ Or ‘I’ve got this gut feeling, I just want to pressure test it’.”

According to Nu’u-Steele, the more different teams use and offer feedback or guide the next steps, the more the research learns from itself, thanks to a layer of AI applied to the process. “We built the ability for all of these different teams to do the testing through one platform, all of that data connect to each other and inform each other,” he said. “Because every time you’re doing a test, that data is compounding. So if I’ve spoken to a certain segment about a certain topic, that is going to add context to a future test if I would like it to. That way it’s also leading to where we’re taking the product, which is the ability to use synthetic in a robust way.”

Ideally counts Omnicom Oceania as a client, and the agency group has used Ideally in over 60 tests and projects, wending its way into the holding company’s Omni platform, Acxiom, RealID and Flywheel. The multi-year deal, according to Daniel Shaw, Omnicom’s Intelligence lead, gives every Omnicom Oceania agency access to Ideally’s consumer research platform, with overnight results coming back to the agencies in the network. He said it costs about one-tenth of a traditional research budget and is getting used across 23 brand categories.

“One of the values here is moving from that mindset of ‘I think’ to ‘I know’,” said Shaw. “What’s great around how Ideally plugs in is, we can be quite targeted in where and how it’s bringing in new insight or a depth of insight into a specific area. That allows us from a QA point of view to know that we’re quite targeted around how we then use that iterative process to feed and enrich what we’re doing.” 

“The Omnicom Oceania partnership is the first time we’ve been embedded end to end across a network, and it’s a model we’re now seeing strong interest in globally. It works because the research sits in service of the creative, not the other way round.”

So how does Ideally ensure that the AI-driven element not start to devise its own conclusions, given that it’s synthesizing various data inputs? Nu’u-Steele explained the company has pre-built research frameworks built into the platform — a research team sits alongside product and engineering teams to ensure consistency across benchmarks, statistical significance, category intelligence.

That said, not all observers of the business are fully convinced synthetic’s the way to go. “Synthetic research can inform early stages of the creative process, but should not be used to drive creative decisions,” said Nathan Richter, senior partner at Wakefield Research. “The best research relies on deep insight into how real humans think and feel, and what they respond to. By definition, AI-generated audiences don’t have the capability to fully replicate human behavior and thinking, just replicate it. When used in isolation, synthetic research risks simply recreating the past rather than uncovering the kinds of novel insights that create a successful campaign.”

That’s not stopping Ideally from pushing ahead. The company just launched Research Assistant, an AI feature that builds consumer studies from a plain-language question or an uploaded brief, drafting up framework, audience, and questions in a fraction of the time it used to take. 

Nu’u-Steele is now based in New York, where Ideally has opened an office. He said the U.S. business has grown 400%, with momentum in the wine (Treasury Wine Estates is a foundation customer) and functional food and beverage categories.

More in Marketing

Your favorite beauty brands are now also entertainment brands

Many of today’s beauty brands are asking if entertainment and advertising can be one and the same. 

By the numbers: How marketers are building the infrastructure for AI search

Marketers are carving out budgets for AI search, but research suggests execution, measurement and content strategies are still catching up.

Future of Marketing Briefing: Creator agencies are becoming entertainment studios — brands are footing the bill

Creator agencies are either rebuilding themselves as entertainment companies or launching as one from day one.