Ideally CEO James Donald says AI is central to his firm’s software, from helping to create surveys to identifying themes responses. But he’s also wary of how some have rushed to bandwagon jump on new industry trends, including synthetic data – where he says some have implemented under-cooked solutions. His firm has its synthetic pilot, but it remains in-house at this point. Photo / Sylvie Whinray
The start-up claims 750% growth over 2024 and 2025 and 350% US revenue growth since it opened its New York office just months ago from customers including Duckhorn, Tilray and Rémy Cointreau.
Donald won’t give financials, but the numbers have been sufficiently attractive to attract new investors. The Series A round was led by Australia’s Shearwater Capital and supported by New Zealand’s Altered Capital plus a second Aussie outfit, Ecliptic VC.
Icehouse Ventures, which led a $6m seed extension round in 2024 and an earlier $2m seed round, also chipped in new funds.
“By helping companies to explore, test, and innovate in real time, the Ideally team are proving Jevons’ paradox within market research,” Altered Capital managing partner McGregor Fea told the Herald.
Jevons’ paradox holds that if a product becomes more efficient, you actually use it more, not less.
“Ideally enables the digitisation of a complex role currently fulfilled by high-cost external agencies, bringing critical market insights in-house and on-demand.”
Ideally’s new product Canvas, also announced today, takes the in-house, DIY, real-time ethos further. It’s designed to taste-test elements of a product or a campaign while ideas are still being kicked around in the early stages of its development.
“Through derisking the innovation process, Ideally helps companies to find product-market-fit faster, cheaper, and more reliably,” Fea says.
Icehouse Ventures chief executive Robbie Paul said, “We are seeing a fundamental restructuring of marketing departments globally, and Ideally is the infrastructure supporting that transition.
“Their ability to reach a $100m valuation in just over two years while maintaining world-class capital efficiency made this a clear investment for us,” says CEO Robbie Paul.”
And Shearwater managing partner Zac Zavos said, “Ideally has taken what was a long, expensive, waterfall research process and rebuilt it as something fast, iterative, and accessible. Once marketers experience the speed of overnight responses, it becomes indispensable.
“They’re moving fast and building the category as they go.”
Ideally is something of a sibling to another hot market research start-up, Tracksuit.
Both were initially bankrolled by well-established Australasian market research and analytics firm TRA (The Research Agency), which remains Ideally’s largest shareholder.
The rise of synthetic data
One of the hottest trends in market research is “synthetic data”, Donald says.
The industry has already shifted from door-to-door and phone to online tools to reach people. For some, the next step is removing the humans altogether by sampling a set of “synthetic” people, or AI bots who have been trained to think like members of the public.
Synthetic data can be used for an entire survey, or to fill in gaps, particularly if a certain demographic has proved hard to reach.
Donald is doing more than keeping tabs on the development. Ideally has a pilot running in-house. But he’s still not convinced the technology is ready for prime time, and says some early attempts by others have been a bit rough.
“There are different types of synthetic data,” Donald says.
“The one everyone got excited about was the quick and dirty one, which was to just ask an LLM [large language model] to act like you and then just get it so wrong, because the underlying data is, like, Quora and Wikipedia.”
Donald says best-practice synthetic data is created by taking questions and responses from a statistically representative population of people, then using that to train a synthetic group.
“But there still need to be guardrails,” he says. “Like, if it’s trained for beer; all your attitudes and behaviours around beer, don’t go and ask ice cream questions.”
Ideally co-founders Joshua Nu’u-Steele (left), James Donald and Brendan Cervin.
The so-called SaaSpocalypse is looming over all software firms right now – the fear that AIs will replicate a SaaS (software-as-a-service) company’s features, or allow upstarts to “vibe code” them at low cost.
Where does Donald stand on the threat of AI eating his firm’s lunch?
“The really important difference here is the type of software business,” Donald says.
“We see the share prices for the likes of Xero and Atlassian – which is a software services product out of Australia – have significantly decreased because they go across lots of different industries, but don’t do anything deep, or have anything proprietary, like data, so they become easier to replicate.”
Xero founder Rod Drury recently argued that the cloud accounting software company does have a data advantage that insulates it from AI, even if investors are not, so far, buying that argument.
The dupes could be cheaper, too.
“We’re a bit different where we go vertical. We solve one problem, which is not for everyone, but it means we go really deep and we build really high-quality data, and that data becomes very valuable because we find lots of different ways to reuse that data.
“So defensibility, meaning hard to copy, is becoming really important and valued by the markets in this new world.”
Chris Keall is an Auckland-based member of the Herald’s business team. He joined the Herald in 2018 and is the technology editor and a senior business writer.