Synthetic Data Launch for Ipsos India

In India, Ipsos has announced it is augmenting its product testing capability with synthetic data, promising faster insights and product launches.

The group says the new offering, which is now rolling out to clients, has been shown in testing to speed up product tests by around 50%, provide richer insights into sub-groups, and save between 20% and 60% on costs. CEO Ipsos India Amit Adarkar says this has been achieved 'without compromising on rigor', and adds: 'Successful deployment of synthetic data hinges on the extent to which synthetic data mirrors real consumer data. With Ipsos' thought leadership in product testing and the extensive on AI model training experience, we are confident to boost market research agility and reduce costs, through cutting edge research.'

Ipsos is involved in partnerships developing synthetic data elsewhere in the world, including two for OOH ad audience measurement: the MOVE2 system with Australia's Outdoor Media Association (OMA) - which will use a 'synthetic audience' of 2.2 million to estimate and predict the visibility and engagement of OOH advertising sites - and a contract with Route, the UK's OOH JIC / currency provider, developing a synthetic dataset representing the travel habits of the population and their journeys for every day of the year.

Anthony Dsouza (pictured), Head of Innovation at Ipsos India, stresses the importance of training such AI-based systems on appropriate real-world data, and states: 'Synthetic numerical data should at least correspond to real-world data on key statistical measures such as means, data distributions, variances, and inter-variable relationships, including correlations... The closer the synthetic data aligns with the real data, the lower the inherent risk involved; however, some risk always persists because synthetic data cannot flawlessly replicate every facet of real data. Consequently, synthetic data should be utilized only when one is prepared to accept a certain level of risk.' Dsouza says using synthetic data in product testing isn't meant to replace human input altogether, but rather to enhance it - by determining the minimum number of human respondents needed to work alongside synthetic data for reliable results.

Group home page: www.ipsos.com .