Funds for British Synthetic Data Firm Electric Twin

In the UK, synthetic audience modelling firm Electric Twin has secured $14m in funding, for international expansion and further development of its prediction technology, including new use cases. The raise includes a previously undisclosed $4m pre-seed round.

Electric Twin was founded by Dr Ben Warner, who spent four years at early AI company Faculty, and two as a senior advisor to the Prime Minister on 'digital and data' during the pandemic; and former army officer Alex Cooper, who was COO at carbon removal firm UNDO for two years prior to starting work on Electric Twin in 2023. The company combines real-world survey data with large language models, social science research and machine learning to create models predicting people's reaction to communications, proposals, products and scenarios.

The new, $10 million round was led by Atomico, with participation from LocalGlobe, Mercuri and Samos Investments, plus angel investors including 'software will eat the world' man Marc Andreessen and former Kantar CEO Eric Salama.

CEO Cooper says the company grew out of experiences helping governments make decisions with limited information, in 2020-21. On the company's blog, he states: 'During the pandemic, my co-founder Ben and I were both in Number 10, watching decisions get made about lockdowns, restrictions, and national measures. Decisions that affected millions of people, made under pressure, with incomplete information about how the public would actually respond. The systems we had for understanding human behaviour simply weren't fast enough or accurate enough at the time. We sat in those rooms wishing we had tools that didn't exist yet. That experience is why we founded Electric Twin. We're now building the tools we wish we'd had.'

Cooper says the company last year ran 'what we believe is the first real-world comparison of synthetic audience predictions against conventional commercial research methodologies' - comparing with eight identical surveys covering 'every major approach the industry uses: aggregator panels, opt-in panels, river sampling, microtask platforms' - and more than eight thousand respondents. Predictions generated in parallel with this using the platform showed a mean absolute error of 0.05 against the baseline constructed from the average of traditional methods. 'Our synthetic audience predictions were indistinguishable from conventional research. '

Web site: www.electrictwin.com .