Book a Demo
Book a Demo
We’d love to show you what we do
We’d love to show you what we do
We’d love to show you what we do.
We’d love to show you what we do.
Book a demo to see how Electric Twin helps teams end guesswork. We’ll show you how we work, what we can unlock for your organisation, and where we can make the biggest impact.
F&Q’s
Need Help?
What is Electric Twin?
Electric Twin builds simulations of a customer's target audiences. They can use these to test decisions at speed and scale, across their entire teams. Test any decision, immediately and iterate at 100,000x the speed.
How does it work?
Electric Twin makes predictions through a system, combining the latest advances in large language models and traditional machine learning, with social science theory and seed data on the target population.
How accurate is it?
With the right dataset we expect accuracy >0.90 NDAM - a distributional accuracy that closely matches that of real-world surveys. We have run tens of thousands of evaluations - independently, with customers and leading academia. These have covered topics of customer sentiment in retail, media, healthcare, AI among others.
What data can I use?
We typically suggest that customers kick-off with a survey or focus group dataset that is closely aligned to their target problem scope. The best dataset to use includes a combination of general questions (e.g. demographics, other) and topic-related questions (e.g. opinions / attitudes on a certain category). For most companies, their existing marketing research works great!
How fast can I get started?
Companies can get started in a couple of days. We typically schedule a demo with your team to understand your use case better and explore what is the best seed data to use for your audience simulation. In almost all cases, we are able to start in a couple of days, onboarding one of your existing surveys.
What if I don’t have a survey that I can use?
There are a number of options. Our team has a lot of experience conducting surveys, making predictions and consulting large enterprises in their understanding of human behaviour. We can work with you to conduct a single survey that you can use to create your synthetic population. Alternatively, we do also have a number of pre-built audiences that we can give you access to out-of-the-box.
What are the main use cases?
We see customers use us in a range of use cases. In almost every single one of our pilots, we have onboarded new departments that were interested to use Electric Twin. Some of the common ones include new product development, product testing, message and campaign testing (text and image), targeting and expanding into new audiences, geographical expansion, testing sensitive content in private. If you are not sure if your use case could work well, reach out to us.
What can it not do?
At the moment, we would not recommend using us as a replacement or automation of tracking polls or brand trackers. We are constantly expanding the scope of the possible. Over the past year, we have shown our engine to be much more powerful than expected, even in areas that we thought may be challenging for us.
Why can’t I just use ChatGPT?
Large language models contain a lot of very valuable information on human behaviour and opinions, but alone: - Struggle to represent target (vs generic) audiences - Revert to the mean - Fail to capture the distribution in a population well - Reinforce the pre-set beliefs, used to create the population
F&Q’s
Need Help?
What is Electric Twin?
How does it work?
How accurate is it?
What data can I use?
How fast can I get started?
What if I don’t have a survey that I can use?
What are the main use cases?
What can it not do?
Why can’t I just use ChatGPT?
F&Q’s
Need Help?
What is Electric Twin?
Electric Twin builds simulations of a customer's target audiences. They can use these to test decisions at speed and scale, across their entire teams. Test any decision, immediately and iterate at 100,000x the speed.
How does it work?
Electric Twin makes predictions through a system, combining the latest advances in large language models and traditional machine learning, with social science theory and seed data on the target population.
How accurate is it?
With the right dataset we expect accuracy >0.90 NDAM - a distributional accuracy that closely matches that of real-world surveys. We have run tens of thousands of evaluations - independently, with customers and leading academia. These have covered topics of customer sentiment in retail, media, healthcare, AI among others.
What data can I use?
We typically suggest that customers kick-off with a survey or focus group dataset that is closely aligned to their target problem scope. The best dataset to use includes a combination of general questions (e.g. demographics, other) and topic-related questions (e.g. opinions / attitudes on a certain category). For most companies, their existing marketing research works great!
How fast can I get started?
Companies can get started in a couple of days. We typically schedule a demo with your team to understand your use case better and explore what is the best seed data to use for your audience simulation. In almost all cases, we are able to start in a couple of days, onboarding one of your existing surveys.
What if I don’t have a survey that I can use?
There are a number of options. Our team has a lot of experience conducting surveys, making predictions and consulting large enterprises in their understanding of human behaviour. We can work with you to conduct a single survey that you can use to create your synthetic population. Alternatively, we do also have a number of pre-built audiences that we can give you access to out-of-the-box.
What are the main use cases?
We see customers use us in a range of use cases. In almost every single one of our pilots, we have onboarded new departments that were interested to use Electric Twin. Some of the common ones include new product development, product testing, message and campaign testing (text and image), targeting and expanding into new audiences, geographical expansion, testing sensitive content in private. If you are not sure if your use case could work well, reach out to us.
What can it not do?
At the moment, we would not recommend using us as a replacement or automation of tracking polls or brand trackers. We are constantly expanding the scope of the possible. Over the past year, we have shown our engine to be much more powerful than expected, even in areas that we thought may be challenging for us.
Why can’t I just use ChatGPT?
Large language models contain a lot of very valuable information on human behaviour and opinions, but alone: - Struggle to represent target (vs generic) audiences - Revert to the mean - Fail to capture the distribution in a population well - Reinforce the pre-set beliefs, used to create the population