A Quick Guide to Building Your Own Synthetic Audiences

Today at the Museums & Heritage Show we did a crash course in how to build your own synthetic audiences.

Of course there are more accurate and technical ways to build refined, scalable and enterprise level synthetics audiences, as we do here at Altair, but the objective behind today’s talk was to highlight that you can all build your own.

It’s not advanced and it will give you a valuable way to interrogate your data quickly and with greater depth than you can now. 

Our approach was based on using Chat GPT, as it is the simplest to use in this use case.

That said, you do need to be careful that you are using a LLM platform where your data stays secure.  You need to ensure that your data is not being used to train models or being released out into the internet!  Be warned…

On that basis, let’s crack on.

Why build your own synthetic audiences?

You’ve got a lot of data and it’s currently sitting in static spreadsheets and pie graphs.  Wouldn’t it be way better to ask questions of it, as if it were a real person?  Of course it would, so plugging your data into an LLM to query and return real-world responses will make a massive difference to your world.

You should also think about building a few.  Why just have one?  You could have a whole focus group to chat with.

Will this replace your analysts, your data team or even you…?!

No, people need to do the analysis and guidance.  We all know LLMs can be so very, very convincing and authoritative, whilst also being very, very wrong.

Making sure humans are at the beginning of the process building and refining is very important.

Making sure humans are at the end to ask follow up questions and make final decisions is crucial.  Your experience guides and decides.

So, how do you create your own synthetic audience?