AI Simulation

Simulation returns a quick summary of likely audience responses. It is built for early hypothesis testing: comparing a handful of headlines or concepts, sanity-checking a price point, or deciding which ideas deserve more budget. It runs on your study data plus the audience and data sources you choose, so the answers are grounded in real respondent data rather than generated from scratch.

A common path is to start with Simulation, move to Synthetic Data for the ideas that look promising, and use Digital Twins to dig into the why. When you need statistical rigor for a final decision, confirm with a fielded study.

Simulation lives inside a completed survey study, on its own Simulation tab next to your results. You need:

• A completed survey study with the questions you want to simulate.
• Access to the feature. Simulation is rolling out in stages and was released to admins first, so ask your Panoplai contact if you don't see the tab.
• An idea of who you want to hear from: a saved audience, or the study's own audience.

Simulation is not offered for every kind of data. It works from survey data in a study, so uploaded files (reports, transcripts, decks) are not simulated directly.

1. Open a completed study and select Results, then the Simulation tab. You will see your past simulations, newest first, each with its name, date, creator and a progress bar showing how many questions are complete.
2. Select New simulation. The wizard has four steps, and Back keeps your choices while Cancel exits without creating anything.
3. Choose your questions. Pick the survey questions you want to simulate. You can run one question or several in a single simulation.
4. Add your screens. If you are testing concepts, headlines or other stimuli, add them here.
5. Choose your audience. Pick a saved audience, re-use one you built earlier, or create one from the Explorer by applying filters and selecting Simulate with this audience. Check the audience summary before you continue.
6. Choose your data sources. Switch the sources you want on or off: your own client data (CDU), the platform's respondent data (PDU) and Global. Answers draw only on the sources that are on. An Include general knowledge option is available if you want the model to draw on general knowledge as well.
7. Review the summary, then run the simulation. Results appear on the same tab when it completes.

Simulation returns a quantified top-line score for each question, with a confidence indication, rather than a row-level dataset. Treat it as a directional read: it tells you which ideas are likely to land and which are not, not the exact percentage you would see in a fielded study.

A coverage and confidence estimate is being added to Simulation results. Until it is available, use the results to rank and filter ideas, and confirm anything you plan to quote externally with Synthetic Data or a human study.

• Start narrow. Simulate two or three headlines, concepts or price points first, then widen once you see a pattern.
• Match the audience to the question. A tighter audience gives a more relevant read, but very small bases are less reliable.
• Save audiences you will re-use. Saved audiences carry their study, file and data-source settings, so repeat runs stay consistent.
• Be deliberate about data sources. Turn sources on or off one at a time to see what drives an answer.
• Use it as a filter, not a verdict. Simulate many ideas, keep the strongest, then validate those with Synthetic Data, Digital Twins or a fielded study.

Need help? Contact your Panoplai customer success manager or email support.