Skip to main content
A user model is a simulated version of a specific group of people. It is built from real human data and predicts how that group would answer questions it has never been asked, or what its members would do on a page they have never seen. User models are built with Anacreon, our simulation engine. Each one simulates hundreds of respondents, each modelled on one real person, so you can see where opinions split rather than only an average.

​
Public user models

Public user models are open to every account. The first built with Anacreon launch in November 2026, starting with US and UK consumer finance and online shoppers. They replace the public models we offered before, which were built with an earlier engine and were less accurate. Until then you can create an account and set up your context, but every simulation needs a user model, so you can only plan and run simulations if your organisation has its own. User Models in the app shows what is coming and lets you ask to be emailed when they launch.

​
Your own user models

With Semilattice Custom, we build a user model of your customers, a target segment or any audience you can describe, from your own research. It is private to your organisation, listed under your organisation’s name in User Models, and tested for accuracy before you use it. A custom model is built from unstructured data about your users: interviews, open-ended survey responses, reviews, support conversations and community posts. The more people the data covers, the better the model answers questions it has not seen. Book a call to scope one.

​
Segments

You can segment a user model into a more specific group, such as one income band or age group, and get feedback from that group alone. In a consumer finance user model, a segment of the higher income bands gives you answers from the people in those bands and no one else. To make one, open a user model in User Models, choose Segment in the bar under the map, and tick the attribute values to keep. The number of matching simulated users updates as you tick, and a segment needs at least 100. Name it, add a description if you want one, and save. You can also ask the agent in a project to make or change a segment.
  • Where it’s listed: with your other user models, labelled Segment of and the name of the user model it was made from. That user model lists its segments under Segments.
  • Accuracy: a segment has no accuracy of its own yet. Its page shows the accuracy of the user model it was made from.
  • Region: the same as the user model it was made from. The region decides where a user model appears on the User Models map and which region filter shows it.
  • Who can use it: everyone in the workspace you made it in, or only you in a personal workspace.
  • Changing it: only the person who made a segment can rename, change or delete it. Once a simulation has used it, its filters can’t change, so that simulation’s results still describe the same people. You can still rename it, and it can’t be deleted.
  • Which user models: you can make a segment from a user model built with Anacreon, but not from another segment.

​
Accuracy

Every user model’s page shows its accuracy, measured by comparing its predictions with how real people answered questions the model was not built from. See Trust & measurement for how it is measured and when to rely on it.

​
Retired user models

The earlier public models have been retired. Simulations you ran with them still open with their results, and the model’s page is marked as deprecated, but you can’t run new simulations with them.