# Explore how different people experience your product.

Use behavior simulation to catch product and usability issues early. See how your user base interacts with a product change, from prototype to staging, and where there is friction in the user journey.

[See how Finch works](/content/products/finch#example/index.html)

## The question
How do different people experience the same product?

## What you examine
UX journeys and product usability

### Generic models weren't built to model human behavior.

Plausible isn't the same as probable. General-purpose AI models are good at generating convincing human responses. But a plausible explanation isn't necessarily an accurate representation of how someone would behave. Finch is built on research into human behavior, with technology designed to simulate how people interpret information, make decisions and act.

### General-purpose models

#### Generate a plausible response.
- A model can describe how a hypothetical user might behave without necessarily reproducing the choices that user would actually make.

### Finch/Behavioral models

#### Simulate the decisions behind it
- Model how differences in goals, experiences and available information influence the paths people take and the decisions they make.

## From product questions to behavioral insights.

Explore how your users might interact with a new product change, before putting it in front of real people.

1. **Set up your agents**  
   ### Model your users  
   Create behavioral agents grounded in real-world clickstream data and enriched with context about your users. Built on Seldon's behavioral technology, designed to capture differences in goals, knowledge, and decision-making.

2. **Bring your product**  
   ### Choose what to test  
   Test a prototype, staging environment, or live product. Finch works best with interactive experiences where agents can navigate and explore.

3. **Run your experiment**  
   ### Run usability tests in silico.  
   Stress-test assumptions, uncover friction, and explore how different users respond. Get insights early and focus human tests where they'll matter most.

## Simulated journeys you can inspect.

Synthetic data extends the reach of real-world data, bringing scale, variation and edge cases. But trust needs to be earned. Simulation works alongside real-world testing, with verification and validation at its core.

### Traceability
#### Follow every journey.
Review the actions and decision points behind a simulated journey. Trace where an agent understood, hesitated or got stuck.

### Transparency
#### Know what informed it.
Inspect the goals, knowledge and context behind each agent's actions, so you can understand why their experiences differ.

### Human oversight
#### Guide the next test.
Use simulated findings to focus your research questions, then test the changes with real people.

## Real world
Simulated journeys  
Compare where decisions align and where they diverge.

## Evidence & evaluation

### Calibrate against real behavior
We work with your team to ground behavioral agents in real-world data, from clickstreams to user research and domain-specific context.

By backtesting their decisions against observed human behavior, we can evaluate and refine how accurately they simulate your users.

[Explore the technology](/content/technology/index.html)

## Stress test your product with behavioral agents.
Bring a problem or a product question that is top of mind and we'll walk through how we can help your team investigate it.

Book a demo
