ONE DECISION
DAY HIKE
Start with one product decision. Review where different personas get stuck and what to test next.
- Scenario design workshop
- Up to four user models
- Journey review with your team
Simulate what users might do before you make a change
Explore how personas might navigate a product. Review where they get stuck and decide what to test next.
Illustrative outcomes
Analytics shows what happened.
Simulation shows what’s
likely next.
Analytics tells you what already happened: what people clicked, where they dropped off, what they bought. Teams use it every day to decide what to do next.
But when you change something, past numbers can’t tell you how people will react. So the decision comes down to a guess, and you wait to see if you were right.
A simulation lets you explore a change with personas that have different goals, expectations, and limits.
Review where the paths split and decide which questions to take to real customers.
Our mission
Bearing helps teams explore how different people might use a product. Follow findings back to the session, then decide what to check with customers.
We simulate the curiosity, confusion, and choices of the people who use your product
one
Define the scenario
Start with a decision. Set the starting conditions, the experience under study, and the outcomes worth watching.
two
Define the personas
Give each agent a reason to be on the trail: goals, prior knowledge, expectations, and limits, grounded in your research and behavioral data.
three
Let them use it
A simulated persona follows a path through the product. Review where the session continued or stopped.
four
Repeat & compare
Compare journeys across personas and versions to see where the trail forks.
five
Read the evidence
Inspect every journey, click, and stall. Leave with your bearings on what is likely, and sharper questions for real customers.
What could you simulate?
Explore how simulation can inform the product decisions your team makes every week.
Other questions you could ask
Watch a simulation run
Each hiker is one agent on the same journey: through onboarding in Fernbook, a pretend bookkeeping app, to its first report. At every step, some turn back and the rest press on.
This example uses a scripted model. Use the results to form questions you can test with real people.
Start with personas, or define your own
Give each persona a goal, prior knowledge, and limits. These examples show how their starting points can differ.
Example personas
A returning expert may skip the steps a first-timer needs. A skeptical evaluator may look for proof before starting.
Personas informed by your customers
Use what your customers have said and done to describe how a persona approaches a task.
- Goals and prior knowledge
- Expectations from past experience
- Time, device, and access needs
Your personabuilt from your data
Ways to work together
Three ways a team could use simulation. The Bearing workspace includes a scripted demo.
ALWAYS-ON SIMULATION
BASE CAMP
Repeat the same scenario across releases to look for changes in where people might get stuck.
- Release-by-release comparisons
- A shared library of inspectable journeys
- Summaries in the tools your team already uses
CALIBRATED TO YOU
HOME RANGE
Use customer research to guide your persona design. Compare simulation results with what people did on the same tasks.
- Custom calibration on your data
- Backtests on held-out tasks
- A standing partnership with your research team
Common questions
What teams usually ask before they run a first simulation.
What is customer behavior simulation?
It is a way to see how people are likely to respond to a product or a change before you ship it. Start with personas that have different goals and limits. Review where each session went and where it stopped.
Is a simulation a prediction?
No. A prediction says what will happen. A simulation shows what looks likely and how sure we are. It tells you where to look and what to test, and a simulated completion rate is never a human conversion rate.
How accurate is it?
Compare simulation results with what real people do on the same tasks. This page shows an illustrative example.
How is this different from asking an AI to pretend to be a customer?
A session gives you a record to inspect. In the workspace, findings link to moments in a recording of the demo product.
Does this replace user research, analytics, or A/B tests?
No. Simulation sits beside them. It helps you form sharper hypotheses and choose what is worth testing, so your time with real customers goes further. Real behavior is still how you find out what is true.
How are the agents built?
Each agent starts with a reason to be there: a goal, prior knowledge, expectations, and limits. Customer research can help you define the starting conditions. The local demo uses scripted behavior.
What can I try in the workspace?
Set up a simulation, start a scripted run, and follow a finding to its moment in the recorded session.
Do I need customer data?
The demo includes example personas and recordings of a fictional product.