Skip to content
Sentia Labs
How-to

Anatomy of a simulation report

The fastest way to judge synthetic research is to read one report. A plain walk through the sample simulation that ships in every new Sentia workspace, and how to read any report like it.

3 min readSentia Labs

The best way to evaluate Sentia is not another essay. It is to open a report and see what you get.

Every new workspace ships with a sample simulation already finished. This post walks through that report in the order you should read it. The sample is authored so a fresh workspace has a full report to explore right away. The structure matches a live run. Use it to learn the reading habit, then judge live runs on your own product.

The setup

The panel is 18 buyer profiles: product teams evaluating analytics tools. They choose between two pricing pages.

  • Version A: usage-based billing
  • Version B: seat-based billing

Open the sample in your workspace and follow along.

1. The topline split

Usage-based wins, 61% to 39%.

Treat that as a direction, not the answer. A single split only tells you which way the panel leaned. It does not tell you whether the lean is trustworthy, shared, or useful.

Keep reading.

2. Who disagreed

By segment, the story changes:

  • Startup buyers prefer usage-based 86%
  • Mid-market prefer usage-based 67%
  • Enterprise buyers reverse it: seat-based wins about four to one

That reversal is the finding.

"Usage-based wins" would ship one page. "Usage-based wins except for the buyers who sign your largest contracts" ships a default page plus an enterprise tier. That is what the report's recommendations conclude.

If a report cannot show disagreement by segment, do not trust the average.

3. Why they chose

Every agent has a written rationale. The enterprise dissent is not just a percentage. It is someone like a VP of Product at a 900-person company, with scar tissue from a usage-billing overage their CFO chased for a quarter.

Their rationale is plain: "Seats times price is the whole invoice." They will pay more for a number that stays the same in January and June.

Spend time here. You want two things:

  1. A surprise: an objection your planning meeting missed, said the way a buyer would say it
  2. A quality check: rationales tied to real constraints beat generic copy that could fit any product

Generic rationales are a grounding warning. Treat the tallies above them carefully.

4. What actually drove the split

The report asks which traits predicted the choice. In this sample, it is not company size.

It is loss aversion (r = -0.81). Of the six agents highest in loss aversion, none chose the usage-based page, in any segment.

That changes the fix. The winning page does not need a different audience. It needs a spend cap and billing alerts that reduce fear of an open-ended bill. That is the report's first recommendation.

A static persona deck would have sorted by job title and missed this.

5. Findings on the page

Concerns land on the actual frames, not in a separate summary:

  • The estimator never defines what a billable "event" is at purchase
  • View-only seats get billed like editors
  • Accessibility issues pin to the same screens

There is also a SUS-style usability score. The report labels it clearly: a model-internal proxy, not a human SUS benchmark. Prefer reports that mark uncertainty instead of presenting every number with equal confidence.

6. The verdict

The winning page is labeled ready, with risk:

  • Ship usage-based as the default
  • Add a spend cap before you scale spend
  • Add a committed-usage tier before enterprise campaigns

Each recommendation names its evidence and the segment it affects. That makes the next step assignable, not inspirational.

A useful report does not only say "A wins." It says what would make the win safe, and who still dissents.

How to read the next report

  1. Treat the topline as a direction, not a forecast.
  2. Check segments before you celebrate. The reversal is often the finding.
  3. Read rationales until you hit a surprise or a generic sentence.
  4. Ask what drove the split. A trait is a design lever. A demographic split is only a targeting note.
  5. Prefer reports that label proxies and gaps honestly.

The sample is in every new workspace on the free tier. Open yours, then run the same structure on a decision you are arguing about this week. If the rationales sound like your users, you have a method worth keeping. If you want a second pair of eyes on that first run, talk to us.

humanagent