pricing teams
Price increase
How far can price move before churn pressure accelerates?
behavioral stress testing
Difference Machine models a customer population under pressure so you can test a decision before you make it. See where behavioral movement begins and how much of the population crosses the boundary.
public data is enough to start. no customer database required.
proof
We analyzed 377,487 public comments around Netflix. A large reversible-friction cohort was building while projected hard churn remained close to 2%.
what it shows
The hard-churn tier was not calibrated to the external estimate. The larger reversible cohort shows the behavioral movement that a churn number alone does not reveal: customers who cancel and return, reassess around releases, and erode revenue without ever showing as a lost subscriber.
projections from public discussion, not measured churn. full method in the report.
Read the Netflix stress test no login · no sales calluse cases
pricing teams
How far can price move before churn pressure accelerates?
subscription businesses
Which customers become unstable when the proposition changes?
product leaders
At what level of disruption does friction turn into defection?
retention teams
Which part of the population can still move back?
the engine, live
Every population begins with its own behavioral baseline. Increase the pressure and Difference Machine recalculates how people move between states. As pressure accumulates, some transitions accelerate. That is the behavioral cliff.
stable, population on autopilot.
this demo uses a sample population. a real model is built from the language, friction structure and behavioral baseline of the population being tested.
what you get
"What happens if we increase subscription price by 10%?" A sentence describing the decision is enough.
Difference Machine derives the behavioral baseline, friction structure and relevant transitions from available population data.
Identify where pressure begins to move people, which transitions accelerate and how much of the population is exposed.
Change the proposed pressure and observe how the projected population response changes.
public data can be enough to begin. first-party data can deepen the model where available.
why trust the output
The starting point comes from observed behavior in the population being tested rather than a generic scoring scale.
I = μ + σ
Frictions can amplify or suppress each other, allowing the model to represent interacting behavioral pressure.
Σ = correlation matrix
The model operates inside fixed behavioral states and valid transitions. Weak evidence can remain unresolved rather than being forced into a confident classification.
0.0 ≤ x ≤ 1.0
Data limitations, calibration choices and external benchmarks remain explicit in the output.
assumptions ∈ output
stress tests
A large reversible friction cohort develops while projected hard churn remains low.
Which friction is closest to triggering brand exit, and where demand may move.
Where subscribers go after cancellation, and how much churn remains commercially recoverable.
Churn that can be won back, separated from exits where retention has little leverage.
beyond customers
The simulation core is unitless. The behavioral structure can therefore be applied to different populations without changing the underlying engine.
customers · employees · voters · patients · users and communities
Behavior responds to density. One complaint rarely changes a population. Enough interacting pressure can. The full argument is in our field note density, not presence.
What decision are you about to make? Tell us the population and the change you are considering. Public data is enough to start.