pricing teams
Price increase
How far can price move before churn pressure accelerates?
behavioral stress testing
Find the point where your customers start to move.
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
The population was already moving.
We analyzed 377,487 public comments around Netflix. A large reversible-friction cohort was building while projected hard churn remained close to 2%.
377,487public comments read
~12%active, reversible friction
~2%projected hard churn
~2.0 to 2.5%independent external churn range
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
What decision are you about to make?
subscription businesses
Subscription change
Which customers become unstable when the proposition changes?
product leaders
Product redesign
At what level of disruption does friction turn into defection?
retention teams
Retention intervention
Which part of the population can still move back?
the engine, live
Push the pressure. Watch the population move.
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.
combined friction0.00
tipping multiplier1.0×
statepassive_inertia
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
From a decision to a stress test.
Bring the decision
"What happens if we increase subscription price by 10%?" A sentence describing the decision is enough.
Build the population model
Difference Machine derives the behavioral baseline, friction structure and relevant transitions from available population data.
Read the population
Identify where pressure begins to move people, which transitions accelerate and how much of the population is exposed.
Stress-test the decision
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
Built for constrained behavioral analysis.
Population-specific baseline
The starting point comes from observed behavior in the population being tested rather than a generic scoring scale.
I = μ + σ
Pressure interaction
Frictions can amplify or suppress each other, allowing the model to represent interacting behavioral pressure.
Σ = correlation matrix
Constrained by design
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
Visible assumptions
Data limitations, calibration choices and external benchmarks remain explicit in the output.
assumptions ∈ output
stress tests
See the engine applied to different populations.
streaming · subscribers
jun 2026
Netflix: the population moves before churn appears.
A large reversible friction cohort develops while projected hard churn remains low.
automotive · owners
jun 2026
Porsche: where the premium is under pressure.
Which friction is closest to triggering brand exit, and where demand may move.
coffee · subscriptions
jul 2026
Coffee subscriptions: recoverable churn.
Where subscribers go after cancellation, and how much churn remains commercially recoverable.
supplements · subscriptions
jul 2026
Supplements: unrecoverable churn.
Churn that can be won back, separated from exits where retention has little leverage.
beyond customers
One engine. Different populations.
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.
Find the cliff before you cross it.
What decision are you about to make? Tell us the population and the change you are considering. Public data is enough to start.
Stress-test a decision
no deck required. no dataset required. a founder reads every request.