reports
Difference Machine reads behavioral movement inside real populations. These published runs show where people become unstable, which pressures sit closest to a decision, and how much of the population can still move back.
Every report uses real public data. Assumptions, calibration and limitations stay visible.
no login required. read the full analysis.
real runs · no login · no sales call
Real populations. Real runs. Each report points the instrument at a different population and publishes what it finds. The value is in the differences: a streaming audience under pricing pressure does not move like a premium automotive population, and coffee subscribers do not leave the category the way supplement subscribers do. The same instrument should produce different answers when the underlying behavioral structure is different.
Where opposition forms, what triggers it, what would defuse it, and how the talk about water compares with the water record.
of 15,896 Texas forum posts are opposed
about 6 in 10
raise water as the main or second concern
roughly 1 in 5
A behavioral read of 377,487 public comments projects roughly 12% of the population in active, reversible friction alongside approximately 2% in hard churn. The hard-churn tier emerged without calibration to the external churn benchmark.
active, reversible friction
~12%
Still moving, still returning, still recoverable.
projected hard churn
~2%
The visible loss is only the end of the transition.
The larger cohort is where the warning appears.
Across more than one million public posts and comments, electric-Porsche resale collapse reached 0.90 on the model's 0 to 1 defection-proximity index in the North American population. Owners discussing that pressure were already naming potential replacements.
north america · resale collapse
0.90
Defection-proximity on a 0 to 1 index.
europe · premium justification
0.60
A separate population produced a different pressure point.
One brand can carry different behavioral risks in different populations.
A defection map built from 8,069 US dog owner decision moments places the fresh dog food price tolerance boundary between $164 and $261 per month. When owners cross it, they do not trade down inside the category: 1,814 completed exits from fresh food stand against 41 switches to a cheaper fresh brand, and most exits happen in a single step.
left fresh food entirely
1,814
Most exits go to kibble, homemade, or raw.
switched to a cheaper fresh brand
41
The retention battle is against category exit, not another fresh brand.
Price sets the boundary. Health decides many of the crossings.
These two reports sit beside each other on purpose. The contrast is the proof.
supplements · subscriptions
63%
unrecoverable exits
Difference Machine extracted 9,030 defection-bearing exits from a 9.1 million post corpus. Across the category, 63% of those exits landed outside the reach of a conventional win-back campaign: customers move toward whole food, bulk ingredients, retail alternatives, or leave the category entirely. Read the supplement churn map
coffee · subscriptions
13%
unrecoverable exits
The same defection-mapping approach found 1,599 typed exits in a corpus of 8.3 million public posts. Here, 13% landed somewhere a retention campaign could no longer follow. A much larger share remains inside the category through another subscription, a roaster, or another recoverable destination. Read the coffee churn map
That difference is why a generic churn benchmark is of limited use. The structure of the population determines where customers can move and which interventions still have leverage.
readouts
Difference Machine is the instrument. The report depends on the question being asked.
Where is the population already moving? Track how pressure changes the distribution of behavioral states before the final outcome becomes visible. Example: Netflix
Which pressure sits closest to a decision to leave? Separate background friction from the pressure most closely associated with actual switching language. Example: Porsche
Where does demand go after the crossing? Map the direction of exit and identify whether the customer remains available to the category. Examples: coffee and supplements
What happens when the conditions change? Apply a named pressure to the population model and observe where transitions accelerate. See the method
Difference Machine does not begin with a universal list of complaints or a fixed churn score. The friction structure comes from the population being studied. The model then reads how those pressures relate to behavioral transitions within that population. That is why the outputs differ.
A useful model should preserve those differences.
working arguments, published openly
How we think about behavioral systems. The stress tests show the output. The field notes explain the measurement choices, assumptions and behavioral ideas underneath the instrument.
A complaint count grows with the size of the corpus. Density tells you how much of the population is carrying the friction.
A departure contains more information than a negative score. Preserving the full path of the exit turns individual statements into a defection map.
Some friction remains relevant for years. Other friction belongs to a market regime that has already passed. The model gives those signals different clocks.
A single brand often lacks enough structure to define the behavioral map around itself. Build the map from the category, then measure the brand inside it.
Historical similarity can identify customers who resemble previous churners. A decision requires a counterfactual question about what happens when the conditions change.
A linear threshold can miss a process that accelerates near a behavioral boundary.
Leaving develops through a transition. The useful intervention window sits while movement is still reversible.
Model population response to a named pressure and keep the limits of the prediction visible.
guide
The field notes examine individual parts of the problem. The churn guide connects them into one practical framework for reading population movement and deciding where intervention still has leverage.
coming next: new populations and new decisions. published runs appear here when the analysis is complete.
your population
Tell us which population matters and what you are considering changing. A price increase. A product change. A new subscription structure. A decision with enough consequence that you would rather see the population move here first. Public data is enough to start.