How Models End

Far more models end than expand. Endings are the normal condition of a testing enterprise, not a scandal, and understanding the ways a model concludes, and why, completes the picture of a model’s life. This lesson covers how models end and how to read those endings.

The ways a model ends

• Expiration. Most models are time-limited demonstrations with a defined end date. When the period ends, the model concludes unless renewed or expanded.
• Non-renewal. A model reaches its end date and CMMI chooses not to continue it, often because the evaluation did not show enough value.
• Cancellation. Less commonly, a model is ended early, by CMMI or under political direction, before its planned conclusion.
• Replacement. A model ends and a successor takes its place, refining the design. The ACO course’s example is exactly this: ACO REACH concludes after 2026 and is replaced by the LEAD Model in 2027.

Endings are the system working

The instinct to read every ending as failure is worth resisting.

Worth remembering: CMMI exists to test models, and testing means most experiments will not prove out. Ending a model that did not demonstrate value is the laboratory functioning as designed, not malfunctioning. A system that never ended a model would be a system that never learned. The honest, evidence-first view treats endings as information, this design did not deliver, or has served its purpose, rather than as embarrassment to be spun.

Reading an ending

That said, not all endings mean the same thing, and a careful reader distinguishes them:

• A model that ended after weak evaluations is evidence the design did not work.
• A model that expired and was replaced by a refined successor is iteration, the field learning and improving.
• A model cancelled early on political grounds is a policy shift, not necessarily a verdict on the model’s merits.

The reason matters. Conflating a political cancellation with an evidentiary failure, or vice versa, misreads what the ending actually tells you about whether the idea worked, exactly the discipline the evaluation course insisted on.

Key takeaways

• Models end through expiration, non-renewal, cancellation, or replacement by a successor.
• Endings are normal for a testing enterprise; concluding models that did not prove out is the system working, not failing.
• Distinguish the reasons: evidentiary failure, iteration, and political cancellation carry very different meanings.