The processes that underlie consumer decisions to invest in the maintenance of a durable good over time are examined. The work centers on a hypotheses that consumers make decisions about whether to repair or replace a good that has suffered a decrease in performance through a process that assesses the value of repair actions relative to two points of reference: the normal rate at which the performance of goods declines as they age (age-indexing), and how the timing and cost of the repair compares to parallel norms for repair expenditures (expenditure indexing). We show how these heuristics can be represented by a cognitive algebra that models maintenance decisions as a series of myopic utility-maximization problems. This process yields outcomes that can approximate those that would emerge from an optimal dynamic maintenance policy in some cases, but significantly depart from optimality in others. The algebra is then used to generate a series of predictions about how maintenance decisions may depart from normative benchmarks that are tested in a dynamic computer-pet ownership simulation. Actual maintenance behavior is characterized by a number of biases that are consistent with theoretical predictions, including a seemingly contradictory tendency to undermaintain and prematurely replace goods of superior value when they were acquired, yet be overly reluctant to part with and over-maintain inferior goods. A discussion of the implications of the work for understanding real-world biases in product care and maintenance behavior is offered.