There are relatively few publications that assess capacity decline in enough commercial cells to quantify
cell-to-cell variation, but those that do show a surprisingly wide variability. Capacity curves cross each
other often, a challenge for efforts to measure the state of health and predict the remaining useful life
(RUL) of individual cells. We analyze capacity fade statistics for 24 commercial pouch cells, providing an
estimate for the time to 5% failure. Our data indicate that RUL predictions based on remaining capacity or
internal resistance are accurate only once the cells have already sorted themselves into “better” and
“worse” ones. Analysis of our failure data, using maximum likelihood techniques, provide uniformly good
fits for a variety of definitions of failure with normal and with 2- and 3-parameter Weibull probability
density functions, but we argue against using a 3-parameter Weibull function for our data. pdf fitting
parameters appear to converge after about 15 failures, although business objectives should ultimately
determine whether data from a given number of batteries provides sufficient confidence to end lifecycle
testing. Increased efforts to make batteries with more consistent lifetimes should lead to improvements
in battery cost and safety.