can rarely be reproduced using the data+code in the journal archive. Recently created archives at top journals should avoid the mistakes of their predecessors. We categorize reasons for archives' failures and identify successful policies.