When breaking up your records into multiple rows by attribute - you then require multiple read units due to how DynamoDB charges you for each row read which costs a minimum of a read unit. So each complete record would have a cost for each attribute as if they were each sized at 4kb (minimum size of a read unit) on read and write.

You also very likely need to use DynamoDB transactions (not really mentioned in the article) or else you're taking big integrity risks with partially incoherent records and so you're paying more for their transaction bookkeeping as well.

Doesn't this seem to map badly to DynamoDB's cost model?