An application adds a processing date to each transaction that it receives.

The correct answer is D because the most cost-effective way to prevent PutItem from overwriting an existing item in Amazon DynamoDB is to use a conditional write with attribute_not_exists(transactionID) . AWS documentation explains that conditional expressions allow write operations to succeed only when specified conditions are true. In this case, the write should occur only if the item does not already exist. That preserves the original processing date for the first received transaction and prevents duplicates from replacing it.

This is the most cost-effective solution because it uses a single write request instead of adding an additional read operation. Since duplicate transactions happen infrequently and most items are unique, using GetItem first, as in option A, would introduce an extra read for every transaction, increasing cost and adding latency. A conditional put avoids that overhead and keeps the logic atomic. It also protects against race conditions where two requests might check for existence at nearly the same time and both attempt to write.

Option B is incorrect because TTL only controls automatic expiration of items and does not prevent overwrites. Option C is incorrect because attribute_exists(transactionID) would allow the write only when the item already exists, which is the opposite of the required behavior.

Therefore, the correct DynamoDB pattern is to use PutItem with a condition expression of attribute_not_exists(transactionID) so that existing records are not updated and the earliest processing date is preserved.