Data consumption plays an important role in Power Virtual Agent PVA licensing. It is important to keep a check on the data storage requirement of PVA implementations. You should know how the data is calculated because PA has special storage requirements for all of its conversations with usual configurations. You just need to keep a check of conversations of the chatbot and any other attachments which are uploaded in the conversation.
Moreover, you should also be familiar with the tables which store conversation transcripts. All the conversations are saved in the PVA bot which has conversationtranscripttable. It is for forecasting data storage that you need to be mindful of the conversation and expected conversation. However, it all depends on the complexity of the conversation.
You can refer to any existing implementation on the table and check its size; once assumed you need to validate it during the operations. Generally, a forecast is done for sizing and licensing costs of one year. It is multiplied above size per conversation with the total number of conversations which are expected in a year. Once you have the number you can calculate the forecasted size of the dataverse that you need for implementation.
In addition to the conversation transcript, you need to check the requirement in which you are expecting users to upload attachments. You also should keep a check of plugin trace logs, audit logs and data consumption. While calculating the size, you should also consider that the data retention policy is an important requirement. There is one bulk deletion job which runs every day and deletes conversations every day too. This is helpful because it optimizes storage requirements.