You are here:
Billing Considerations for Enriched Index
Enriched indexing impacts the consumption of credits used for billing for orgs operating Data 360 under a Data Coud license. With enriched search indexes, you are processing unstructured data and making calls to a LLM to generate enriched chunks.
When you create a search index configuration with unstructured data, your data is stored and processed in Data 360. There are three billing components for unstructured data processing and indexing in Data 360: data storage, data processing, and data queries. Each component has a distinct applicable usage type.
In addition to unstructured data billing implications, enriched indexes also consume flex credits.
This feature has access to Digital Wallet, a free account management tool that offers near real-time consumption data for enabled products across your active contracts. Access Digital Wallet and start tracking your org's usage. To learn more, see About Digital Wallet.
| Digital Wallet Card | Usage Type | Usage Type Description | Notes |
|---|---|---|---|
| Data Services | Unstructured Data Processed | Usage is calculated based on the amount of unstructured data that is processed. For example, if the search index processes 100 PDF documents that are 1 MB each, usage is calculated as 100 MB. If the search index processes five audio/video files that are on average 100MB each, usage is calculated as 500MB. In Data 360, unstructured data may be chunked and vectorized using an embedding model. Usage is computed only once across both these activities. For example, if one 100 MB PDF document is chunked and vectorized, usage is computed as 100 MB, not as 200MB. |
The cost of creating a search index remains the same for vector search and hybrid search. If a video file of 1GB is transcribed, chunked, and vectorized, usage is computed as 1GB. A DMO and all its file attachments are treated as a single unit for processing purposes. For example, if Data 360 processes incremental changes either to the fields on a source DMO or to a file attachment on that DMO, all file attachments are reindexed. |
| Data Services | Data Queries | Usage is calculated based on the number of records processed. The count of records processed depends on the structure of a query as well as other related factors such as the total number of records in the objects being queried. |
For vector search queries against unstructured data, the number of vectors in the search index are counted. For hybrid search queries against unstructured data, the number of vectors and keyword records in the search index are counted. In a typical search index, the number of keyword records is the same as the number of vectors. |
| Data Storage | Storage Beyond Allocation | Usage is calculated based on the amount of storage used above the amount allocated. | Every file ingested and table created, including unstructured data lake objects (UDLO), unstructured data model objects (UDMO), CDMO, or index data model objects (index DMO), count toward Data 360 data storage, including the following.
|
| Flex Credits | Standard Prompts Basic Prompts Advanced Prompts |
Usage is calculated based on two factors: the number of direct requests to the LLM via the LLM gateway, and whether the gateway uses a Salesforce managed large language model. The specific category depends on the model that is used. See Large Language Model Support to find out which usage types apply. All Standard, Basic, and Advanced prompts process up to 2,000 tokens per prompt. Token usage is rounded up in 2,000-token increments. All Standard, Basic, and Advanced prompts that exceed this limit will be metered as multiple prompts, with each additional 2,000-token chunk counting as a new prompt. For example, a prompt with a total of 6,500 input and output tokens will be metered as 4 prompts. Tokens are units of data processed by the AI models. |
With enriched indexes, calls to the LLM to generate enriched chunks are counted as Standard Prompts. For more information see, Flex Credits Billable Usage Types. |
Unstructured data and search index are not available for orgs operating Data 360 under the Customer Data Platform (CDP) license. For more information on how Data Cloud usage is billed, refer to your contract or contact your account executive.
Understand Enriched Index Costs
Enriched indexing enhances chunks with generated intelligence, such as entities, summaries, topics, and answerable questions, to improve intent-driven retrieval. Although this process increases discoverability and grounding for AI assistants, it raises indexing costs because every chunk produces additional embeddings and metadata chunks.
The number of chunks generated from a dataset depends on factors such as the chunking strategy (passage extraction, section-aware) and chunk size. Together, these variables determine how efficiently content is chunked and the cost of enriched indexing.
Example : Credit Consumption Calculation
Consider a sample data of 5 MB consisting of PDF documents such as product catalogs or technical guides. Using a structure-aware chunking strategy with a maximum chunk size of 512 tokens, assume Data 360 search chunking generates approximately 16 chunks per MB. Let’s calculate the total LLM requests that were made to generate the enriched chunks and the flex credits consumed.
Total Chunks = 5 MB x 16 chunks per MB = 80 chunks
Each LLM request processes 4 chunks.
Total LLM Requests = 80 chunks ÷ 4 chunks per LLM request = 20 requests
Each LLM request includes token usage. Let's calculate the total tokens used per request.
- Prompt Instructions take 2,100 tokens
- Tokens for chunks processed = 4 chunks x 512 tokens = 2,048 tokens
- Total output tokens = 1,500 tokens
Total Tokens Used per Request = 2,100 + 2,048 + 1,500 = 5,648 tokens
Assume 2000 tokens is one unit of Salesforce Flex Credit, and billing rounds up to the nearest unit. Number of billing units required for the 5 MB data set is:
- 5,648 tokens ÷ 2,000 tokens per unit = 2.824 billing units
- 2.824 billing units rounds up to 3 billing units.
Assume each billing unit consumes 4 flex credits for a Standard Prompt, in this case, each LLM request consumes:
3 Billing Units x 4 Flex Credits per unit = 12 Flex Credits per LLM request
So, for 20 LLM requests: Total Flex Credits used = 20 requests x 12 Flex Credits per request = 240 Flex Credits
Alternatively, if the dataset consisted of files producing more chunks per MB, the enrichment workload increases significantly. For the same 5 MB data set, let's say a structure-aware chunking produces 121 chunks per MB.
Total Chunks = 5 MB x 121 chunks per MB = 605 chunks
Each LLM request processes 4 chunks.
Total LLM Requests = 605 chunks ÷ 4 chunks per LLM request = 152 requests
Let’s calculate the total flex credits consumed for 152 LLM requests based on how total tokens used per request, total billing units, and flex credits consumed per billing unit are calculated.
Total Tokens Used per Request = 2,100 + 2,048 + 1,500 = 5,648 tokens
Total Billing Units = 5,648 tokens ÷ 2,000 tokens per unit = 2.824 units
2.824 billing units rounds up to 3 billing units.
Total Flex Credits Consumed for 3 billing units = 3 x 4 credits per unit = 12 Flex Credits per LLM request
Total Flex Credits consumed for 152 LLM requests = 152 requests x 12 Flex Credits per request = 1,824 Flex Credits
These sample calculations show how the number of chunks generated affect the number of LLM requests made which ultimately impacts the flex credits consumed.
