Calculated Insights are the SQL-defined aggregations that power most of what business users actually do with Data Cloud. They calculate things like "lifetime spend per customer," "engagement score over the last 30 days," and "predicted churn probability." They are also the single most common source of unexpected credit consumption in Data Cloud deployments.
The cost structure of Calculated Insights is straightforward in theory and surprising in practice. Each insight runs on a defined refresh schedule; each run processes some volume of records and computes some amount of result data; both processing and result storage accrue credits. The interaction between refresh frequency, record volume, and computation complexity produces credit consumption patterns that the proposal-stage estimates rarely capture accurately.
How Calculated Insights consume credits
Each Calculated Insight is defined as a SQL query against the Data Cloud lakehouse, configured with a refresh schedule. When the insight refreshes, three credit categories accrue. Compute credits accrue based on the data volume processed by the query—roughly proportional to the size of the underlying tables joined and filtered. Storage credits accrue based on the size of the resulting insight table, which persists in the lakehouse. Activation credits accrue separately if the insight is published to a downstream system like Marketing Cloud or a CRM object.
| Refresh frequency | Annual runs (1 insight) | Relative credit cost | When appropriate |
|---|---|---|---|
| Every 15 minutes | ~35,000 | 100x | Real-time scoring needs only |
| Hourly | ~8,760 | 25x | Hourly operational dashboards |
| Every 4 hours | ~2,190 | 6x | Intra-day reporting |
| Daily | 365 | 1x (baseline) | Most analytical use cases |
| Weekly | 52 | 0.15x | Cohort analysis, periodic reports |
The credit multiplier across refresh frequencies is roughly proportional to the run count, but it is not strictly linear. Fifteen-minute refreshes incur additional overhead from scheduler invocation, partial result materialization, and downstream pipeline triggering. Hourly is often the operational sweet spot for use cases that genuinely require intraday freshness.
The five most common over-consumption patterns
1. Real-time refresh on insights that are not used in real time
The most common over-consumption pattern is real-time or near-real-time refresh on insights consumed by daily or weekly reports. An insight refreshed every fifteen minutes that powers a Tuesday-morning leadership dashboard is consuming roughly 100x the credits that a daily refresh would consume—for value that is not differentiated by the refresh frequency. A periodic audit of refresh frequency against downstream consumption is the single highest-yield optimization in most environments.
2. Insights duplicated across teams
Without governance, marketing, sales operations, customer success, and analytics teams each build their own versions of the same logical insight. We routinely see three to five variants of "engagement score" or "customer lifetime value" in a single tenant, each refreshing on its own schedule and consuming its own credits. Consolidating into a canonical insight, with team-specific views built on top, typically eliminates 20-40% of insight-related credit consumption.
3. Joins that pull more data than needed
Calculated Insights are written in SQL, and SQL written without query-cost awareness frequently joins large tables when smaller dimensional tables would suffice. The compute credit cost of a query is roughly proportional to the data processed. A SQL review of the top-twenty insights by credit consumption typically identifies meaningful optimization opportunity in roughly half of them.
4. Insights with no consumer
Insights are easy to create and easy to forget. The catalog accretes insights that were built for a one-time project, served their purpose, and have been refreshing on schedule ever since. A quarterly review of insight consumption—who reads it, what dashboard or activation depends on it—typically allows pruning 15-25% of the catalog with no operational impact.
5. Refresh frequency drift
An insight created with a daily refresh frequency, when first promoted to a production dashboard, gets bumped to hourly to satisfy a stakeholder request. The stakeholder forgets about it. The insight continues to refresh hourly for years. Refresh frequency tends to drift upward and rarely downward without explicit review.
The negotiation implications
Calculated Insights matter at the negotiation table for two reasons. First, they are a significant component of Data Cloud credit consumption, and the negotiated credit pool needs to be sized realistically against the expected insight footprint. Over-sizing creates committed shelfware; under-sizing creates uncomfortable overage conversations. Second, the optimization potential in the insight catalog—the 40-60% reductions described above—creates real leverage at renewal. A customer who can document insight optimization wins and a defensible projected steady-state consumption is in a meaningfully stronger position than a customer who is renewing on a flat extrapolation of last year's burn.
Insist on a credit consumption report
The order form should obligate Salesforce to produce a quarterly credit consumption report broken down by category—ingestion, identity resolution, calculated insights, activation, storage. Without this report, the customer is forced to rely on the Data Cloud Setup UI for consumption visibility, which is functional but not optimized for governance reporting. The contractual obligation to produce the report creates a forcing function on both sides to track consumption deliberately.
Negotiate a true-down right
The optimization potential in Calculated Insights means that a customer's actual steady-state consumption is often dramatically lower than the initial commitment. Without a true-down right, that gap becomes shelfware. With a true-down right, it becomes a real reduction in the renewal commitment. The true-down right is the single most valuable clause in any consumption-priced Salesforce contract, and Data Cloud is the SKU where it produces the most savings.
An operational playbook
The Calculated Insights catalogs with the healthiest economics share three operational practices. A quarterly insight audit reviews each insight's refresh frequency against its actual downstream consumption, with explicit owner sign-off on any refresh frequency more frequent than daily. A monthly credit consumption report ranks the top-twenty insights by credits consumed, with SQL review on any insight that has shifted significantly. And a governance gate requires any new insight to declare its expected refresh frequency and downstream consumer before it is created. None of these practices are technically demanding; they are governance disciplines that most customers do not apply to consumption-priced infrastructure until the budget pain becomes unignorable.
The customers we advise who apply these disciplines from the beginning of a Data Cloud deployment achieve the 40-60% optimization gains within the first quarter. The customers who wait until the budget pain hits typically achieve the same gains, but only after explaining a credit overage to their CFO. The disciplines are the same; the cost of waiting is real.
The streaming-vs-batch decision
Calculated Insights can be configured to run in batch mode (scheduled refreshes) or in streaming mode (continuous incremental calculation). The choice has significant cost implications. Streaming mode delivers near-real-time freshness but consumes credits continuously; batch mode delivers periodic freshness at much lower steady-state cost.
The right choice depends on the downstream consumption pattern. An insight consumed by a real-time scoring API benefits from streaming. An insight consumed by a Tuesday-morning leadership dashboard does not. Most insights, in practice, fall into the second category, but many are configured for streaming because the option exists. A periodic audit of streaming-mode insights against actual freshness requirements consistently identifies opportunities to revert to batch with no operational impact.
The aggregation-level question
Each Calculated Insight is defined at a specific aggregation level—per customer, per account, per region, per product, per channel. The aggregation level affects both the computation cost (more granular aggregations are more expensive to compute) and the storage cost (more granular aggregations produce larger result tables).
A common pattern in immature deployments is to define insights at the most granular aggregation level possible, on the theory that downstream consumers can roll up as needed. The compute and storage cost of this approach can be substantial. A right-sized approach defines each insight at the aggregation level its consumers actually need, with separate insights defined at coarser levels for use cases that need aggregation. The right-sizing typically reduces total insight credit consumption by 15-25% with no loss in analytical capability.
The hierarchical insight pattern
Mature Calculated Insights catalogs frequently use a hierarchical pattern: foundational insights computed at fine granularity, and rollup insights computed from the foundational insights at coarser granularity. The pattern reduces total compute cost because the foundational insights are computed once and reused across rollups, rather than each rollup independently scanning the underlying data.
The pattern requires deliberate design and ongoing governance. Without governance, the foundational and rollup insights drift apart; with governance, the pattern produces meaningful cost savings while improving the consistency of insight semantics across the organization.
The insight lifecycle
Calculated Insights have a lifecycle—they are created for a specific business need, mature into operational dependencies, and eventually become candidates for retirement when the business need changes or the use case migrates. Most organizations manage the create-and-mature phases but neglect the retire phase. The result is an insights catalog that accretes over time without ever shrinking.
A quarterly insight retirement review—identifying insights that have not been read in the last 90 days, validating that no downstream activation or dashboard depends on them, and retiring them—is a low-effort, high-yield governance practice. The customers we advise who institutionalize this practice maintain insight catalogs that stay near steady-state size, with the credit consumption that comes with that discipline.
The relationship to Data Cloud's broader credit pool
Calculated Insights credits are drawn from the same overall Data Cloud credit pool as ingestion, harmonization, and activation. The total credit consumption is the sum across all four categories, and the negotiated commitment should reflect the steady-state expectation across all of them rather than the maximum of any one.
This fungibility creates real flexibility for the buyer. A deployment that aggressively optimizes harmonization may have credit headroom for additional Calculated Insights; a deployment that adds new activation destinations may need to compensate by retiring lower-value insights. The right contractual structure allows this fluidity—a single committed credit pool with explicit allocation guidance, rather than separate sub-pools that constrain each category independently.
Where to position Calculated Insights in the broader negotiation
Calculated Insights are best negotiated as part of the broader Data Cloud commercial conversation rather than as a separate line item. The right framing is the total expected steady-state credit consumption across all four categories—ingestion, harmonization, calculated insights, activation—with the negotiated capacity sized against that total plus a defined growth buffer. The contractual obligations on Salesforce should include a quarterly usage report broken down by category, so the customer can see how each category is consuming credits over time and rebalance the operational mix accordingly.
Customers who negotiate Calculated Insights as part of this integrated view consistently achieve better outcomes than customers who treat each Data Cloud component separately. The aggregate negotiation produces volume leverage; the category breakdown enables ongoing optimization; the fungibility creates operational flexibility. The combination is what separates the top-quartile Data Cloud contracts from the median.