How Semantic Layers Unify Metric Definitions Across Queries

01: Regional Split vs. Refund Month Change Are Two Different Operations

"Split last month's net sales by region." This follow-up only changes the viewing angle, assuming the metric model already defines region — for example, the major region where the transacting store belongs . The national total and regional breakdowns may be computed by different SQL queries, but both can use the same order scope, refund rules, and statistical timestamp. "Region" cannot be arbitrarily picked from a table column. Viewing by transacting store's region asks how much business each major region did; viewing by shipping address asks where the business originated. Both dimensions are useful, and reports should state which is used.

"Refunds deducted by occurrence month" changes the metric itself. Pushing refunds back to the original order month suits reviewing how much of that month's orders ultimately remained. Deducting by refund occurrence month suits observing the current period's refund impact on sales performance. Both calibers can coexist with different audiences. The problem arises when an Agent mixes the second algorithm into the first dashboard while keeping the title, annotations, and historical comparisons unchanged. If October's extra 20,000 yuan appears in an auto-generated business summary, a single caliber adjustment could be misread as business growth.

This is why dashboard editing and metric changes must be handled separately . Business users can freely swap dimensions or shorten time ranges. When calculation rules truly change, the new caliber's name must be visible, and its effective date clearly stated. Otherwise, every new dashboard adds another metric definition that the data team must manually verify.

02: Claude Certifies a Query, but the Caliber Can Still Change Upstream

Claude Data generates queries for charts; viewers can inspect the actual SQL and last refresh time. Teams can certify approved queries and attach usage instructions for Claude. Editing a query invalidates its prior certification. Users with write permission on the Data artifact can initiate certification. Teams managing certified queries via Git can use GitHub PRs for review. For an organization, whether dashboard write permission should also include the authority to publish official calibers is a separate governance decision. Claude Data help docs: https://docs.anthropic.com/en/docs/claude-data

Certification has an easily overlooked boundary. Suppose a dashboard always reads a view named "Net Sales," and the data team swaps the refund logic inside that view. The chart SQL remains unchanged, yet re-execution follows the new rules. Claude's public docs explain the de-certification mechanism when the query is edited. They do not specify whether the certification badge updates automatically when an upstream view or metric definition changes. Troubleshooting such drift requires more than the chart SQL; you need to trace which metric or view versions it referenced. Claude Data help docs: https://docs.anthropic.com/en/docs/claude-data

The same metric version can yield different results on different days. Under the original-order-month refund caliber, a new refund posted in November for an October order reduces October's sales by 20,000 yuan. Here the data participating in the calculation changed, not the rule. Reviewing numbers in a monthly business meeting at minimum requires retrieving the rule used at that time and the data snapshot accessed. Teams needing to preserve month-end values can archive published results or retain snapshots sufficient for recomputation. This lets them distinguish between a rule adjustment and a late-arriving refund.

03: Semantic Layer Lets Different Queries Reuse the Same Metric Caliber

The same net sales metric: finance wants monthly totals, regional heads want store-level breakdowns, channel teams want source-based splits. If every chart stores its own full calculation logic, a rule change forces a manual hunt for which charts still use the old algorithm. The semantic layer centralizes these shared rules into the metric model , including which time field to use, at what granularity to aggregate, which dimensions are allowed for grouping, and what relationships join the data. Take dbt's MetricFlow as an example: an application submits a metric and a set of dimensions, and the semantic layer generates SQL from the model. The same net sales definition can serve national totals, regional comparisons, and store details. dbt semantic layer architecture: https://docs.getdbt.com/docs/build/semantic-layer

An order refunded in two installments illustrates why a metric name alone is insufficient. An order paid 1,000 yuan, first refunded 100 yuan, then 200 yuan. If a one-row-per-order table is joined directly to a one-row-per-refund table and both amounts are summed, the order amount gets counted twice. The SQL would output 1,700 yuan. For order-grained net amount, the two refunds must first be aggregated to the order grain (300 yuan). If refunds need to be tracked by occurrence month, the model must retain each refund's timestamp to allocate them to the correct month. Wrong granularity makes the numbers wrong even if the formula matches the finance definition.

When integrating a Data Agent, you can either feed the caliber documentation to the model for reading, or let it call the semantic query interface directly. In the first approach, which refund date field to use and how to join the order table still depend on the model-generated SQL. In the second, the Agent can first discover available metrics, their supported dimensions, then submit time range and grouping conditions, leaving the calculation to the metric model. dbt's MCP server provides such tools. Claude's official docs also state that Claude Data can use dbt, Looker, or Cube definitions via existing MCP connectors. During deployment, it's worth tracing a chart's execution log end-to-end to confirm which path the official metric query actually took. dbt MCP tools: https://github.com/dbt-labs/dbt-mcp · Claude Data help docs: https://docs.anthropic.com/en/docs/claude-data

Pay special attention to how the Agent handles missing dimensions. A user asks "split net sales by refund initiation channel," but the model only has order transaction channel. If the system substitutes a similarly named field , it may produce a nice-looking chart that answers a different question. Official dashboards can hold such requests for confirmation; ad-hoc analyses can continue exploring but should annotate the algorithm and dimensions used on the chart. If both refund attribution rules will be used long-term, they can be built as separately named metrics for business users to choose explicitly.

04: Four Requests to Validate an Agent-Generated Business Dashboard

Validation can start with a few orders whose expected results are known. Prepare one cross-month refund and one multi-installment refund, write down the expected outcomes, then run the same test data through the following four requests. Each time, verify the caliber shown on the chart, the actual metric hit, and the execution log. Looking only at the final number often fails to catch two different paths that coincidentally produce the same result.

Request: "Split October net sales by region." Validation: Confirm the split uses the same metric and refund rule as the total. Under the same data snapshot and filters, if regions are mutually exclusive and collectively exhaustive, the sum of this additive metric across regions should equal the total.

Request: "Change refunds to deduct by occurrence month." Validation: The chart must label the new caliber. In the example, East China's October figure returns from 80,000 to 100,000 yuan because the 20,000 yuan previously deducted from October moves to November.

Request: "Keep dashboard SQL unchanged, update upstream caliber." Validation: Check whether old results can trace back to the original definition, whether new results show the definition has been updated, and whether affected dashboards need re-validation.

Request: "Share the dashboard with a colleague who lacks the underlying connection." Validation: Using the recipient's account, confirm the visible results, the connection used by the query, and whether the viewer can alter the query scope.

The last item relates to report distribution. With Claude Data's real-time sharing enabled, colleagues inside the organization can see results fetched via the sharer's data connection, even without that connection themselves. Sharing covers only the queries required by the dashboard ; query changes require editor approval. Therefore, the publisher must verify not just numerical correctness but also the aggregation level. A regional total visible company-wide may need restricted distribution once drilled down to individual stores. Claude Data help docs: https://docs.anthropic.com/en/docs/claude-data

Next time that chart is opened in a business meeting, its most valuable property is comparability . Whoever asks, however many extra regional layers are added, everyone still knows which rule set "net sales" uses. If a 20,000 yuan refund truly moves from October to November, the meeting can reconcile that variance on the spot and proceed to discuss which caliber should drive decisions.

References

Anthropic. November 10, 2026 release: Claude Data general availability, certified queries and real-time sharing in beta. Introducing Claude Whiteboards, Claude Sheets, and Claude Data

Anthropic Help Center. Query certification, de-certification on edit, existing metric definition integration, real-time sharing query and permission scope. Get started with Claude Data

dbt Labs. Semantic layer components and how MetricFlow generates SQL from metric models. dbt Semantic Layer architecture

dbt Labs. Discovering metrics, querying available dimensions, and executing metric queries via MCP. Introducing the dbt MCP Server