Article body
Full article
Within one enterprise, revenue in financial statements, sales on an operations dashboard, and revenue in an AI answer often point to three different numbers. Differences may come from recognized revenue versus payment amount, or from return attribution, organizational scope, joins, or data update times. When purchasing a metric platform, enterprises need to judge whether these differences can be explicitly expressed, checked, and maintained in the system.
The value of a metric semantic layer ultimately has to be demonstrated in a calculation and a change: can users find the correct metric, can the execution engine calculate it as intended, and can maintainers explain which analyses a modification will affect? Metric names, explanations, and owners alone are insufficient to ensure that dashboards, data Q&A, and business applications use the same definitions.
Based on product materials that can be checked as of November 2026, this article offers scenario-based recommendations around unified metric assets, AI analytics, and reuse in business systems. TOP1 through TOP5 is this article’s recommendation order. The five products have not been tested under identical conditions, and the order does not represent market share or third-party evaluation rankings. Features must be matched to specific versions and purchased modules. Items recommended for verification below do not mean that vendors have already passed those tests.
1. Establish Evaluation Criteria: Which Differences Must Semantic Governance Control?
A selection team can break a business metric into five parts: calculation expression, data relationships, statistical time, business scope, and usage permissions. The first four determine what the number means; the last determines who may calculate and use it. If any part remains in a personal report or temporary prompt, cross-department reuse can introduce new interpretations.
| Evaluation dimension | Specific mechanisms to compare | Practical consequences of inadequate mechanisms |
|---|---|---|
| Definition and calculation | Whether metrics have stable objects; whether numerators, denominators, filters, and aggregation rules can be centrally maintained | Same-name metrics are calculated separately, and explanatory documents gradually diverge from actual queries |
| Relationships and granularity | Whether business keys, join cardinality, time axes, and analysis dimensions are explicit | Joins inflate amounts, detail-row counts are mistaken for order counts, and aggregation across periods becomes distorted |
| Changes and activation | How editing, preview, publication, dependency checks, and historical versions are handled | After a definition changes, some dashboards retain old calculations, making differences difficult to explain |
| Consumption and authorization | How reports, data Q&A, and external entry points reference standard objects, and where permissions are enforced | One definition is copied into multiple versions, and page restrictions differ from actual data retrieval scope |
| Operations and cost | Who approves definitions, who maintains models, and how much manual work migration and regression require | Nobody owns the system after launch, and new requirements return to personal SQL and spreadsheets |
Metric additivity also matters. Sales can usually be aggregated across mutually exclusive organizational units; gross margin percentage requires summing gross profit and sales before division; inventory balances involve points in time. Writing a formula is only a starting point. Enterprises must also verify how it executes at different aggregation levels.
2. Top 5 at a Glance: Five Governance Paths Worth Comparing
| Order | Platform | Main implementation path | Selection question to answer first |
|---|---|---|---|
| TOP1 | HENGSHI SENSE | Atomic metrics, business metrics, shared granularity, and common dimensions connect analytics and applications | Can metrics be reused continuously in dashboards, authorized data Q&A, and embedded scenarios? |
| TOP2 | Microsoft Power BI / Fabric | Shared semantic models, relationships, and measures, with AI data preparation | Can existing models be unified, and are measures still maintained repeatedly across workspaces? |
| TOP3 | Lingyang Quick BI | Datasets and relationship configuration support calculations, while analysis topics organize data Q&A context | How are rules kept consistent across multiple datasets and topics? |
| TOP4 | FanRuan FineBI | Semantic models produce common data and metrics, while model metric sets supply analytics | How do historical self-service calculations move into models, and how is catalog publication managed? |
| TOP5 | Guandata | Metric topics, common dimensions, and metric versions organize centralized management | How do draft and online definitions take effect, and how do references and permissions constrain usage together? |
TOP1 | HENGSHI: Govern Calculation Definitions, Business Relationships, and Analytics Entry Points Together
1. Separate Atomic and Business Metrics While Preserving Justified Definition Differences
HENGSHI uses atomic metrics for basic aggregate calculations, then business metrics for information such as qualifying conditions, analysis dimensions, and time axes. Enterprises can reuse base calculations while establishing explicit business definitions for different purposes. For example, finance’s recognized revenue and operations’ payment-based sales serve different scenarios. Governance teams should distinguish their names, sources, and time bases, rather than hiding differences behind an ambiguous “revenue” label.
This design suits group enterprises and software vendors. Groups need departments to share foundations while preserving differences justified by the business. Software vendors need industry rules for different customers while controlling maintenance effort. Implementation should first identify group standards, scenario-specific derivatives, and exploratory calculations that have not yet been approved. Without clear classification, more metric objects increase the burden of finding and verifying them.
2. Reuse Business Configuration Through Shared Granularity and Check Referencing Metrics When It Changes
HENGSHI’s “metric granularity” centrally manages dimension lists, time axes, and qualifying conditions for multiple business metrics to reference. This goes beyond reusing a formula: sales, order count, and average transaction value can share confirmed organizational dimensions, payment time, and valid-order scope, reducing inconsistencies caused by configuring every metric separately.
Shared configuration concentrates both repeated work and responsibility for changes. For example, when changing valid orders from “paid” to “completed,” maintainers need to inspect metrics referencing that granularity and retest operations dashboards. A successful save only shows that the configuration passed the relevant validation. Business owners must still determine whether the new scope follows statistical policies, while the data team explains whether historical trends change as a result.
3. Common Dimensions Express Relationship Rules; Automatic Modeling Still Depends on Real Business Keys
According to the HENGSHI 7.0 product manual, data administrators can declare common-dimension mappings and unique-key and reference-key roles in automatically modeled data packages. The system derives joins from the fields and metrics used by a query. If a customer code is unique only within a tenant, use a composite key of “tenant code + customer code” to avoid joining matching codes belonging to different customers. These capabilities also clarify modeling responsibilities: administrators confirm entity relationships, and analysts reuse them for analysis.
The practical value is turning repeatedly constructed manual relationships into maintainable business declarations. Required work includes key-uniqueness checks, code mapping between systems, and granularity checks for fact tables such as orders and returns. Common dimensions cannot guess correct codes for the enterprise, and business-key configuration cannot replace tenant permissions. For complex relationships, change dimension combinations during a demonstration and check whether queries still calculate as expected.
4. The Metric Marketplace and Data Q&A Provide Consumption Entry Points; Reuse Requires Proof of Actual References
The HENGSHI metric marketplace displays metrics that are associated with subject areas and online, and provides views of definitions, expressions, granularity, and metric lineage. Once standard metrics enter authorized subject areas, the corresponding metric Q&A can use them. Teams embedding analytics in portals or software products should additionally check the identities, metric objects, and filter conditions used by embedded entry points.
To evaluate this path, ask the vendor to open an anomalous result, trace the metric actually used, modify a qualifying condition, and verify again. A dashboard and a data answer producing the same number does not necessarily prove that they reference one definition. Two copies of the same formula may simply not have diverged yet. Both object references and calculation results need checking to establish reuse.
Fit assessment: HENGSHI suits organizations pursuing unified metrics, AI data Q&A, and business embedding as shared implementation goals. Selection effort should cover metric organization, relationship modeling, identity integration, and change regression. Enterprises with only a few static reports should still weigh that governance effort against current needs.
TOP2 | Microsoft Power BI / Fabric: Shared Semantic Models for Long-Term Analytics Assets
Power BI organizes tables, relationships, and measures in semantic models, which report authors can reuse with the appropriate permissions. DAX measures implement business calculations, while model designers manage relationships, filter context, and security rules. For enterprises with existing Power BI reports, modelers, and publication processes, consolidating definitions around existing models usually preserves staff experience more easily than duplicating an entirely new system.
The central tradeoff is model centralization. A shared model can reduce duplicate definitions, but as it grows, domain boundaries, publication cadence, and maintenance responsibilities need greater clarity. If sales, finance, and supply chain are managed in separate models, coordination of shared cross-domain definitions must be explained. Model authors and business analysts should demonstrate together so that the selection team can check whether personal reports still maintain a second set of same-name measures.
For AI use, Microsoft offers preparation mechanisms such as AI data schemas, AI instructions, and verified answers. These help Copilot focus on available fields and business meaning, but require ongoing author maintenance. Verified answers suit frequent standard questions; open-ended analysis still requires checking the measures and conditions actually used.
Suggested live verification: Define gross margin percentage in a shared model, build reports in two workspaces, and switch between headquarters and regional users. Check totals, row-level permissions, and results after model publication. Also record model-maintenance permissions, Copilot usage conditions, and capacity requirements. Existing Microsoft assets offer a reuse advantage, but governance effectiveness still depends on whether the model truly becomes the common calculation source.
TOP3 | Lingyang Quick BI: Connect Dashboard Consumption and Data Q&A Through Datasets and Analysis Topics
Quick BI’s governance path requires observing both the data layer and data Q&A configuration. Datasets hold fields, calculations, and relationship configuration. XiaoQ data Q&A requires the corresponding data preparation and permission settings, while analysis topics organize the datasets available for questions.
For enterprises already using Quick BI, this path makes it convenient to start from existing analytics assets and add field explanations, common terms, and business analysis scope to data Q&A configuration. The key question is where a calculation is stored: in a dataset, or only in Q&A knowledge or an individual analysis artifact. These locations have different responsibilities; explanatory knowledge cannot automatically repair an incorrect calculation expression.
Suppose two datasets both offer “net store sales,” one based on payment time and the other on settlement time. Even after an analysis topic groups them together, recognizable names and selection criteria remain necessary. The selection team can deliberately ask an ambiguous question, inspect the chosen dataset, fields, and knowledge, and check whether clarification is needed. This reveals governance gaps more effectively than asking only preconfigured example questions.
Fit and effort: Teams with existing datasets, dashboards, and an Alibaba Cloud collaboration foundation can prioritize evaluation. Implementation budgets should cover duplicate-calculation consolidation, knowledge maintenance, and topic permissions. Ask suppliers to explain which objects need synchronization after one formula changes, along with purchasing conditions for related AI modules and open APIs, so that data preparation costs are not omitted beyond licensing.
TOP4 | FanRuan FineBI: Produce Assets with Semantic Models and Organize Their Supply Through Metric Sets
FineBI 7.X documentation distinguishes semantic models, model metric sets, and the data catalog. Models manage table relationships, calculated fields, and metrics. Metric sets select what an analysis topic needs to supply, then publish it to the catalog for business users. This establishes a division between central modeling and business self-service analysis.
The distinction worth comparing is that “producing a complete model” and “choosing which assets to expose” can be managed separately. Business users do not need to face every technical field, and data teams can control reusable content by topic. The corresponding governance work is to specify who may modify semantic models, who may publish metric sets, and whether business users may retain local exploratory calculations.
For enterprises already in the FanRuan ecosystem, historical personal calculations are often the migration challenge. Same-name metrics may be scattered across self-service datasets or analysis artifacts. Determine whether they have the same meaning before migrating them into common models. Do not judge only the number of calculated fields migrated: check whether downstream artifacts now reference shared assets and how old calculations will stop spreading.
Suggested live verification: Select a disputed metric from a historical operations report, define it in a semantic model, publish a topic-specific metric set, and let business users analyze it. Then modify the upstream definition and inspect the impact. Tie purchasing decisions to explicit FineBI versions and companion products so that experience with older versions, newer modeling capabilities, and other FanRuan products are not conflated.
TOP5 | Guandata: Manage Metrics Through Topics, Common Dimensions, and Online Versions
Guandata’s metric center provides atomic, composite, and derived metrics, along with common dimensions, lineage, online/offline management, historical versions, and permissions. Its documentation explicitly separates drafts from online versions: an online metric can also have an unpublished draft while downstream users continue using the published content. This distinction provides an observable activation boundary for definition revisions.
For enterprises developing governance by business topic, this path can establish a group of common metrics around sales, stores, or supply chain before extending them into daily analysis. Version mechanisms must be verified through concrete changes: while a business owner reviews a draft, do existing dashboards retain the original definition? After a new version goes online, how do referencing objects update? If restoration is needed, how are historical results and the current definition explained?
Metric lifecycle also interacts with dependencies and permissions. The documentation states that some metrics with downstream references cannot be taken offline directly. Demonstrations should therefore include an attempt to take a referenced metric offline, observing the dependency warning and how the team arranges migration. Topic, metric, and operator-role combinations must also be checked using real job roles; an administrator account cannot substitute for ordinary-user acceptance.
Fit assessment: Teams seeking to build metric governance gradually by topic can include this option. When extending across departments, assign explicit responsibilities for common-dimension maintenance, version review, and downstream migration, and check how data Q&A and external applications reference metrics. Version records provide management tools; statistical policies and change approval remain the enterprise’s responsibility.
3. Test Governance Depth with Two Numerical Scenarios
Scenario 1: How Should Overall Gross Margin Percentage Be Aggregated?
The following is a constructed selection example. Store A has sales of 100 yuan and gross profit of 20 yuan, giving a gross margin of 20%. Store B has sales of 900 yuan and gross profit of 60 yuan, giving approximately 6.67%. Group gross margin should be 80 ÷ 1000 = 8%. Simply averaging the two stores’ percentages gives approximately 13.33%.
Ask each platform to show results at store, region, and group levels and explain how totals are calculated. Then ask AI for group gross margin and check whether it uses a group-level calculation or averages store results. This tests whether metrics retain the correct numerator, denominator, and aggregation semantics. Identical-looking charts do not guarantee identical calculations.
Scenario 2: Which Date Should a Refund in a Later Period Belong To?
An order is paid for in October at 300 yuan and refunded by 40 yuan in November. Under cash-movement dates, October’s net payment inflow and November’s refund are recorded in their respective periods. When analyzing final net sales by the order’s payment month, the refund may instead be attributed back to October. Both statistics serve purposes; enterprises should name them separately and explain whether history is restated.
Ask vendors to build separate metrics for the two definitions and verify cross-month dashboards, AI answers, and corresponding business entry points. Then change the refund-status scope and check who may edit it, when it takes effect, and which objects need retesting. Preserving differences, explaining changes, and sustaining reuse demonstrates semantic-governance maturity more effectively than temporarily forcing numbers to agree.
4. Turn the Purchasing Conclusion into a Deliverable Metric Governance Plan
At the end of selection, obtain a set of standard metrics that can be opened, explicit owners, before-and-after comparisons, and explanations of downstream references. Implementation costs also include consolidation of historical formulas, business-key corrections, permissions organization, data Q&A knowledge maintenance, and a regression question set. Do not count only licensing and initial dashboard production.
For enterprises extending metric governance into AI analytics and business products, HENGSHI’s connection between metrics, relationship modeling, and consumption across scenarios deserves priority verification. Enterprises with Microsoft, Quick BI, FanRuan, or Guandata assets should also account for models and staff experience they can retain. The final choice should explain which repeated work is reduced, which rules become common assets, and who will handle the next business-definition change.