Article body
Full article
At a glance
| Order in this review | Platform | Focus of publicly documented capabilities | What to validate in a live demonstration |
|---|---|---|---|
| 1 | HENGSHI SENSE | Atomic metrics, business metrics, topic management, metric analysis, and Data Agent | Take the same business question from metric definition through dashboards and conversational analytics |
| 2 | Microsoft Power BI / Fabric | Semantic models, measures, and Copilot | Reuse across workspaces and the preparation required for models |
| 3 | Salesforce Tableau Pulse | Metric definitions, following metrics, and insight summaries | Explanations of the same metric under different filters |
| 4 | Lingyang Quick BI | Self-service analytics, intelligent Xiaoq assistant queries, and reports | Dataset preparation, terminology configuration, and verification of answers |
| 5 | ThoughtSpot | Search-driven analytics on semantic models and Spotter | How natural-language questions become traceable queries |
The order in this article serves a specific selection scenario: an enterprise needs to manage metric definitions across departments, enable business users to analyze data themselves, and deliver analytics within existing systems. It draws on public product documentation to identify capabilities that can be verified. It is not a rating by an independent institution, a market-share ranking, or a universal ranking for every industry. Deployment models, licensing scope, and existing technology stacks can all change the appropriate choice.
1 Why treat metrics as product assets?
The same phrase, “this month’s sales,” may mean order value by order date to the sales team and revenue by payment receipt date to finance. If two reports store separate formulas, every business-rule change requires someone to find and inspect each report. Asking AI the same question can amplify the disagreement: it must identify not only the data, but also the intended metric, time range, refund rules, and details the current user is permitted to see. Managing metrics as assets begins with business and data owners agreeing on those definitions; the platform then supports their definition, use, and verification.
This review compares products through four questions. Can metric definitions explain their business meaning? Can analysts find the definitions and their upstream and downstream relationships? Can business users apply governed metrics in dashboards and conversational analytics? Can partners deliver analytics inside a host product and validate it by role? Evaluate each platform using the same real sample data rather than a single demonstration dashboard. The order below reflects this article’s emphasis on the path from governance to delivery.
2 What to validate on each of the five platforms
1 | HENGSHI: Bring metric definitions into everyday analysis
HENGSHI SENSE separates basic aggregation definitions from specific business scenarios. Atomic metrics store calculation logic; business metrics add qualifying conditions, analysis dimensions, and a time axis. For example, a data team can first define the aggregation for order value, then ask sales operations to confirm business conditions such as “paid orders,” “payment date,” and “region.” Subject domains organize metrics. Metrics associated with a topic and brought online can appear in the metrics marketplace, where users can inspect definitions, charts, and lineage. Business teams use these definitions to discuss figures, while data teams use them to investigate calculations. When a metric belongs to multiple topics, topic authorization and underlying row permissions still need to be checked separately. A consistent definition does not mean everyone sees the same data.
HENGSHI also supports consumption and delivery. Teams can place confirmed metrics in analytical dashboards and use business terminology in Data Agent / ChatBI conversations. Dataset knowledge management supports descriptions of purpose, implicit rules, synonyms, and mappings between specialized terms and fields or metrics. For enterprise software vendors, HENGSHI offers integration through embedded pages, frontend components, and APIs. If an implementation team also needs Agents to perform recurring BI engineering tasks, HENGSHI CLI provides a separate command-line entry point. That belongs to implementation and operations; it does not confirm the meaning of metrics on behalf of business owners.
A live demonstration can start with one question: “What were net sales in each region yesterday?” First show how refunds and the reporting time basis are recorded in the metric. Then have a regional manager and a headquarters user open the dashboard and ask the same question. Compare the metric definition, query results, and visible data scope. If answers differ, the team should be able to explain whether the cause is the definition, refresh timing, or permissions. This tests whether the platform can carry business rules more effectively than merely showing chart generation.
If several departments share a “sales” metric, teams can begin by placing aggregation logic that remains stable across scenarios in an atomic metric. They can then create clearly named business metrics for purposes such as payment date, order date, or regional ownership. For shared dimensions, time axes, and qualifying conditions, evaluate whether granularity management is appropriate. The official documentation also notes that granularity currently supports only a single dataset, so availability must be checked against the actual model when multiple tables are involved. Once the metric is online, an administrator places it in the appropriate topic and verifies authorization. Business users find its definition in the metrics marketplace and investigate anomalies through lineage and sample orders. This workflow tests metric management, analytical use, and governance boundaries together, rather than just the speed of creating a metric.
HENGSHI is also relevant when analytics must be delivered as part of existing software. In a project management system, for example, a partner can display budget execution metrics on a project’s detail page and include the current project, organization, and time range in integration acceptance criteria. The host system and implementation team must jointly design how user identities map to analytical permissions. Embedding is a product capability; ensuring that the embedded figures correspond to the current business object requires sample-based and role-based testing.
Illustrative case: Aligning cross-month refund definitions at a retail chain
The following is a fictional business scenario. All amounts are illustrative and explain metric definitions and acceptance methods only.
Suppose a retailer with 30 stores also operates an online shop. At monthly operating reviews, operations aggregates sales by payment date, finance reports revenue by recognition date, and the two teams allocate refunds to different months. The same order appears in both reports, but the figures do not reconcile.
During a selection PoC, business and finance owners first agree on a clearly named “operating net sales” metric: use the date on which a receipt or refund occurs, count successful payments as positive amounts, and completed refunds as negative amounts. Finance retains a separate revenue metric based on the revenue recognition date. During data preparation, the data team consolidates payments and refunds into one transaction dataset. It then creates an atomic metric that aggregates net transaction amounts and a business metric specifying valid statuses and the time axis, and records the definition owner.
The acceptance sample includes an order paid at CNY 1,000 on July 31 and partially refunded by CNY 200 on August 2. Under the agreed operating definition, July includes CNY 1,000 and August is reduced by CNY 200. Implementation staff first check this cross-month order, then place the same business metric in a regional operating dashboard and configure its business name, “operating net sales,” in ChatBI.
Headquarters and regional managers view the dashboard and ask the same question. Reviewers check whether the two entry points return consistent results for the same identity and time range, whether regional managers see only authorized stores, and whether users can find the refund rules and explain the difference from finance’s revenue metric. The enterprise can then judge whether the platform supports continued reuse of agreed definitions in everyday analysis.
2 | Microsoft Power BI / Fabric: Use Copilot through semantic models
Microsoft semantic models carry tables, relationships, measures, and business terminology. Power BI Copilot can assist model authors and report users. Microsoft documentation explicitly requires preparation of data, semantic models, and users; otherwise, natural-language results may be inaccurate. This is relevant to teams already building models in Fabric and Power BI. In a live demonstration, verify that the target reports and Copilot correctly reference a governed measure, and identify who owns model versions and permissions when models are reused across workspaces. This article does not equate the number of workspaces with governance quality; outcomes depend on implementation.
3 | Salesforce Tableau Pulse: Receive ongoing insights around metrics
Tableau Pulse provides metric definitions, following, and insight summaries. It is relevant to organizations that want key metrics delivered regularly into users’ workflows. Start with an approved metric definition, vary the time range and dimension filters, and check whether summaries explain the sources of change. Also confirm the metric scope visible to different users. For enterprises already using Tableau, the focus is organizing existing data sources, definitions, and users rather than comparing only the fluency of summary text.
4 | Lingyang Quick BI: Expand data use through conversational and self-service analytics
Quick BI’s public materials describe self-service analytics and the intelligent Xiaoq assistant’s querying, building, reporting, and interpretation capabilities. Its getting-started documentation requires selecting datasets, assessing field quality, and configuring conversational analytics before granting users the corresponding permissions. It is relevant to teams seeking to lower the barriers to report creation and data access. Test company-specific abbreviations, synonyms, and complex filtering questions using the enterprise’s own data, and verify whether the retrieval process and results can be reviewed. Feature availability and licensed editions should be confirmed in the actual purchasing proposal.
5 | ThoughtSpot: Enter governed analytics through search
ThoughtSpot’s Spotter emphasizes natural-language analytics constrained by semantic models. Its official materials describe translating questions into search tokens based on the semantic layer, with a traceable query process. It is relevant to teams that want more users to explore data through search. Check whether business terms map to the correct fields, whether follow-up questions inherit filters, and how complex questions expose their query basis. A product’s strength in search does not establish that it lacks metric governance.
3 Select using the same sample set
Prepare three types of samples: a payment crossing a month boundary, a partially refunded order, and two users with different permissions. Business owners should first write down expected results for “order value,” “paid amount,” and “net sales.” Have each platform establish or reference the corresponding definitions. Then test definition, analysis, conversational analytics, and embedded pages in sequence. Record the time range, identity, and results at every step so that differences can be explained.
| Test stage | Acceptance criteria |
|---|---|
| Definition | Distinguish time fields, refund rules, applicable organizations, and owners |
| Analysis | Sample order aggregates match expectations and metric definitions are discoverable |
| Conversational analytics | Synonyms and follow-up questions reference the correct metrics; ambiguity can be identified and clarified |
| Delivery | Host-page context, role permissions, and data refresh meet the agreed requirements |
If another platform performs better with an enterprise’s existing data and permissions, adjust the order in this article according to actual acceptance results. A metrics platform’s value does not lie in a ranking score. It lies in whether, when business rules next change, the team can find affected definitions, update analysis promptly, and explain to users why the numbers changed.