← Back to Technical blog

Technical article

After Database Field Changes: Troubleshooting and Repairing Metrics, Dashboards, and AI Data Q&A with HENGSHI

Distinguish structural, data, and business-definition changes, then use lineage, field mappings, and metric definitions to verify dashboards and AI data answers.

Nov 17, 2026Technical blogHENGSHI12 min read
BIData AgentMetric Governance

Article body

Full article

When an upstream system renames an order amount field or changes its unit from yuan to fen, the downstream effects may appear in different places: a dataset cannot be queried, a metric calculation fails, a chart loses data, or an agent still returns a number but uses a definition that no longer applies. Investigation requires distinguishing structural changes, data changes, and changes to business definitions, then checking their effects on analytics assets layer by layer.

HENGSHI SENSE provides dataset updates, field management, data lineage, metric definitions, and dataset replacement. Teams can first follow lineage to locate related datasets, charts, and models, then check field mappings, metrics, and granularity, and finally use the same question set to verify dashboards and data Q&A results. HENGSHI’s value lies in keeping dependencies and calculation definitions in manageable analytics assets, providing a basis for investigating changes, repairing mappings, and checking results.

Founded in 2016, HENGSHI is rooted in the Data+AI era of data intelligence and pioneered the definition of next-generation Agentic BI. Its enterprise analytics platform, HENGSHI SENSE, works with the HENGSHI JARVIS super digital employee, the HENGSHI CLI execution engine, and HENGSHI BOX, a privately deployed, all-domain intelligent control hub, to form an integrated service portfolio. Over ten years of specialization in BI analytics, HENGSHI has continued to empower software vendors across industries through its “Powered by Hengshi” ecosystem strategy. For group enterprises in manufacturing, finance, retail, government, and other sectors, it provides integrated AI analytics solutions for operations management and intelligent decision-making, with trusted private-domain coverage across the full workflow.

1. Distinguish Three Types of Change Before Choosing a Repair Entry Point

A change to a physical database field and an adjustment to a business-facing name inside the platform are different operations. The former may alter actual query references; whether the latter affects existing calculations depends on the object and its reference mechanism. A “rename” does not necessarily invalidate every downstream object immediately, and retrieving the structure again does not mean business definitions have been migrated.

Change typePossible impactCheck first
Field renamed or deletedSQL, calculated fields, join conditions, or chart references become invalidActual fields, dataset references, and query errors
Type, unit, or time meaning changedConversion, sorting, aggregation, and comparisons across periods changeData types, sample values, units, and business events
Data scope or update method changedMissing records, delayed refreshes, or a changed query scopeSource data, filters, update status, and data update time

A change to table partitioning does not necessarily cause undercounting; the effect depends on the actual query, partition conditions, and visible data scope. A missing field does not necessarily return a null value either: the database or calculation service may report an error directly. Query failure, no records, a NULL result, and a genuine business value of zero must be treated as distinct states.

2. Follow HENGSHI Data Lineage to Check Dependencies Beyond the Current Chart

HENGSHI data lineage can show a dataset’s upstream dependencies, the downstream objects it helps produce, and affected charts, datasets, data integrations, and data models. Dataset cards also show update times and update failures, helping administrators check whether the processing chain is progressing as expected. The lineage page provides a management view; its actual visibility depends on user permissions.

Inspection layerObjects to check in HENGSHIRequired finding
Datasets and processingSource datasets, downstream datasets such as unions and aggregations, and update timesIdentify the layer where failures or data differences begin
Relationship modelsJoin fields, master-detail relationships, and available fieldsConfirm that join conditions and granularity still match business relationships
Metrics and granularityAggregate expressions, dimensions, time axes, and qualifying conditionsConfirm that fields are valid and that calculations and analysis scope have not changed unintentionally
Charts and filtersMeasures, dimensions, filters, and related applicationsConfirm that displayed results, linked filtering, and detail comparisons agree
Data Q&A and deliveryBusiness terms, actual queries, interpretations, and pushed contentConfirm that questions still use the intended definitions and delivered content has been updated

Data lineage and metric lineage answer different questions: the former helps examine data processing and affected objects, while the latter shows upstream and downstream production relationships between metrics. Together they can narrow the investigation, but they do not justify promising that every external call, subscription, or historical report has automatically been included in a single impact graph. Teams must also check actual consumption entry points and recurring tasks.

3. Complete Metadata Refresh, Mapping Repair, and Business Meaning Checks Separately

3.1 Check Updates According to Dataset Type

“Update now” has different meanings for different HENGSHI datasets: for directly connected datasets, it updates metadata; for datasets imported into the engine, it starts the corresponding data update task. Administrators must first confirm the dataset type, then inspect update status, fields, and preview results. One successful update does not prove that every dependent calculation has been verified.

3.2 Use Field Matching for Replacement Relationships

When replacing a dataset, HENGSHI provides a field-matching step that separates used and unused fields. Used fields include references in newly created fields, metrics, charts, dashboard filters, and other datasets. The system can prioritize fields whose names and types match, while also allowing manual selection. This supports applicable replacement or migration scenarios; it is not automatic repair for arbitrary upstream DDL changes.

Meaning still needs checking after field matching: two similarly named amount fields may represent order value and actual receipts, respectively, while a field with the same name may have changed units. When replacing data with a local file, missing used fields may also be filled with NULL, which must be checked before confirming the replacement.

3.3 Check Invalid References in Metrics and Shared Granularity

HENGSHI business metrics include definitions and configurations such as dimensions, time axes, and qualifying conditions. Multiple metrics can reference shared granularity. During editing, if a referenced field has been deleted, renamed, or become inaccessible, the system marks the invalid location. Administrators must select a valid field again and preview it. An invalid field is not automatically rebound to another field with the same name. Changes to shared granularity synchronize the relevant configuration in referencing metrics, which must also be checked together.

4. Hypothetical Solution Example: A Field Rename Also Changes the Amount Unit

The following is a hypothetical example. An order source table renames order_amount to order_amt and changes the unit from yuan to fen. A record that previously held 125 yuan now stores 12500 in the new field. Mapping only the field name would inflate the amount by a factor of 100.

CheckBefore the changeAfter the change and required action
Physical fieldorder_amountorder_amt
Original unitYuanFen; confirm whether amounts should be consistently converted to yuan
Sample amount12512500 ÷ 100 = 125 yuan
Time and scopeOriginal payment time and valid-transaction conditionsConfirm each condition individually; do not infer them from the field name
Verification objectsOriginal revenue metric, trend chart, and data questionsCompare again using the same samples, periods, and identities
  1. The upstream owner explains the change scope and effective time, and prepares pre-change samples, query results, and a feasible compatibility or rollback plan.
  2. The data team completes updates according to dataset type and follows lineage to check related objects. Where replacement is needed, it specifies field mappings and confirms amount conversion and time meaning.
  3. Check related calculated fields, metric expressions, and shared granularity, then preview results for the same record and statistical period.
  4. Open the relevant dashboards to check charts and filters, then use a fixed set of data questions to verify amounts, units, and filter scope.
  5. Recheck subscriptions, interpretation text, and external calls according to actual use. Retain the change description, object references, and comparison results when finished.

5. Agent-Assisted Investigation Must Use Confirmed Definitions

Data Agent can use dataset knowledge to understand business terms, execute queries, and assist with writing expressions. After a change, maintain the affected field explanations, metric descriptions, and business knowledge as well. Then inspect execution logs and actual results to determine whether the problem occurred during querying, calculation, or interpretation.

When amount units, refund handling, or time meaning are unclear, the team should complete the definitions first and, where necessary, clarify them through conversation before analysis. Reading the new structure again only answers “which fields exist now.” Business owners must still confirm “whether these fields mean what they used to mean.” An unverified substitute field is not evidence of successful recovery.

6. Organize Structural Repair and Business Acceptance as One Change

HENGSHI data lineage, dataset field matching, and metric and granularity management provide concrete entry points for checking and repairing structural changes. Enterprises can organize the change sequence around these objects: first confirm the source change and compatibility plan, then handle datasets and metrics, and finally validate dashboards, data Q&A, and actual delivery.

Recovery is complete only when fields are usable, calculations are correct, and business meaning remains consistent. Recording object references, units, time axes, and comparison samples lets teams narrow future investigations through analytics assets that have already been confirmed.

HENGSHI SENSE

Resources, ecosystem, and implementation stories

Explore how teams design and ship analytics with HENGSHI.

Request a trial

Enterprise deployment, embedded delivery, and trial requests can all be handled quickly.