Configuring Table and Column Properties
A sales manager opens a Power BI report and finds two fields named Product ID. One is a business identifier that people recognize. The other is a technical key used by the model. Nearby, Revenue appears as an unformatted decimal, Order Number invites automatic addition, and several internal columns compete with the fields people actually need. The report may still produce a chart, but is the semantic model ready for someone else to use correctly?
That question gets to the purpose of table and column properties. The PL-300 course blueprint explicitly includes configuring table and column properties. This isn’t only about making a field list look tidy. These properties influence how report authors discover, interpret, and use the model’s fields. A technically functioning model can still create confusion if its names, visibility, organization, formatting, data categories, or default summarization choices don’t match the meaning of the data.
A semantic model is the Power BI model that contains tables, relationships, metadata, calculations, and the structures used by reports. Metadata is information that describes the model’s objects. A friendly table name, a column’s display format, and whether a technical key is visible are all examples of metadata decisions. They don’t necessarily change the underlying source value. They change how that value is presented and made available to people building reports.
Start by identifying the requested outcome and the Power BI layer that owns it. If a source value must be cleaned, split, replaced, or reshaped during refresh, that points toward Power Query and its M transformations. If the requirement concerns how a loaded field is named, organized, formatted, summarized, categorized, or exposed to report authors, that points toward semantic-model properties.
A calculation expressed in Data Analysis Expressions, or DAX, belongs to the model’s calculation logic. A choice that applies only to one visual belongs at the report layer. Publishing and access decisions belong to the Power BI service.
That separation matters because nearby choices aren’t interchangeable. Renaming a column in Power Query can change the name that reaches the model. Giving a loaded model field a business-friendly name addresses how it’s presented in the semantic model. Changing a label in one visual may improve that single visual while leaving the shared field confusing everywhere else. All three actions can alter words that someone sees, but they operate at different layers and have different reach.
For the sales model, the analyst first translates the business request into concrete model behavior. Report authors should find recognizable sales fields. Internal keys should stay out of their way when those keys aren’t intended for reporting.
Revenue should have a format consistent with its financial meaning. Order Number shouldn’t encourage an inappropriate default aggregation merely because its stored values are numeric. Related fields should be organized so that a person can understand what belongs together.
Names are the most immediate part of that experience. A business-friendly name should communicate the field’s meaning without forcing every report author to decode source-system abbreviations. That doesn’t mean every short name is bad or that every field needs a long description embedded in its name. It means the model’s exposed vocabulary should match the vocabulary used by the people working with the data. Product Code and Product Key, for example, need distinct names if they represent distinct concepts.
Visibility answers a different question. A column can be necessary to the model while being unnecessary or potentially confusing to a report author. A technical key may support model structure even though nobody should drag it onto a chart.
Hiding that field, when appropriate, reduces clutter and directs attention toward business-facing fields. Hiding isn’t the same as deleting. The decision is about exposure in the reporting experience, not a claim that the underlying column has no model purpose.
That qualification is important. You shouldn’t hide every identifier merely because it contains the word key or ID. Product Code might be a useful business identifier that appears on invoices, search results, or operational reports.
Product Key might be an internal value with no business meaning. Their names may look similar, but their roles differ. The requirement and the field’s meaning determine which one should be exposed.
Formatting also belongs to meaning. Suppose Net Revenue contains decimal values. A report author shouldn’t have to rediscover the intended financial presentation every time the field is used. A model-level format choice can provide a consistent default experience.
Formatting doesn’t prove that the underlying values are correct, and it doesn’t repair source-data quality. It describes how a value should ordinarily be presented. A well-formatted wrong number is still wrong, so presentation and data validation remain separate responsibilities.
Default summarization needs the same care. Numeric storage doesn’t automatically imply that addition is meaningful. Revenue commonly supports aggregation in a sales analysis.
Order Number may contain digits but still represent a label rather than a quantity. Adding order numbers produces a number, yet that number has no useful business interpretation. The model’s default summarization choice should reflect what the field means, not just which data type happens to store it.
Data categories and metadata organization provide more context to the authoring experience. A data category helps describe the kind of information a field represents. Metadata organization helps related model content appear in a structure that report authors can navigate. These settings aren’t decorative rewards added after the real modeling is complete. They’re part of the model’s usable contract with the people who build reports from it.
Now consider what happens when a second analyst receives this model. They create a sales page and search for product information. If the field list presents clear business names, keeps irrelevant technical columns out of view, and offers sensible formatting and summarization defaults, the model guides them toward reasonable choices. It doesn’t remove the need for judgment, but it reduces avoidable mistakes. If instead every source column is exposed with cryptic names, the report author has to reverse-engineer the model before doing useful analysis.
This is why there isn’t one universally correct combination of properties for every Power BI model. A setting is not appropriate merely because the interface permits it. The same type of column can serve different business roles. One numeric identifier may be a technical key that should be hidden.
Another may be an important code that users need to see. One decimal field may represent an amount. Another may represent a rate that requires a different presentation. The modeler has to connect each choice to the field’s actual meaning and intended audience.
The sales team later reports a problem. Net Revenue looks reasonable in an ordinary summary, but Order Number is being treated like a value that should be added. The analyst shouldn’t begin by changing unrelated query transformations or redesigning every visual. The observable symptom concerns the default way a model field is offered for use. That directs attention back to the column’s semantic-model properties.
A disciplined troubleshooting process begins by reproducing the behavior. The analyst uses Order Number in a controlled report context and records what Power BI offers or produces. They compare that result with the expected business behavior, which is to treat the order number as an identifier rather than an additive quantity. Then they inspect the property that governs default summarization and correct the mismatch. Finally, they repeat the original test and check another case that uses the same field.
That last check matters because a successful configuration action isn’t evidence that the requirement has been satisfied. Power BI accepting a property change tells you that the change was applied. It doesn’t tell you whether the field now behaves appropriately for its business purpose. Correctness has to be observed in the layer where the behavior occurs. For a model property, that means inspecting the semantic model and then checking the report-author or consumer experience affected by it.
The same reasoning applies to a hidden key. The analyst shouldn’t stop after selecting a visibility setting. They verify that the business-facing Product Code remains available while the internal Product Key no longer distracts report authors. They also confirm that the model still supports the report behavior that required the key. The test connects both sides of the decision: the technical field remains useful to the model, and the reporting experience exposes the appropriate business field.
Renaming deserves similar verification. Suppose Product Code and Product Key were both shortened to Product ID in an attempt to simplify the model. The names now look friendly, but they erase an important distinction. A report author can’t confidently tell which identifier belongs in a table.
The correction isn’t simply to choose prettier wording. It’s to choose names that preserve meaning. Verification means that someone examining the model can distinguish the business code from the internal key without guessing.
Formatting can fail more quietly. Net Revenue might display as a plain decimal in one place because a report author manually formatted a visual and assumed the model was fixed. Another report built from the same semantic model could still inherit an unsuitable default.
To determine which layer owns the correction, ask whether the requirement applies only to one visual or should follow the field as part of the reusable model. A visual-specific requirement can stay with the visual. A shared default for the business field points back to the semantic model.
This layer test prevents a familiar but inefficient response to Power BI problems: changing the first related setting you recognize. If a confusing name exists throughout the shared model, editing one visual’s label treats only one symptom. If a source value itself is malformed, changing its display format doesn’t clean it. If a report author can see an internal key, adding a DAX calculation doesn’t address that visibility problem. Restate the requirement first, identify the owning layer, and then inspect the mechanism that directly controls the result.
You can make the validation more reliable by separating normal and edge conditions. In the normal case, confirm that a report author can find Net Revenue, use its intended default presentation, and distinguish Product Code from Product Key. Then check less convenient conditions.
Search for fields rather than relying on where they happened to appear. Build a second report page or use another report context to see whether the behavior comes from the shared model rather than one customized visual. Use an identifier containing numeric values to confirm that its meaning, not its appearance, guided the summarization decision.
Filters can also reveal the difference between a field property and a broader model problem. Suppose Net Revenue is formatted correctly, but its result becomes unexpected under a particular filter. The visible symptom involves the same field, yet formatting may have nothing to do with the wrong value. Reproduce the filter state, compare expected and actual results, and inspect the layer that governs the calculation or model behavior. Don’t keep changing column metadata when the observed failure belongs to relationships, calculation context, or another part of the solution.
Refresh conditions require the same restraint. If a renamed or formatted field looks right but refreshed data contains incorrect values, a semantic-model display property isn’t a substitute for examining query output or source data. If refreshed values are sound but the authoring experience remains confusing, then the table and column properties deserve attention. The symptom tells you where to begin, while controlled testing tells you whether that first diagnosis was correct.
A useful acceptance check for the sales model has several parts. Can a report author locate the fields using business language? Are fields with different roles named distinctly? Are technical keys hidden when they’re not intended for reporting, while useful business identifiers remain available?
Do formatting and default summarization choices agree with each field’s meaning? Are related fields organized coherently? Most importantly, can the analyst demonstrate those outcomes in the model and in the resulting report-author experience?
Notice that none of those checks asks whether the analyst remembers a particular button sequence. Interface locations can change, and different editing experiences can present properties differently. The durable skill is understanding what the property controls. When you know that visibility governs exposure, formatting governs default presentation, summarization should follow semantic meaning, and naming communicates business concepts, you can evaluate a model even when the interface isn’t familiar.
There’s also a governance consequence, though the decision still begins with usability. A shared semantic model gives multiple report authors a common set of fields. If each author has to reinterpret technical names, hide unwanted keys independently, or correct presentation one visual at a time, the shared model isn’t carrying enough of the intended meaning. Appropriate table and column properties make the reusable model a clearer starting point. They don’t eliminate report-level choices, but they reduce unnecessary reinvention.
Return to the original sales report. Two fields named Product ID weren’t merely an untidy detail. They concealed different roles. The unformatted revenue field failed to communicate its ordinary presentation.
The additive order number invited a mathematically possible but meaningless result. The exposed internal columns made the model harder to navigate. Each symptom points to a property decision, but each decision still has to be justified by what the field represents.
The analyst’s correction is therefore deliberate rather than uniform. They preserve a visible business-facing Product Code, give the internal Product Key a distinct name and hide it when it isn’t meant for report authors, apply an appropriate default presentation to Net Revenue, and configure Order Number in line with its role as an identifier. They organize the exposed metadata around how the business understands sales. Then they test those outcomes through the authoring experience instead of trusting that the configuration was saved.
That final verification is what separates a setting from a solution. A property can be present and still be poorly chosen. A visual can appear correct while masking an inconsistent model default. A normal case can pass while a different filter or refreshed data condition exposes another cause. By comparing expected and actual behavior at the correct layer, the analyst can explain not only what was changed, but why the resulting semantic model now meets the business requirement.
Configuring table and column properties is ultimately an exercise in preserving meaning. Business-friendly names and organized fields help people find the right content. Appropriate visibility keeps necessary technical structures from becoming reporting distractions. Formatting and default summarization guide ordinary use, but only when they match what the field actually represents.
The reliable approach is to begin with the business requirement, identify whether the behavior belongs in the query, semantic model, calculation, report, or service layer, and change the property that directly governs it. Then verify the result where report authors and consumers encounter it. When the expected behavior is observable in that context, you have more than a saved configuration. You have evidence that the semantic model supports its intended use.
Choose table and column properties that help consumers interpret and use the model correctly. Which field behavior in a report would reveal a misleading format, summary, or category setting? A focused checklist and the lesson's exact labels are available in the companion episode reference. This lesson was produced by Bare Metal Cyber dot com.