What is Tableau? From a table to a useful dashboard

Tableau helps you explore data from different angles and build interactive reports. You connect a source, choose fields, decide how to aggregate values and arrange views. A normal Excel file is enough to begin. The interesting part starts when you can use the chart to answer a specific question.

I treat Tableau as an analyst’s workshop. It lets me investigate an idea, compare groups and prepare a dashboard for other people. I still have to decide what revenue means, which months are complete and whether a table relationship is duplicating orders. A colorful mistake remains a mistake. It just looks more convincing in a meeting.

What Tableau actually does

A basic workflow starts with a data connection, continues with a worksheet containing a visualization and ends with several worksheets arranged in a dashboard. Readers can inspect the result and use the filters you expose. Available connections depend on the edition: an Excel file, a text file and a company database are different types of sources. The connection documentation explains the options.

This is useful for questions about changes over time, differences between categories and areas that deserve investigation. A monthly sales view can reveal a seasonal pattern. A cost ranking can point to a category worth discussing with its process owner. The chart alone does not establish why a difference exists.

Before opening the application, I write one sentence naming the reader and what they need to find out. “A sales dashboard” leaves too much room for decoration. “Check whether revenue growth is accompanied by rising delivery cost” provides a direction. I can identify the measures I need and leave unrelated ideas outside the first version.

Desktop, Public, Cloud, Server and Prep

The Tableau name covers several tools and ways of working. I separate building an analysis locally, publishing it on the internet and collaborating inside an organization. That makes the task the starting point for choosing a tool, rather than starting with a shopping list.

Desktop currently has Public, Free and Professional editions. The official comparison distinguishes their uses: Public supports learning and public visualizations, without commercial use; Free supports local analysis, including business use; Professional supports publishing to Cloud or Server. These distinctions reflect the documentation checked on October 29, 2026. I check it again before choosing an edition.

Cloud and Server differ in responsibility for hosting and maintenance. The provider handles that part for Cloud; an organization takes responsibility for its Server environment. Tableau’s deployment guide explains the models. As a beginner, you do not need to deploy a server to learn how to build accurate charts.

Prep belongs earlier in the process, when data needs repeatable preparation. The Prep documentation covers filtering, reshaping and combining sources. I choose where those operations belong based on the existing process. Some may already happen in a database or another tool. An extra processing stage should solve a specific problem.

A first project: sales and delivery cost

Consider a small synthetic example. Each row represents a completed month in one channel. All amounts use the same illustrative unit, and the cost includes delivery only. This simplification keeps attention on the report’s logic rather than on a complicated business model.

January has total revenue of 200 and delivery cost of 30. February has revenue of 300 and delivery cost of 40. Delivery cost increases, but its share of revenue falls from 15% in January to approximately 13.3% in February. You can derive all of these numbers directly from four rows, without an elaborate model.

Showing cost alone could prompt an unnecessarily quick alarm. Showing only the percentage would also leave something out: it does not explain how much the channel mix matters. The store’s share stays at 10%, while online stays at 20%. The proportions of revenue change. That gives me a useful reason to add a channel comparison beside the overall trend.

I would not call revenue minus delivery cost the company’s profit. Other costs are absent from this example. Naming a measure is part of the analysis, rather than a cosmetic task left until the end. If the report needs to describe profitability, I first agree on the definition with the person responsible for that result.

I also define the denominator for cost share. For the whole business, I divide total cost by total revenue. An unweighted average of channel percentages would give 15%, while February’s cost of 40 divided by revenue of 300 gives approximately 13.3%. The channels have different revenue weights. I record that definition so another person can reproduce the result when a third channel is added. A brief note beside the input is enough for this first project.

Prepare the table before connecting

I start with column names and the unit of observation. What does one row mean? If the source contains both individual orders and ready-made monthly totals, adding them together can double the result. A chart type cannot repair that mixture of detail levels.

I keep decorative titles, merged headers and manually entered totals outside the data area. There is a separate column for the date, channel and each measure. I represent a month with an actual date for its first day, rather than a collection of ambiguous abbreviations. Amounts are numeric, without a unit name appended inside every cell.

One file is enough for this example. I do not force several tables into the project simply to make it look advanced. If order lines are added later, I need to check the level at which delivery cost is recorded. Repeating an entire order’s cost on each line can produce a very persuasive but incorrect report.

After connecting, I compare record counts and basic totals with the input. Formatting comes later. If preparing a file is difficult, start with importing CSV into Excel and identifying column types. You can switch tools, but an ambiguous date tends to travel with the data.

Dimensions, measures and the view’s detail

In Tableau, I use channel as a dimension that divides results into groups, and revenue as a measure to aggregate. Adding another dimension can change the level of detail in the view. Understanding that matters more than memorizing the location of an icon.

A separate distinction concerns continuous and discrete fields. Blue discrete fields create headers; green continuous fields create axes. I avoid reducing this to “blue means text and green means numbers.” Dates can be continuous or discrete too. The dimensions and measures guide explains how these concepts relate.

My main question is what one mark on the chart represents. Is a point a month for the whole business, or a month for a particular channel? Adding channel to Color can create separate series. Adding an order identifier to Detail may sharply increase the number of marks. Those actions change the question being asked of the data.

In the first worksheet, I place month on the horizontal axis and total revenue on the vertical axis. I check that dates appear chronologically and both channels are included. Then I compare the first point with the control table. January should show 200. If it shows 100, I investigate a filter or an extra grouping before changing the line color.

Turn worksheets into a dashboard

For a first version, I need a revenue trend, a delivery cost share trend and a channel comparison. I do not try to squeeze the entire business onto one screen. The reader should understand the relationship between measures and identify an area worth discussing.

I combine the worksheets in a new dashboard. The dashboard documentation covers arranging views and adding interaction, including using one worksheet as a filter for others. After adding an interaction, I check which elements actually respond. An inconsistent filter can place revenue for one channel beside cost for the entire business.

Every chart gets a measure name and a unit. Beside delivery cost share, I also provide the underlying amounts so that a small percentage cannot hide a large change in scale. For this exercise, I prefer separate, clearly labeled views to a dual axis whose alignment could suggest an accidental relationship.

I also write a brief reading instruction, such as “Select a channel to compare its cost share across both months.” If three charts require half a page of instructions, I simplify the layout. I discuss readers, questions and decisions further in the dashboard article.

Reviewing the first version

I walk through the report as a new reader. First, I identify the date range without opening filters. Next, I check whether the title describes the whole business or a selected channel. I select one channel and try to return to the complete view. If clearing the selection is confusing, a reader can get stuck too.

Then I trace one number back to specific input rows. In this exercise, selecting online should leave revenue of 100 and delivery cost of 20 in each month. That is a simple control point. If the result differs, I record the filter state and investigate. Matching the grand total alone does not establish that every breakdown is correct.

Finally, I review the dashboard on the screen where it will be used. A layout that works on a large monitor may require scrolling on a laptop. I check that the legend, units and key values remain readable. This inspection belongs to delivering an analysis, alongside describing its source and definitions.

Does the data refresh automatically?

Building a dashboard does not establish its update process. Tableau distinguishes live connections, extracts and published sources. A live connection queries the source; an extract is a separately stored copy of data. Refresh behavior depends on that choice and the environment, as described in Refresh Data Sources.

I record three things beside a report: the period covered by the data, the last successful refresh and the person responsible for the source. Those dates need not match. Successfully reading an old file still produces old data. A reassuring “today” label can hide that distinction if nobody asks what it means.

For a learning project, a manual refresh with an honest coverage note is sufficient. At work, the team needs an agreed process. I ask what happens when a column is renamed, a file is missing or the source is delayed. I do not describe a report as live when its input is produced once a week.

Tableau Public and portfolio publishing

Tableau Public is useful for showing work based on data that can be disclosed. One important update to older tutorials is that Desktop Public Edition also supports saving locally. Saving on your computer and publishing to a profile are separate actions. The saving guide explains both.

For a portfolio, I choose open or synthetic data. Before publishing, I inspect details beyond the first screen, including field names, descriptions and source contents. Hiding a worksheet or omitting a person’s name from a chart does not establish that company data belongs on a public profile.

I include the question, definitions and limitations with the project. A recruiter or another analyst should be able to understand why I chose a view and what I checked before drawing a conclusion. A set of filters is difficult to assess on its own. A brief explanation turns a gallery of images into an understandable example of analytical thinking.

What Tableau makes easier, and what I still decide

Tableau makes it convenient to move between breakdowns of a dataset. I can notice an unusual month and inspect it by channel. I still have to distinguish an observation from its cause. A rising delivery cost share might involve a price change, a different order mix, a data issue or several factors at once.

I do not begin learning by trying to cover every feature. I choose one problem and build a version I can explain. If aggregation becomes confusing, I return to a small table. If the business question is unclear, a conversation with the reader usually helps more than another visual effect.

The working environment also matters when choosing a tool. Available sources, sharing requirements and responsibility for maintenance affect the decision. Comparing Tableau with another application becomes meaningful in that context. I explore this in Tableau or Power BI, rather than treating a tool name as a career plan.

I also record the boundary of the project. This example covers two months and two channels. It cannot establish annual seasonality or forecast future sales. That limitation helps keep the result useful within its actual scope, without attaching a broad story to a tiny sample.

When the first reading prompts another question, I record it separately. We might need order counts, delivery distances or a product breakdown. Every new field brings another definition and source to check. I expand the analysis gradually, keeping a way to compare the new version with the earlier, simpler result.

An exercise for your first evening

Enter the four synthetic rows from this article and build two views: revenue by month and revenue by channel. Check that January totals 200, February 300 and the whole period 500. Explain in your own words why both views describe the same dataset while showing a different breakdown.

Next, add delivery cost and compare its share of revenue. Observe what happens when you select the store channel alone. Write one finding and one limitation. The finding might concern the channel mix; the limitation might be the absence of other costs. Avoid adding a cause that these four rows cannot reveal.

Finally, ask someone to read the dashboard without your help. Pay attention to their questions and mistakes. Those observations help improve titles, units and layout. I consider the first useful project complete when its result is understandable and consistent with the input. I add further features when a concrete need appears.