Business intelligence once meant building a report, publishing it and moving on to the next request. Three changes have made the job harder to do casually. Copilot in Power BI and the AI features in Tableau can answer questions directly from a model, so a badly named or badly built model now produces wrong answers in plain English. Microsoft Fabric has pulled Power BI closer to the data platform. Reports have also become code, stored in Git and promoted through environments. Power BI developer skills, and BI skills in general, now lean towards semantic modelling and engineering discipline over chart-building.
What follows applies to Power BI and Tableau, with notes where the two differ, and most of the reasoning carries across to Looker or Qlik. You get eight core skills with a check for each, the working habits that matter, what is rising and falling in value, a matrix by level and a way of testing candidates beyond tool menus.
Core technical skills for a BI developer
Relationships and role-playing dates
Beyond a basic star schema, good developers handle the awkward cases: tables with several date columns, many-to-many links and filters that must flow one way only. They avoid bidirectional relationships unless they can explain the ambiguity involved. How to check it: say orders have an order date and a ship date, and users want to view revenue by either from one report. Ask how they would model it. Good answers cover an inactive relationship with USERELATIONSHIP or a second date dimension, and when each one suits report users better.
Calculation logic in DAX or Tableau
Basic measures are easy to generate. The skill lies in knowing how evaluation works when a calculation meets real calendars and partial periods. Indian companies mostly report on an April to March financial year, which trips up default time intelligence. How to check it: for Power BI, ask for a financial year-to-date measure, then a comparison with the same period last year when the current month is incomplete. For Tableau, ask where a context filter and a dimension filter sit relative to a FIXED expression in the order of operations, and how that changes the result.
Power Query and query folding
Transformations are cheapest when they run at the source. Query folding means Power Query pushes steps down as native SQL, and one careless step can end that. Incremental refresh depends on it. How to check it: say a table set up for incremental refresh still takes as long as a full load. Ask where they would look. A strong candidate checks whether the date filter folds, uses View Native Query or the query diagnostics, and may suggest moving logic upstream into a view or dbt model.
Storage modes and cost at the source
Import, DirectQuery, Direct Lake and composite models each trade speed, freshness and source load, as do Tableau extracts and live connections. A choice made for convenience can fall on someone else’s budget. How to check it: describe a finance pack refreshed daily and an operations page that must be close to live, both on a Snowflake warehouse. Ask which mode suits each and what the report does to warehouse credits. Good answers notice that a busy DirectQuery page can keep a warehouse running all day.
Security beyond a role filter
Row-level security is only one layer. Developers should know dynamic rules based on the signed-in user, object-level security for hiding columns, and how workspace roles and sharing interact with them. How to check it: say a manager tests a row-level rule in the workspace and can see every region, and ask why. A person with real experience knows those rules apply to Viewers, and that workspace Admins, Members and Contributors see all the data. Ask how they would test the rule properly before release.
Version control and deployment
A single PBIX on someone’s laptop is a risk. Power BI Project files and TMDL make models readable in Git, deployment pipelines move content between environments, and Tabular Editor helps with bulk changes. Tableau teams have their own routes through projects and content migration. How to check it: say two developers changed the same PBIX and one overwrote the other’s work the day before month-end. Ask how they would set things up so this cannot recur. Look for source control, separate environments and parameterised connections.
History and changing dimensions
Businesses restructure. Customers move between regions and products change category, and someone must decide what past reports should show. How to check it: say a large customer moved from the South region to the West last quarter. Ask whether its past sales should move too, and how they would support both views. Strong candidates ask the business what it wants first, then talk about slowly changing dimensions and an as-of join, usually built with the data engineers.
Report design for a decision
A good report page puts the answer first and removes everything else. Knowing chart types matters less than knowing what a particular person needs at a particular moment. How to check it: show a crowded page with a dozen visuals and several slicers, then say the regional manager reads it on a phone before the morning call. Ask what they would keep. Good candidates ask what that manager decides, and they cut hard. Watch also for colour used for meaning instead of decoration.
Working skills that separate good from average
- Starting with the decision. Asked for a dashboard, strong developers ask what action it should trigger and who takes it. Many requests shrink or disappear after that conversation.
- Retiring reports. Usage data usually shows that many reports go unopened. Good developers check it and propose removal, which reduces maintenance and leaves fewer rival numbers in circulation.
- Coaching self-service users. Analysts will build their own reports whatever anyone says. A strong BI developer gives them certified models to build on and teaches them enough to stay out of trouble.
- Keeping calm at month-end. A refresh failure during closing affects finance deadlines. Dependable people keep monitoring in place, tell users early and have a fallback ready.
Rising and falling skills in 2026
Mattering more. The semantic model has become the interface for AI, so clear table and measure names, descriptions and AI instructions now affect the quality of Copilot answers. Reviewing generated DAX matters too, because a measure can look right at row level and be wrong in totals. Fabric knowledge helps where a company has adopted it, particularly Direct Lake and how capacity gets consumed. Treating reports as code, with Git and deployment pipelines, has moved from unusual to expected on larger teams. Governance, including sensitivity labels and endorsed models, also matters more as personal data spreads through self-service reports.
Mattering less. Heavy transformation inside the BI tool is giving way to logic in dbt or the warehouse, where it can be reused. Hand-assembling routine visuals counts for less now that assistants can draft a page. Remembering DAX syntax is less important than knowing whether the result is right. Older platforms such as QlikView and Crystal Reports are mainly migration knowledge.
BI developer skills matrix by level
| Skill area | Junior | Mid-level | Senior |
|---|---|---|---|
| Modelling | Builds simple star schemas | Handles role-playing dates and mixed grains | Sets shared, certified model standards |
| DAX or Tableau logic | Writes basic measures and fields | Solves time and context problems | Reviews others’ logic for correctness and speed |
| Data preparation | Cleans data in Power Query or Prep | Keeps queries folding and refreshes incremental | Decides what moves upstream |
| Performance | Removes obvious slow visuals | Picks storage mode and diagnoses queries | Plans capacity and source load |
| Security | Applies existing rules | Builds dynamic row-level rules and tests them | Designs the access model and workspace roles |
| Deployment | Publishes under guidance | Uses Git and environments | Sets release process and tenant or site rules |
| Report design | Follows the team template | Designs for one audience and decision | Sets design standards and retires reports |
| Stakeholders | Clarifies requirements with help | Runs requirement sessions alone | Leads metric agreement across departments |
How to assess BI skills in a hiring process
A polished dashboard says little about what sits behind it, and portfolios often show only the surface. Set the scope in your BI developer job description first, then use exercises that open up the model.
- Audit an inherited file. Give a small PBIX or workbook with a bidirectional relationship, a hard-coded measure and an unused table. Ask for a short list of what they would fix and in what order.
- Role-play a requirements meeting. Have a panel member play a sales head who wants a dashboard “with everything”. Score the questions asked, and whether the candidate leaves with a defined metric and user.
- Check a generated measure. Share a DAX or Tableau calculation produced by an assistant that fails at subtotal level. See whether they test it against known figures before accepting it.
- Ask about something they removed. People who have run a reporting estate can name a report they retired or a model they merged, and they can describe the argument that followed.
Red flags on a CV: a count of dashboards as the headline achievement; Power BI Desktop work with no mention of the Service, refresh or security; no SQL anywhere; identical project descriptions for Power BI and Tableau; and no named audience for any report.
Frequently asked questions
Is the PL-300 certification worth asking for?
It is a fair sign that someone has covered the Power BI basics, and it is useful for juniors. It does not show modelling judgement or experience with a messy tenant, so keep it as a preference and test skill directly.
Does a Power BI developer need Microsoft Fabric experience?
Only if your company runs Fabric or plans to. Good modelling and DAX carry across, and the additions, such as lakehouses and Direct Lake, can be learned. Where Fabric is central, look for someone who has worked with capacity limits in practice.
Can an Excel analyst become a BI developer?
Many do, since they already understand the business and care about the numbers. Their usual gaps are dimensional modelling, SQL and the habit of building for other users. Pivot table skill alone is not enough, so look for evidence of those gaps closing.
How much Python should a BI developer know?
For most roles, none is required. SQL matters far more. Python becomes helpful for automation or for working with data scientists, and it is a bonus rather than a requirement unless the role is drifting towards analytics engineering.
If reporting is the problem you need solved, Zilo AI can help you hire BI developers in India, whether on contract, on contract-to-hire or as permanent staff. Data engineers and analytics architects for the wider chain are on our IT talent page. Tell us which tool you run and what the business needs from it.