Python, Charts & Storytelling · Lesson 6 of 6
Data Visualisation and Storytelling: Charts and Dashboards That Drive Decisions
A correct analysis that nobody understands changes nothing. Learn how to pick the right chart, design a clean dashboard in Power BI, Tableau or Looker Studio, and present findings so people act on them.
Two analysts find the same insight: returns are eating the profit on budget phones. The first sends a spreadsheet with 14 tabs. The second sends one chart titled "Budget phones: 1 in 5 is returned — mostly for battery complaints" and a recommendation. Guess whose work changes the decision.
Visualisation is not decoration. It is how your analysis reaches the people who act on it.
Choose the chart from the question
| Your question | Use | Avoid |
|---|---|---|
| How does it change over time? | Line chart | Pie chart, 3D anything |
| Which is biggest / compare categories? | Bar chart (horizontal if labels are long), sorted | Unsorted bars |
| What is the share of a whole? | Stacked bar, or a pie with 2–4 slices | Pie with 9 slices |
| How are values spread? | Histogram, box plot | Average alone |
| Are two numbers related? | Scatter plot | Two separate charts |
| One key number | KPI card with comparison ("₹42 L, −6% vs Sep") | A chart for one number |
Five design rules that fix most charts
- The title states the finding — "Delhi grew fastest in Q3 (+18%)" — not "Sales by city".
- Remove clutter: gridlines, borders, legends you do not need, decimal places nobody reads.
- Use colour to highlight, not to decorate: grey for everything, one strong colour for what matters.
- Start bar charts at zero. A bar axis starting at 90 makes a 2% difference look like 200%.
- Label directly — put values and series names on the chart instead of making people match a legend.
Honesty matters: truncated axes, cherry-picked date ranges and dual axes that exaggerate a link are the most common ways charts mislead — sometimes by accident. If a chart surprises you, check the axes before you present it.
Designing a dashboard
A dashboard answers the same questions every week for a specific person. Before building, write down who uses it and the three to five questions they need answered.
- Top row: 3–5 KPI cards (revenue, orders, average order value, return rate) each with a comparison to last period or target.
- Middle: the main trend (line chart) and the main breakdown (sorted bar by city or category).
- Bottom: detail table for people who want to dig in.
- Filters (date, region, category) at the top or side — and the same filters on every page.
Power BI, Tableau and Looker Studio all follow this layout. Power BI is very common in Indian corporates, Tableau in analytics teams and consulting, and Looker Studio is free and connects easily to Google Sheets — ideal for a portfolio.
Telling the story: the one-slide structure
When you present a finding, use this order:
- Context — what we looked at and why. "Returns cost us ₹18 lakh last quarter."
- Finding — the one thing that matters, with one chart. "60% come from two budget phone models."
- Why — the supporting evidence. "Return reasons: battery complaints in 7 of 10 cases."
- Recommendation — a specific action with expected impact. "Pause Model X and raise the issue with the supplier; this could cut returns by about a third."
The "so what?" test: after every chart, ask "so what should someone do differently?" If you cannot answer, the chart belongs in your notes, not your presentation.
Common mistakes freshers make
- Showing everything they calculated instead of what matters.
- Explaining the method for five minutes before giving the answer. Lead with the answer.
- Using a different colour for every bar.
- No comparison: "Sales were ₹42 lakh" — compared to what? Last month, target, last year?
Try it yourself: Take one chart you made earlier. Rewrite its title as a finding, remove three unnecessary elements, highlight one bar, and write the four-line context–finding–why–recommendation note to go with it. Compare before and after side by side.
Common interview questions
Try answering each one aloud before you open it.
1How do you choose the right chart?
Start from the question: line charts for trends over time, sorted bar charts for comparing categories, histograms or box plots for distributions, scatter plots for relationships, and KPI cards for single key numbers with a comparison.
2What makes a good dashboard?
It is built for a specific user and their few key questions, puts KPIs with comparisons at the top, shows the main trend and breakdown clearly, uses consistent filters and minimal clutter, and highlights what needs attention.
3How would you present an insight to a non-technical manager?
Lead with the finding and recommendation, support it with one clear chart and a short explanation of why, avoid jargon and method details unless asked, and quantify the expected impact.
Practise this
Build something with it
FAQs
Power BI or Tableau — which should I learn?
Either. Check job postings in your target city and companies. The skills — data modelling basics, good chart choice and clean layout — transfer between them in a few weeks.
How many dashboards should my portfolio have?
One or two strong ones built on real public data, each with a short write-up of the questions and key findings, beat many generic ones.
References & further reading
Finished this lesson?
Mark it complete. Create a free account to save it on every device.