Lesson 11 of 11 · 10 min · Telling the story

From dashboard to decision: insights, recommendations and presenting

Covers: General data storytelling practice for analysts. Updated 13 Nov 2026. Part of the Data Analyst roadmap.

What you will learn

  • The difference between a number, an observation and an insight
  • Write recommendations that someone can act on
  • Structure a short stakeholder presentation
  • Answer questions and handle challenges to your numbers
  • Turn a project into a portfolio case study

In simple words

A dashboard shows what happened. A manager wants to know what it means and what to do. "Sales fell 12 percent" is a fact. "Sales fell 12 percent because mobile checkout is failing on one Android version, costing about ₹2 lakh a week, so we should fix it this week" is an insight with a recommendation. Learning to move from the first to the second is what makes an analyst valuable.

Words to know

  • Insight: a finding that explains why something happened and why it matters.
  • So what: the consequence of a finding for the business.
  • Recommendation: a specific action, with an owner and a measure of success.
  • Stakeholder: a person who depends on or decides from your analysis.

From number to insight

Three levels, each more useful than the last:

  1. Data: "Mobile conversion is 3.3 percent."
  2. Observation: "Mobile conversion fell from 4.7 to 3.3 percent last week while desktop was flat."
  3. Insight: "The fall started with the 14 October Android release; payment failures on that version rose to 28 percent, costing roughly ₹2 lakh of sales a week."

To reach an insight, keep asking: compared with what (target, last period, other segments)? Why (segment, funnel step, event)? So what (money, customers, risk)? What would we do about it? An insight usually has a number, a comparison, a cause and a consequence.

Check the insight before sharing: is it true (reconciled with a source)? Is it new to the audience? Is it actionable? Would someone make a different decision because of it? If not, it is trivia.

Writing a recommendation

A recommendation answers "what should we do?" Write it so a busy person can say yes or no:

Part Example
Action Roll back app version 4.2 or release the payment fix
Owner Mobile engineering lead
When By Friday
Impact Recover about ₹2 lakh a week in sales
Cost and risk Rollback removes two small features for a week
How we will know Mobile conversion back above 4.5 percent within a week; payment failures under 3 percent

Offer options when the choice is not obvious (fix forward, roll back, do nothing) with a recommended one and the reason. State your confidence and what you do not know ("the cause is confirmed in logs for 80 percent of failures; we are checking the rest"). Separate facts, interpretation and opinion clearly.

Structuring a presentation

For stakeholders, lead with the answer, not the method. A reliable structure for 5 to 10 minutes:

  1. The headline: the one-sentence conclusion and the decision needed. ("Mobile checkout is failing since 14 October; we need to roll back today.")
  2. The context: the goal and the metric, in a sentence.
  3. The evidence: two to four visuals, each with a message title, building the argument: the problem, where it is concentrated, the cause.
  4. The impact: in rupees, customers or risk.
  5. The recommendation and next steps: owner, date, measure.
  6. Appendix (not presented): methods, definitions, data quality notes, extra charts.

Rules of thumb:

  • One message per slide or page, shown in the title.
  • Numbers with comparisons, rounded sensibly (₹2 lakh, not ₹1,98,742).
  • Show the chart, not the table; keep detail in the appendix or the live dashboard.
  • Use the audience's words, not statistics terms ("most customers" instead of "the median").
  • Tell a short story: situation, problem, cause, solution, the way a person would explain it.
  • For a live dashboard walk-through, start from the KPI cards, narrate one path (a trend, then a breakdown, then a drill-down) and stop to state the point; do not click through everything.

Know your audience

Audience They want You give
Executive the decision, the risk and the money headline, impact, recommendation; one or two charts
Manager what to change in their area segments, drivers and an action plan
Analyst or engineer how it was calculated definitions, queries, assumptions, data quality notes
Customer or client the effect on them simple language and what you will do

Ask beforehand: who is deciding, what decision, how much time, what do they already believe? Send a one-paragraph summary before the meeting, so the discussion is about the decision.

Handling questions and challenges

  • "Is this number right?" Have the reconciliation ready: the source, the definition, and how it matches the finance or system figure.
  • "Why did you exclude X?" Know your filters and exclusions and state them upfront on the page.
  • "What about Y?" If you do not know, say so and promise a follow-up date; never invent an answer.
  • "Correlation versus cause?" Be honest about what the evidence shows and what needs a test.
  • Disagreement: ask what evidence would change their mind, and agree on a way to test it.
  • Do not hide bad news or weak data; state uncertainty plainly, with ranges where you have them.

Responsible use of data

  • Show only what is needed; avoid exposing individuals' personal data in reports and screenshots.
  • Do not mislead by choosing time ranges or axes to support a conclusion.
  • Be clear when a result is based on a small sample or a short period.
  • Declare conflicts: if the analysis affects your own team's targets, say so and let someone else review it.
  • Follow your organisation's data policies and the law on personal data.

A portfolio case study

Employers hire analysts partly for judgement. Turn a project into a one-page story:

  1. The question: who needed what decision? ("Which courses should the institute promote next quarter?")
  2. The data: source, size, period, limits. (Use public data or fake, clearly labelled data; never a real company's private data.)
  3. The work: cleaning decisions, the model (a star schema diagram), key measures (a screenshot of a DAX measure with a one-line explanation).
  4. The findings: two or three insights, each with a chart and a number.
  5. The recommendation and how you would measure success.
  6. What you would do next and what you learnt.
  7. Links: the report or screenshots, the GitHub repository, a short video walk-through.

Practise explaining it in 90 seconds and in 5 minutes. In an interview, "what did you find, and what would you do?" matters more than "which chart did you use?".

Common mistakes

  • Presenting findings without a recommendation.
  • Starting with the method instead of the answer.
  • Too many charts and decimals.
  • Hiding uncertainty or overclaiming cause.
  • Not reconciling the numbers before presenting.
  • Using jargon the audience does not use.

Try it yourself

  1. Take a chart from any report and write the data, observation and insight versions of its message.
  2. Write a recommendation using the six parts (action, owner, when, impact, risk, success measure).
  3. Outline a five-minute presentation for a manager using the structure above, with message titles for each slide.
  4. Write the answers to three tough questions you might get about your numbers.
  5. Draft a one-page portfolio case study from a project you have done.

Check yourself

What turns an observation into an insight?

Adding the comparison, the cause and the consequence: why it happened and why it matters.

What should every recommendation include?

A specific action, an owner, a date, the expected impact, any cost or risk, and how success will be measured.

Why lead with the conclusion in a presentation?

Decision makers have limited time; the headline and the decision needed set up the evidence and keep the discussion focused.

What do you do when you cannot answer a question about the data?

Say you do not know, note it, and commit to a follow-up with a date, instead of guessing.

Interview questions

How do you present a finding to a non-technical stakeholder?

Lead with the conclusion and the decision needed, use plain language and one or two clear visuals with message titles, give the impact in business terms, and finish with a specific recommendation and next steps.

Walk me through a project where your analysis changed a decision.

Describe the business question, the data and method briefly, the insight with numbers, the recommendation, what was decided and the measured result.

What would you do if your analysis contradicted a senior person's belief?

Check my data and definitions, present the evidence neutrally with the uncertainty, ask what would change their mind, and propose a test where possible.

Summary

  • Move from data to observation to insight: compare, explain and state the consequence.
  • Recommendations name the action, owner, timing, impact, risk and success measure.
  • Lead with the answer, use message titles, few visuals and plain language; tailor to the audience.
  • Reconcile numbers, state uncertainty, and turn projects into short portfolio stories.

Going deeper

See how much of this you can explain

A free mock interview asks you new questions on Power BI and dashboards for Data Analyst and scores every answer, with what to fix.