Pebellslearn by building
0%
RoadmapsMatrixCoursesSandboxSign in
Exploring and Explaining Data  ›  Capstone Project

Communicating Results

Exploratory Data Analysis 15 minA decision-ready summary gives the answer first, then the key evidence with its uncertainty in the reader's units, then specific caveats, then the next step.

Communicating Results

The Quick Checkout analysis is finished: framed, tested, checked for Simpson's paradox and charted. Now it has to change a decision, and that happens in a short message read between meetings. A write-up that opens with "We extracted data from three tables and applied a two-proportion z-test" loses its reader before the result arrives. The work is only as useful as the first three lines people actually read.

The rule

Answer first, then evidence, then caveats, then the next step.

A decision-ready summary has four parts, in this order:

  1. The answer: the recommendation or finding in one sentence a busy reader could repeat.
  2. The evidence: the two or three numbers that justify it, with their uncertainty, in units the reader cares about: orders, rupees, customers.
  3. The caveats: specific limits, each with what you are doing about it. Not a list of every possible doubt.
  4. The next step: what happens now, and who decides.

Here is the Quick Checkout summary for Pantry's leadership:

Recommendation: switch every user to Quick Checkout this week.

Why: in a two-week randomised test with 28,800 users, it raised checkout conversion from 10% to 11%. That is about 2,000 extra orders and ₹1 lakh of margin a month; even the cautious end of the estimate adds around 560 orders.

Watch-outs: refunds and page speed showed no harm. The test ran for two weeks, so some of the lift could be novelty; we will keep measuring.

Next: product ships on Monday, and we report conversion weekly for a month.

Notice what is missing: the p-value, the test's name, the SQL. Those belong in an appendix for reviewers who want to check the method. Translate the statistics instead: "a gap this large would be very unlikely if the checkout made no difference" says what p = 0.006 means without the jargon.

The shortcut

Climb the "so what?" ladder: after each number, ask "so what?" until you reach something the reader decides with.

From a statistic to a headlineStep 1 of 4

  1. The statistic

    Conversion rose 1 point, with a 95% interval of 0.28 to 1.72 points.

A final check before sending: read only the first two sentences. If someone who stopped there would make the right decision, the summary works.

Drills

Answer fast — trust the rule.

quick drill

Which opening line works best for Pantry's leadership?

200,000 users reach checkout each month. How many extra orders does the cautious end of the interval, 0.28 points, give?
200000×0.28100= ?200000 \times \tfrac{0.28}{100} = \,?
At ₹50 of margin per order, what are those 560 extra orders worth each month, in rupees?
560×50= ?560 \times 50 = \,?
quick drill

How should you say "p = 0.006" to a non-technical reader?

quick drill

Where do caveats belong in the summary?

Remember

  • Answer first, then evidence, then caveats, then the next step.
  • Climb the "so what?" ladder until a number becomes orders, rupees or customers.
  • Trap: opening with the method, or burying the answer under every possible doubt.

Practice — level up

0/1 solved

Write your answer, then compare it with a rubric and a model answer.

ExplainSummarise the express launch

Write a four-part summary (answer, evidence, caveats, next step) for Pantry's leadership about the express delivery launch, using the difference-in-differences estimate of 6 extra weekly orders per 1,000 Pune customers. Pune has 50,000 customers. Then explain why you ordered the parts the way you did.