Data Analysis Deck

About this template

One question in ten slides

A 10-slide format for explaining an analysis of log, transaction or survey data to decision makers who did not run it. The cover subtitle reads “from defining the analysis question to trend analysis, a 10-slide template”: one analysis answering one question, not a dashboard replacement. Use a dashboard for ongoing metric tracking and this file for a single question such as “why did churn rise?” The strings at the top of the cover – “DOT MATRIX 10 x 8” and “FOCUS R03 – C06” – are decoration; replace them with a company name or an analysis code, or delete them.

Filling order

1. The agenda has five lines – analysis question, data scope, hypotheses and findings, key metrics, trend analysis – and all five have slides behind them. Only the first two come as a divider plus a body slide; the other three are a single slide each.
2. Write the three conclusion lines on slide 4 first. Its lead line says to pin down in one sentence what you wanted to know, and the defaults are the difference between churned and retained customers, three testable hypotheses, and the deciding metric fixed in advance. Each line is numbered 01/03, 02/03, 03/03, so it is a slot you cannot exceed; method comes later.
3. Slide 6 takes the data definition. Under a lead line about naming which table and how many records, it ships with 4.18M transaction log records over 24 months, a join to the member attribute table, and 1,932 excluded test accounts – period, join target and exclusions in three fixed slots. Without them, nothing downstream can be verified.
4. Slide 7 is a two-column comparison. The left column, hypotheses, holds three: bigger discounts raise repeat purchase, app users buy more, weekend revenue beats weekdays. The right column, findings, answers each: repeat purchase falls above a 20% discount, app users show a 1.7x basket size, weekday evenings peak. The layout exists so that rejected hypotheses stay on the left; keep all three pairs.
5. The four KPI cards read 4.18M records analyzed, a 24-month window, a 1.7x repeat-purchase gap and a p-value of 0.03 – already analysis figures, so change the numbers and name the test under the p-value card. The 1.7x on the card is the same figure as the 1.7x basket size on slide 7, so edit both together.
6. The four trend bars run week 1 at 100, week 4 at 66, week 12 at 43 and week 24 at 35 – a retention curve after sign-up. For any declining metric change only the labels; for a rising metric you must also reverse the bar heights.

Sample lines

• Conclusion, e.g. “Seven-day churn rose from 24% to 31%; 78% of the increase occurs at first checkout on mobile web.”
• Data definition, e.g. “2025-11-01 to 2026-01-31, 42,180 new sign-ups; internal accounts and under-7-day observations excluded.”
• Recommendation, e.g. “Cut checkout fields from nine to four and A/B test for two weeks.”

Computation and the chart

Churn = churned / observed x 100; if the denominators differ between periods, never show the rate alone – write “31% of 42,180”. Contribution = segment increase / total increase x 100. The bars on slide 9 are four shapes, not a chart, so resize each as (value / largest value) x the tallest bar’s height. The slide carries no unit caption at all, so state what 100 represents and whether a decline means improvement. The closing slide holds only a headline about the action the numbers point to, plus a thank-you, so put the recommendation with an owner and a date there.

Common mistakes

• Turning correlation into cause: coupon users may simply be loyal customers already. Say “is associated with”.
• Omitting sample size – 31% of 400 and 31% of 42,180 carry different weight.
• Clearing the left column of slide 7 and keeping only results, so the next analyst repeats the hypotheses that already failed.
• Adding preprocessing rules or limitations verbally when they are not in the agenda. Duplicate the divider-and-body pair and add an agenda line instead.