Free resource

The Data Playbook.
Analyse it. Chart it. Land it.

A free, practical guide to analysing data, drawing it honestly and presenting it to people who have to decide something. No sign-up. Open a folder and play with the examples.

00Start hereTwo minutes, then pick a folder.
0.1How this playbook works

Analysis first, then the chart, then the room. That is the order most data work fails in.

Someone asks a vague question, the analysis answers a different one, and the slide ends up decorating a decision that was already made. Each folder fixes one stage.

My test for everything here: would this change what someone does on Monday morning? If not, it is decoration.

0.2Who it is for

Anyone who makes or reads charts at work. No coding needed.

You do need to be willing to be wrong in public, which is the real skill.

This is my practical view, drawn from working in insight and analytics. It is not peer-reviewed research and it does not speak for my employer.

01Analyse it properlyBefore anyone sees a chart
1.1Start with the decision, not the dataset

No decision, no analysis. Write the sentence before you open the data.

A decisionName it, and name the decision-maker.
A dateSpecific, measurable and time-bound.
The right sizeNarrow enough to answer, not a fragment of something bigger.

If you cannot finish the sentence: nobody has agreed who decides. That is why everyone keeps asking for "a bit more analysis".

1.2Break the problem into a driver tree

Split the outcome into parts that do not overlap, until each part can be measured.

Tap a box to split it further. Two to four levels is usually enough.

Write a hypothesis against each branch and test it. Keep the ones that survive in the story. Your audience does not need a tour of every dead end.

Watch for overlap: if the same customer sits in two boxes, your arithmetic will flatter you.

1.3Check the data before you trust it

Five minutes of reconciliation beats finding out in the meeting.

Count itDo the totals match the source system or the finance number?
Check the edgesMinimums, maximums, blanks, zeros and dates in the future.
Check the definition"Active customer" means something different in every team.
Check the historyDid a definition or system change during the period?
Ask who made itData is collected by people with incentives.
1.4Separate signal from noise

Know how much a metric normally moves before you explain a move.

  • Compare like with like: same season, same length, same population.
  • A 20% rise from five customers is five customers.
  • Look at forty cuts of the data and one will look interesting by chance. Say how many you looked at.
1.5Correlation, cause and what you can claim

Match your claim to your evidence. These are three different sentences.

Habit to build: before presenting, write down the two best alternative explanations and say what you did about them.

1.6Say how sure you are New

Give a range, say how confident you are, and name what would change your mind.

WeakChurn will fall 4.2% next quarter.
BetterWe expect churn to fall by 2 to 5 points. We are fairly confident in the direction and less sure of the size, which depends on how many customers take the offer.
0246best guess
02Draw it honestlyCharts people read correctly
2.1Pick the right chart

Choose the chart from the question, not from what the software offers.

2.2Credibility killers

One bad chart makes people doubt the other twenty.

Us: 96
Rival: 100

3D effectsThey distort the thing you are measuring.
Dual axesYou can make any two lines look related.
Eight-slice piesNobody can rank them.
Cherry-picked datesStart where it suits you and someone will notice.
Missing unitsPounds, thousands, percent or percentage points?
2.3Guide the eye

Start grey. Add one loud colour for the thing that matters.

  • Delete gridlines, borders and legends you do not need.
  • Label bars and lines directly, so the eye does not ping-pong to a legend.
  • Use size, bold and position sparingly. Each one is a vote for attention.

Squint test: blur your eyes. If the one thing that matters is not the first thing you see, redraw it.

2.4Give it context

A number alone means nothing. Add one reference.

62% satisfaction could be good, bad or catastrophic. Against last year and the target, it is a modest gain that still misses the goal.

Annotate the chart itself. A short note at the point where something changed ("price rise", "new billing system") does more than a paragraph beneath it.

2.5A case for tables

Charts show patterns. Tables let people find their own number.

RegionSales (£k)Change
North1,240+4.1%
Midlands980-2.6%
South1,615+7.3%
  • Right-align numbers and use the same decimal places.
  • Keep rules light, or use none and rely on spacing.
  • Highlight the row that matters. Do not colour everything.
2.6Small multiples New

More than four or five lines? Split them into a grid.

Use the same scale on every panel, order panels by size or pattern rather than alphabet, and give the panel you are discussing the highlight colour.

2.7Colour that survives real eyes

Never rely on red against green alone. Some of your audience will not be able to tell them apart.

Red and green

Bad
Good

A safer pair

Bad
Good

Colour-blind-safe set (Okabe-Ito). Tap to copy.

A navy, pink and cream palette you can borrow

03Land it with peopleWhere most analysis dies
3.1Write titles that say so what

A topic title says where to look. An insight title says what to think.

Customer churn by tenure

Revenue by channel

Satisfaction trend

Read only your slide titles, in order. They should tell the whole story on their own.

3.2Problem, evidence, implication, decision, action

Lead with the answer, then give the reasoning.

Senior people can always ask for more detail. They cannot get back the five minutes you spent building to a reveal.

3.3Presenting under pressure

Know your one sentence.

Cut to 30 secondsWhat would you say? Say that first.
Backup in the appendixNot in your head.
Pause after the key numberSilence reads as confidence.
"I will find out by Thursday"A perfectly good answer. Then do it.
3.4Handling the sceptic

Challenge is a gift, even when it does not feel like one.

Work out who is likely to object, and why, before you walk in. Sometimes the objection is about the data. Often it is about what the finding means for their budget or their team.

Name the limitSay it before they do.
Ask what would change their mindIf the answer is "nothing", it is not a data conversation.
Agree definitions earlySources and definitions, ahead of the meeting.
3.5When you have lost the room

Stop, re-anchor, and go straight to the recommendation.

You will see it: phones come out and questions drift. Cut the detail. You can always send the rest afterwards.

"Let me step back. The decision we need today is..."

04ToolkitUse it today
4.1Pre-flight checklist

Run this before anything leaves your hands.

0 of 12 checked

Your ticks are saved in this browser only.

Free to use and share with credit. Practical guidance, not peer-reviewed research. Spotted an error or want a topic covered? Tell me.

Want to go further?

Working on something tricky?

If you are wrestling with how to analyse, measure or present evidence in your organisation, I would like to hear about it.

Get in touch