How to present data to executives so they decide
The short answer
To present data so executives decide, start from the decision and the decider, then write the message as one sentence with a number. Check the data for unfair comparisons before charting. Choose the chart people read most accurately, title it with the finding, keep bar axes at zero, make it accessible, and put the ask and recommendation on page 1.
Want the tools, not just the method? The The Data Storytelling and Dashboard Kit for Managers kit turns this guide into ready-to-use templates, spreadsheets and printables, with an instant download.
- No analysis until someone names the decision, the decider and the date
- Position on a common scale is read most accurately; angle and area least
- Truncated bar charts exaggerate differences even when viewers are warned
- Titles steer what people take as the message: state the finding
- Colour is never the only cue; text 4.5:1 and chart marks 3:1 under WCAG 2.2
- Dashboards for recurring questions; one-page memos for one decision
What are the steps to present data to executives?
The method has ten steps. The first three happen before you draw anything, and they do most of the work. Teams that skip them usually produce long packs of accurate charts that answer nobody's question, then spend the meeting arguing about definitions.
Each step is small. The decision brief takes about fifteen minutes with the person who asked. The big idea takes as long as it takes to write one honest sentence. Data checks are a short routine you run on every dataset, not a research project. The time comes back later, because you build fewer charts and the meeting decides.
- 1. Name the decision, the decider and the date (a decision brief).
- 2. Write the big idea: one sentence with a point of view and what is at stake.
- 3. Check the data: source, definitions, completeness, outliers and fair comparisons.
- 4. Choose the chart from what it must show.
- 5. Design it: a finding title, direct labels, one highlight colour, accessible contrast.
- 6. Keep it honest: zero-based bars, labelled zooms, one axis.
- 7. Build the story answer-first: a one-page memo or a short deck.
- 8. Put recurring questions on a small dashboard with a KPI dictionary.
- 9. Run the meeting as a decision meeting with a pre-read 48 hours ahead.
- 10. Rehearse the hard questions and pre-wire the sceptic.
Why start from the decision instead of the data?
Every chart answers a question. If nobody chooses the question, the chart answers whatever the data makes easy, and packs grow to dozens of slides that are each true and collectively useless. A decision has three parts you can write down: a verb such as approve, stop or fund; a person who decides; and a date.
Official guidance points the same way. The UK Government Analysis Function asks every chart to carry a headline title that communicates its message, and Statistics Canada asks whether your audience needs to make decisions or simply be informed before you choose a visual. The most useful question to ask the requester is: what evidence would change your mind? It tells you which chart to build.
Write the brief in nine fields: the request in the asker's words, the decision, who decides and who influences them, the decision date, what the audience already believes, what evidence would change their mind, a draft big idea, the format, and the data you need with its owner. If the asker cannot name a decision, write the question you will answer instead and confirm it by email. Keep the answer short.
Which chart should you use for which message?
Start from what the chart must show: change over time, comparison, ranking, part-to-whole, distribution, relationship or variance to target. Each has a best first choice.
The ranking behind those choices comes from research. In 1984 William Cleveland and Robert McGill found that people judge position along a common scale most accurately, then length, direction and angle, then area, with shading and colour saturation last; position judgements were about 1.96 times as accurate as angle judgements. Jeffrey Heer and Michael Bostock replicated the ranking in 2010 with crowdsourced participants. That is why sorted bars usually beat pies for comparison.
Two practical rules follow. Use position and length for the comparison that matters, and save colour and size for directing attention, because the eye picks out one strong colour almost instantly but reads amounts from colour poorly. And when two charts are equally valid, choose the one your audience already reads every month: familiarity cuts reading time.
| The chart must show | Best first choice | Avoid |
|---|---|---|
| Change over time | Line chart | Pie; dual axes |
| Compare categories | Bar chart sorted by value | Pie; 3-D |
| Ranking | Sorted horizontal bar | Alphabetical order |
| Part-to-whole | Sorted bars or one stacked bar | Many-slice pies |
| Distribution | Histogram or box plot | An average alone |
| Relationship | Scatter plot | Two lines on two axes |
| Variance to target | Diverging bar of actual minus target | Truncated bars |
What does a worked example look like?
Take a fictional six-store grocery chain in Calgary whose margin has slipped by a point in a year. Shrink, the stock lost to waste, damage and theft, rose from 2.8% to 3.5% of sales. The chain total suggests a chain-wide problem. Split by store, the picture changes: two stores opened in 2026 carry almost all the excess.
The big idea that follows: shrink at the two newest stores costs about CAD 42,740 a month above target, and moving them to the ordering method the other four use should recover about CAD 32,660 a month by June. The chart is a sorted bar of shrink by store with the two new stores highlighted, titled with that finding.
Notice what the split does. A chain-wide fix would have spent money at four stores that were already near target. The store chart points the money at the two stores where it can recover most, and it gives leaders a number to weigh against the cost of a pilot. The table below is the data behind the chart; the right-hand column is shrink above target in dollars a month, the figure that turns a percentage into a decision.
| Store (March) | Net sales (CAD k) | Shrink % | Above 3.0% target (CAD a month) |
|---|---|---|---|
| Calgary NW | 1,480 | 2.9 | 0 |
| Calgary SE | 1,310 | 4.6 | 20,960 |
| Airdrie | 1,260 | 2.8 | 0 |
| Chestermere | 1,210 | 4.8 | 21,780 |
| Okotoks | 1,170 | 3.1 | 1,170 |
| Cochrane | 1,050 | 3.0 | 0 |
| Total | 7,480 | 3.5 | 43,910 |
How do you keep charts honest?
Start bar and column charts at zero. Bars are read by length, so a cut axis inflates differences. Brenda Yang and colleagues at Duke University found the effect in five studies, in 83.5% of participants, and it persisted after people were taught about truncation. Michael Correll, Enrico Bertini and Steven Franconeri found that explicit visual cues of truncation did not remove the effect.
Line charts are read by position and slope, so a zoomed axis can be honest when the change that matters is small, provided the axis is labelled and the full range is available. Avoid dual axes, which can make any two series look linked; size circles by area, not radius; and do not pick a start date that flatters the trend.
When someone asks you to start a bar axis higher so a small gap shows, offer an alternative instead of a refusal: plot the difference itself (the gap to target in points) as bars from zero, use a dot plot, or show the trend as a labelled line. Keep the full-range version in the back-up so you can answer a challenge in the room in one sentence.
Reading helps; measuring tells you what to work on. These AI-graded assessments on AssessAll pair with this topic:
How do you check the data before charting it?
Check the source, the definition, completeness, outliers and whether totals reconcile. Then check the comparisons. Totals can reverse group patterns when the mix changes, as in the 1973 Berkeley admissions data, where pooled figures suggested bias against women but department-level data did not. Small groups produce extreme results by chance. Unusually bad results tend to improve on their own, so a 'fix' applied to the worst unit can look like it worked when it did not.
When two things move together, write 'rose at the same time as', not 'caused', unless you have a test.
Record the checks so others can trust them. A simple status per dataset (READY or CHECK) is enough: READY needs an agreed definition, outliers looked at, a like-for-like comparison and nearly all the expected rows. A dataset marked CHECK can still be shown, but only with a plain-words caveat on the page, never silently.
How do you make charts accessible?
About 1 in 12 men and 1 in 200 women of Northern European ancestry have red-green colour vision deficiency, so colour must never be the only cue. WCAG 2.2 sets the reference: text contrast of at least 4.5:1 (3:1 for large text), chart marks needed to understand the content at least 3:1, and a text alternative for every chart. WCAG 2.2 was approved as the international standard ISO/IEC 40500:2025 in October 2025.
Practical steps: label lines and bars directly, add the word to every traffic light, use a tested palette such as the UK Government Analysis Function's, check every colour pair, and write alt text that gives the message rather than describing shapes.
Which laws require WCAG, for whom and from when differs by market and changes: public-sector rules in the UK, New Zealand and Canada, the European Accessibility Act for some products and services in the EU, and the ADA Title II web rule for state and local governments in the US. Treat accessibility as good practice for every internal pack and check the legal position with your adviser for anything published.
What are the most common mistakes?
Most mistakes are habits, not gaps in skill. Each one below has a simple fix in the method.
- Starting analysis before anyone names the decision.
- Topic titles ('Sales by store') instead of findings.
- Pies with many similar slices.
- Truncated bar charts 'fixed' with a warning label.
- Red and green as the only signal.
- Two definitions of the same KPI.
- Dashboards with a tile for every request.
- Walking through the pack in the meeting instead of deciding.
- Publishing AI-drafted numbers without re-checking them.
When should you use a dashboard, a memo or a deck?
Use a dashboard for questions that recur every week or month; the Office for National Statistics notes that dashboards suit high-level, frequently updated indicators and are not well suited to explaining specific findings. Use a one-page memo for a single decision with a deadline: the ask, options including do nothing, a recommendation, two charts and the risks. Use a deck when a group needs to build agreement live, with slide titles that tell the story on their own.
Whatever the format, put the ask first. A busy reader on a phone should know what you need from them in the first ten seconds.
For a monthly business review, two gates protect the meeting. A readiness check before the pre-read goes out (any critical gap means no-go), and a pre-read sent at least 48 hours ahead so the meeting can start with questions instead of a walk-through. Log every decision the same day with an owner and a date.

Put the method to work this week
The Data Storytelling and Dashboard Kit gives you the decision brief, chart chooser, Excel dashboard, MBR pack and scripts to apply every step above, with a worked example to copy.
Sources
- UK Government Analysis Function (Office for National Statistics): Data visualisation: charts
- UK Government Analysis Function (Office for National Statistics): Data visualisation: colours
- World Wide Web Consortium (W3C): Web Content Accessibility Guidelines (WCAG) 2.2, W3C Recommendation
- Statistics Canada: Data Visualization: An introduction (catalogue 892000062020014)
- Office for National Statistics (ONS) Service Manual: Dashboards (data visualisation guidance)
- US National Library of Medicine (MedlinePlus Genetics): Color vision deficiency
- Journal of the American Statistical Association: Cleveland, W. S. and McGill, R. Graphical Perception: Theory, Experimentation, and Application to the Development of Graphical Methods. JASA 79(387), 531-554
- University of Washington Interactive Data Lab (ACM CHI 2010): Heer, J. and Bostock, M. Crowdsourcing Graphical Perception: Using Mechanical Turk to Assess Visualization Design. CHI 2010, 203-212
- Duke University Scholars (Journal of Applied Research in Memory and Cognition): Yang, B. W., Vargas Restrepo, C., Stanley, M. L. and Marsh, E. J. Truncating Bar Graphs Persistently Misleads Viewers. JARMAC 10(2), 298-311
- arXiv / ACM CHI 2020: Correll, M., Bertini, E. and Franconeri, S. Truncating the Y-Axis: Threat or Menace?
- Science (copy hosted by University of Iowa): Bickel, P. J., Hammel, E. A. and O'Connell, J. W. Sex Bias in Graduate Admissions: Data from Berkeley. Science 187, 398-404
- University of Maryland (ACM CHI 2018): Kong, H.-K., Liu, Z. and Karahalios, K. Frames and Slants in Titles of Visualizations on Controversial Topics
More from the Bodhih family
Questions people ask next
How long should a data presentation to executives be?
As short as the decision allows. A one-page memo of 350 to 450 words with two charts suits most single decisions; a monthly review pack works well at around nine pages with back-up behind it.
Should every chart have a title that states the finding?
Yes. Research shows titles steer what readers take as a chart's main message, so state the finding accurately and put the measure, unit and period in a subtitle.
Are pie charts ever acceptable?
Occasionally: UK government guidance allows them for five or fewer categories where one is clearly dominant. For comparing similar values, sorted bars are read more accurately.
Is it ever acceptable not to start an axis at zero?
For bar and column charts, no. For line charts, a zoomed axis can be honest if it is labelled and the change shown is the change that matters.
How many KPIs should a dashboard show?
Keep the first screen to the few KPIs that answer leaders' recurring questions; eight or fewer is a practical limit. Every KPI needs an owner, formula, target and threshold.
Can AI tools help with data storytelling?
Yes, for drafting briefs, titles, alt text and outlines. But AI can state false numbers confidently, so re-check every figure and never paste personal data into unapproved tools.
What tools do I need?
A spreadsheet is enough to start. Excel, Google Sheets, Power BI, Looker Studio and Tableau all support the method; the skill is in the steps, not the software.