Our board pack used to run to sixty one pages. Twenty two of those were charts. In the meeting, members referred to four of them.
We now produce nineteen pages with nine charts, and all nine get discussed. Here is what changed in how we brief Data Visualization Services for executive reporting.
A board pack is not a data warehouse tour
The habit that produced sixty one pages.
Each function sent its own section with its own charts, because each function wanted its work represented. The pack grew every year and nobody removed anything.
The result was a document that showed everything and argued nothing. Board members read the summary and skipped the rest, which is a reasonable use of their time.
One claim per page
The rule that cut the pack by two thirds.
Each page now states a single claim at the top, in a sentence. The chart below it exists to support or qualify that claim.
If a page cannot be summarised in one sentence, it is two pages or it is not needed. That test removed eleven pages in the first attempt.
Show the number, the target, and the trend
Three things, always together.
A value alone invites the question. A value with a target and a direction of travel answers it before it is asked.
Our previous charts frequently showed a line with no target on it. Members would ask whether that was good, and the answer took three minutes of explanation that should have been printed on the page.
Experienced Data Visualization Services practitioners treat this as a baseline requirement rather than an enhancement.
Put the definition on the page
Not in an appendix.
Each number carries a short note saying how it is calculated and from what period. Our revenue figure has a footnote about treatment of deferred items, because that treatment changed and the change explains an apparent dip.
Before this, half the discussion time went on what the number meant rather than what to do about it.
Keep the format identical every month
Boards read by pattern recognition.
Our pack has the same pages in the same order with the same charts in the same positions every month. Members know where to look, so they spend their attention on the change rather than the layout.
Redesigning the pack to make it fresher is a service to the person producing it and a disservice to everybody reading it.
Make bad news as visible as good news
Credibility depends on it.
We have a standing page for things that moved against us, with the same prominence as the page on things that moved in our favour.
Our previous packs buried difficulties in the commentary. One board member noticed and started reading the commentary first, which is not the outcome anybody wanted.
Build on one modelled layer
The technical precondition.
Our sixty one page pack was assembled from nine source systems by four people over a week, with figures that occasionally disagreed between sections.
An AWS data warehouse modernization gave us one modelled layer where the definitions are applied once. The pack now assembles in a day, the sections agree, and the arguments about whose number is right have stopped.
Separate the pack from the operational dashboards
They serve different readers.
Our operational dashboards are detailed, filterable, and refreshed hourly. The board pack is nineteen fixed pages produced monthly with written commentary.
Trying to serve both audiences with one artefact produced something too shallow for operators and too busy for the board. Good Data Visualization Services work starts by asking which of the two you want.
Be careful with generated commentary
A specific caution.
We trialled several AI Tools For Business to draft the narrative around each chart. The drafts were fluent and occasionally stated a cause the data did not support.
We kept them for the descriptive paragraphs and write the interpretation ourselves, because the interpretation is the part the board is actually buying.
Review which pages were discussed
The feedback loop almost nobody runs.
After each meeting we record which pages were referred to. A page not discussed for three consecutive meetings is reviewed and usually removed.
That single practice is how sixty one pages became nineteen. Nobody had to argue for cuts, because the record made the case.
What our brief specifies
- One claim per page, written as a sentence.
- Every number with a target and a trend.
- The definition printed beside the number.
- An identical layout month to month.
- Equal prominence for unfavourable movement.
- One modelled data layer underneath.
- A record of which pages were discussed.
Common questions
How long should a board pack be?
Shorter than yours is. Ours works at nineteen pages for a business that previously sent sixty one.
Who writes the commentary?
The accountable executive, not the reporting team and not a tool. The numbers can be assembled for you. The interpretation cannot.
Should the pack be interactive?
Ours is not. Fixed pages read in a meeting beat a dashboard nobody opens beforehand.
How do you handle a number that changed definition?
Print both, state the change, and keep the old series for a year. Quietly restating history destroys trust in the whole pack.
Who should choose the charts?
The reader, with help. Data Visualization Services practitioners can propose the form, and the executive accountable for the number approves what appears.
How should this be briefed?
In writing, with the claim per page specified. The best Data Visualization Services engagements we have run started from the claims and worked back to the charts.
Can tools help with assembly?
Yes, and that is where we use them. Several AI Tools For Business handle the descriptive assembly well and the causal reasoning badly.
Where we ended up
Nineteen pages, nine charts, one claim each, one modelled layer, and a record of what actually gets discussed.
The pack is less impressive to look at than the sixty one page version. It is read, which the previous one largely was not.
