Industry
Part of Reporting: what beginners should know in 2027
Reporting examples worth studying before you start
Reporting shapes worth knowing: twelve recurring report types, the question each one answers, and what each of them costs to produce every period.
Most recurring reports are one of about twelve shapes. Knowing the shapes matters because each one answers a different question, carries a different maintenance cost, and fails in a different way.
What follows is a catalog rather than a ranking. Read it to find the shape that matches the question you were actually asked, which is often not the shape that was requested.
What to take away
- The request almost always names a format. The right response names a shape, and the two rarely match on the first pass.
- Cost per period varies by an order of magnitude across these twelve. Nobody asks about that until year two.
- Four of the twelve are designed to be read only when something is wrong. Those are the cheapest and the most neglected.
The twelve shapes
1. The operational tally. A short list of counts for a shift or a day: volume in, volume out, backlog, oldest open item. Read at a fixed moment by people with authority to move work. Fails when it grows past one screen.
2. The exception list. Only the items outside a rule: breaches, overdue cases, accounts past a limit. An empty page is the good outcome, which makes it the one shape people forget to check when it stays empty for a while.
3. The statistical control view. The same measure plotted against limits derived from its own past variation, so ordinary noise stops triggering conversations. The introduction to control charts in the NIST handbook explains the reasoning better than any internal document will, and the reasoning is the part that transfers even when the arithmetic is simplified.
4. The period pack. The monthly or quarterly management document: primary numbers, comparisons, commentary, appendix. The most expensive shape to maintain and the one most likely to be produced long after anyone reads it.
5. The variance report. Actual against plan or against forecast, decomposed into the parts that explain the gap. Useful only if the decomposition is real; without it, this shape is two columns and a subtraction.
6. The cohort view. A group defined by when it entered, followed as it ages, with incomplete cells marked. It answers whether something is genuinely improving or whether the mix simply changed, which a period-over-period view cannot. The logic is the same as a cohort study, and so are the traps: an incomplete cohort read as if it were complete produces a confident wrong answer.
7. The reconciliation. Two independently produced figures shown side by side with the bridge between them. It exists to be boring. When it stops being boring, something upstream broke, and it has done its job.
8. The funnel or stage view. Counts at each stage of a process with the drop between stages. Its failure mode is time: the stages are measured at one instant, so a fast-moving process and a slow one look identical.
9. The ranked list. Top or bottom items by some measure, used to decide where to intervene. Sort by the measure, never alphabetically, and always show how many rows are below the cut.
10. The distribution view. Not the average but the spread: how many cases fall where. This is the shape most often skipped and the one that changes the most minds, because a mean and a median that disagree are visible in a way that no summary row makes them.
11. The freshness and failure report. Which feeds landed, when, and which did not. Nobody requests this. Every estate with more than a handful of reports needs it, because the alternative is discovering a stale number through a reader.
12. The retirement register. What was decommissioned, what replaced it, and what is scheduled to go. It is the only shape that shrinks the other eleven.
Matching a shape to the question
| The question underneath | Shape that answers it | Shape usually requested |
|---|---|---|
| Is anything wrong right now? | Exception list, control view | Period pack |
| Why did the total move? | Variance report with decomposition | Two lines on one chart |
| Is this genuinely improving? | Cohort view | Period over period |
| Where should we intervene? | Ranked list | Full table of everything |
| Do these two systems agree? | Reconciliation | A meeting |
| Is the typical case what we think? | Distribution view | An average |
The right column is not stupidity. It is what people ask for when they have only ever been shown one shape, which is why offering a second shape in the scoping conversation changes more outcomes than any improvement to the first.
Cost per period, honestly
Three of these are nearly free once built: the exception list, the freshness report, and the control view, because each runs unattended and produces nothing when nothing is wrong. Three are expensive forever: the period pack, the variance report, and the reconciliation, because each needs a human to write or check something every cycle.
Budget accordingly. A team that has agreed to four period packs has committed most of a person, and that commitment is invisible in every plan until somebody counts it.
Before you build any of them
Write down which shape you chose and why the alternative was rejected. That single sentence is what lets a successor change the shape later without relitigating the whole request. Chart selection follows the same discipline, and the guidance on choosing charts published by the Government Analysis Function is a good external reference to settle arguments about the picture once the shape is agreed.
Related reading on this site
The production and close discipline behind all twelve is in building a report worth keeping. For panel shapes on a screen rather than report shapes in a pack, see twelve panel shapes. Deciding what promise a given artifact carries is covered in the three publication tiers, the cohort and variance shapes lean on the definitions work in writing a metric definition, and the variance shape only makes sense alongside how forecasts are built and reviewed.
Common questions
Can one artifact be two shapes?
Usually badly. The common pairing that does work is an exception list with a link into a fuller table, because the second shape is not on the page until the reader asks for it.
We were asked for a dashboard and I think the answer is an exception list. How do I say that?
Offer both for one period and ask which one they opened. That conversation is short and it settles itself, and the loser is retired with no argument.
Which shape should a new team build first?
The freshness and failure report, before anything anyone requested. Every other shape rests on knowing whether the data arrived, and it is the one nobody thinks to ask for.
Is the retirement register really a report?
It is, and it should be shown to whoever funds the work. A list of things that no longer need to exist is the clearest evidence a team can produce that it is not simply accumulating obligations.
