Build brilliant Power BI reports with a disciplined schematicLesson 4 of 8
The five page archetypes
Give overview, trends, breakdown, exceptions and drillthrough distinct jobs.
What you'll do
Assign five analytical questions to the spine's page archetypes and sketch a two-page synthetic report. You will preserve context across navigation and define honest fallbacks. The examples teach the default schematic; the v2 archetype CLI integration remains Preview.
Give each page a distinct job
An archetype is a question and an information arrangement. It does not guarantee that the model contains the fields or qualified measures needed to answer it. Start from the reader's decision, then check availability before reserving panels.
| Archetype | Question | Main arrangement |
|---|---|---|
| Executive Overview | Are we succeeding, and where must we act? | KPI row, dominant trend, ranking, queue |
| Trends & Drivers | When did it change, and which segments moved? | Primary trend, variance, comparable segment trends |
| Breakdown/Composition | Where does the result come from? | Ranking, composition, Matrix |
| Detail & Exceptions | Which records require action? | Population KPIs, detail, severity, aging |
| Drillthrough | What happened to this entity? | Identity, two cards, history, transactions |
The usual page order is Overview, Trends & Drivers, Breakdown, then Detail & Exceptions. Drillthrough is hidden from ordinary navigation and entered contextually. Do not repeat the full KPI band everywhere if it displaces the analysis needed for that page's question.
Orient, explain and compare
Executive Overview uses four contextual KPI slots and a dominant outcome history. A support ranking highlights adverse impact when qualified variance exists; otherwise rank actual contribution. Without valid time, use a ranked outcome comparison instead of a fabricated trend. Without a validated exception predicate, call the queue Top contributors.
Trends & Drivers expands temporal analysis. Signed variance belongs in chronological order, while drivers use meaningful ranked categories. Comparable small multiples share scales and require supported native mappings. A volume/rate combination needs an explicit paired question and named axes; it does not establish causality.
Breakdown compares related business dimensions. A composition visual requires exclusive additive parts or qualified shares. For overlapping counts or ratios, use ranked comparisons and recompute totals in the correct scope. Other means the excluded population, not the difference between rounded percentages and one hundred.
Make detail actionable
Detail & Exceptions shows a declared predicate, readable records and an All records route. Use roughly six to eight useful columns: entity, date, owner/status and relevant exposure or comparisons. Sort severity, exposure and then a stable entity key. Aging bins follow numeric business order.
This is a fictional schematic. Its layout illustrates the question; its sample values do not constitute an observed report. When no validated exception predicate exists, use Detail and omit invented exception counts and rates. Vertical detail scrolling can be appropriate; horizontal table scrolling is not the default answer to too many columns.
Keep drillthrough contextual
Drillthrough requires one entity, a stable field binding and native Back. An unfiltered or multi-entity entry must not display a false single-entity title. History and comparisons use that entity's scope, and the supporting transactions stay within authorised access.
Interaction exercise:
Select a territory on Overview.
Open Detail with the intended period and business scope.
Drill through to one entity, then use Back.
Reset and verify the declared default state returns.Produce the review packet
Sketch Overview plus the page that best supports its next question. Name the fields, comparisons, default sorting and missing-data behaviour for each slot. Mark unsupported pages suppressed and remove their navigation targets. Do not preserve an empty page simply to complete the five-page set.
Your output is a page map and an interaction checklist. Separate schema/layout checks from actual click evidence and rendered-value checks. A well-chosen archetype improves explanation; qualification still applies only to declared measures, observed states and retained evidence.
Check your understanding
- When should Top contributors replace an Exceptions queue?
- Why should a ratio breakdown avoid a casually calculated Other category?
- What must happen when drillthrough receives no single-entity context?