Business Intelligence with Power BI

L09 · Interactivity

Interactivity

Learning objectives. By the end of this lesson you will be able to:
  • Distinguish the three filtering surfaces — slicers, the Filters pane, and cross-filtering — and choose deliberately among them
  • Configure visual interactions (filter, highlight, none) per pair of visuals
  • Separate drill down (levels within one visual) from drillthrough (a page about one thing)
  • Build bookmarks and buttons that switch views without confusing state
  • Design for the actual reader: exploration pages versus answer pages

Estimated time: 60–75 minutes of reading and follow-along, before practice.

Why this matters

Since Lesson 1 you have known the punchline: a published report is interrogated, not read. This lesson is where interrogation becomes a designed experience rather than a happy accident. The stakes are real in both directions. Underbuilt, a report answers only the exact question its author anticipated. Overbuilt — five slicers, three drill paths, bookmarks everywhere — it becomes a cockpit no VP will touch, and a reader who accidentally leaves a filter on somewhere ships a wrong number to a board deck. Interactivity is power steering: the reader drives, and you decide what the car can do.

The three filtering surfaces

Students conflate these constantly; keep them separate and half the lesson is won:

SurfaceWho controls itVisibilityUse it for
SlicerThe readerOn the page, state always visibleThe dimensions readers legitimately explore: region, year, category
Filters paneAuthor (mostly)Collapsed by default; per-visual, per-page, or all-pagesDesign-time scoping: "this page is 2026 only," "this visual excludes returns"
Cross-filteringThe reader, by clicking visualsInvisible until you notice everything movedFree exploration — every visual is secretly a filter control
The invisible-filter problem A slicer's state is visible; a Filters-pane selection and a clicked-bar cross-filter are easy to miss. The classic failure: a reader clicks the Furniture bar, forgets, reads the card, and reports Furniture's number as the company's. Two defenses, both authorial: keep reader-facing filtering in visible controls (slicers) wherever possible, and — the professional's habit — every page carries a clear-filters path (a reset bookmark button, next section). Design as if every filter will one day be forgotten, because it will.

Visual interactions: filter, highlight, none

Click the Furniture bar and every other visual reacts — but how they react is a choice. Filter recomputes the target to Furniture only. Highlight keeps the full total visible and saturates Furniture's share within it — honest for part-to-whole contexts, ambiguous for cards. None disconnects the pair entirely. Edit interactions (Format → Edit interactions) sets this per pair of visuals, and the default is not always right: a KPI card being "highlighted" shows an unchanged number over a changed context — readers misread it, and switching that pair to Filter (or None, with a reason) is routine craft. The discipline: for every visual, decide what a click elsewhere should do to it, and make the boring explicit choice.

Drill down: levels within one visual

Give the category bar a hierarchy — Category → Subcategory — and the drill arrows appear. Drill into Furniture and the bar re-renders as its four subcategories: Storage $6.81M, Bedroom $6.23M, Seating $5.89M, Tables $5.87M — a remarkably even split, which is itself a finding (no single furniture line carries the category; compare Lighting, where you already know one cheap lamp family drives volume). Drill down answers "what is inside this bar?" without leaving the visual. Two mechanics worth teaching your hands: the split-arrow expands all categories at once (the L01 date hierarchy move), while click-drill descends into one; and the visual's header shows which level you are on — read it before you quote a number.

Drillthrough: a page about one thing

Drill down stays in the visual; drillthrough navigates: right-click the Furniture bar → Drill through → "Category Detail," and the reader lands on a dedicated page already filtered to Furniture — its monthly trend, its subcategory table, its return rate, its top products. Build one detail page and every category bar in the report becomes a doorway to it. The craft requirements: the drillthrough page declares its filter loudly (a title measure like "Detail: " & SELECTEDVALUE(Products[Category]) — one new function, doing exactly what its name says), and the automatic back button stays where readers can find it. Drillthrough is the answer to "can I click into this?" — the single most asked question in any dashboard review.

[Screenshot l09-drillthrough — the Category Detail page filtered to Furniture: dynamic title, subcategory bars (Storage $6.81M leading), monthly line, back button]
A page about one thing. The dynamic title is the page saying which thing.

Bookmarks and buttons: designed states

A bookmark captures a page state — filters, slicer selections, which visuals are visible — and a button applies it on click. Three uses cover nearly all honest needs:

  • Reset. A "Clear filters" button applying a bookmark of the pristine page — the invisible-filter problem's antidote, on every exploration page.
  • View switching. "Revenue view / Units view" buttons swapping which visuals are visible (bookmarks can show/hide). Two claims, one page, reader's choice — remember Lesson 1's units-versus-dollars: this is that lesson as UI.
  • Guided states. "Show me the December story" — a bookmark that sets the filters a narrative needs. Used sparingly, this is storytelling; used everywhere, it is a slideshow wearing a dashboard's clothes.

The failure mode is state confusion: bookmarks that capture data filters when you meant only display changes (the bookmark's options control this — "Data" vs "Display"), and pages where a bookmark leaves slicers visually stale. Test every bookmark the way a hostile reader would: click it after doing something weird.

Exploration pages versus answer pages

The lesson's design frame: every page is one or the other. An answer page (the board pack) states findings — few interactions, maybe a year slicer, titles doing the talking. An exploration page (the analyst's cockpit) hands over the keys — slicers, cross-filtering, drillthrough. The classic dashboard sin is building one page for both audiences and serving neither: executives lost in controls, analysts starved of them. When in doubt, split the page and let each be honestly itself. (Your capstone will be graded on exactly this separation.)

Knowledge check. A reader reports "the dashboard says revenue is $24.8M" — but the card should read $66.81M. No slicer shows a selection. What most likely happened, and which design habit prevents the escalation?
Knowledge check. Merchandising wants to click any category anywhere and land on a full page about it — trend, subcategories, returns, top products. Drill down or drillthrough, and why?
Knowledge check. On drilling into Furniture, its four subcategories land within $1M of each other (Storage $6.81M to Tables $5.87M). A colleague calls the drill "a dead end — nothing to see." The better read:

Common mistakes

  • Filter state only in the Filters pane. Readers do not open it. Visible state or a reset path — ideally both.
  • Default interactions everywhere. Highlighted cards mislead; audit every pair with Edit Interactions at least once per page.
  • Drillthrough pages that whisper their filter. The dynamic title is not decoration; it is the page's honesty caption — the map lesson again.
  • Bookmark sprawl. Reset, view-switch, and a few guided states. A page needing ten bookmarks is a slideshow that should be one.
  • One page for both audiences. Cockpits for analysts, answers for executives. Split rather than compromise.

Summary

  • Three filtering surfaces: slicers (visible, reader's), the Filters pane (authorial scoping), cross-filtering (invisible, everywhere). Design as if every filter will be forgotten.
  • Interactions are per-pair choices — filter, highlight, none — and highlighted cards are the default worth overriding.
  • Drill down descends within a visual (Furniture → four ~$6M engines); drillthrough lands on a page about one thing, with a dynamic title as its confession.
  • Bookmarks + buttons = reset paths, view switches, guided states — and nothing more.
  • Every page is an exploration page or an answer page. Decide which, and build it as that.