Business intelligence development reference
Design single-screen monitoring views that show what’s happening, what needs attention, and where to go next.
A dashboard puts the essential signals for a recurring decision on one screen. Its readers can see what is happening, spot what needs attention, and decide whether to act or investigate.
Start with the monitoring task. Keep the signals that change a decision, give each one enough context to interpret it, and put the most consequential information where the eye lands first. Treat the dashboard as the front door to deeper work. Few’s dashboard writings provide the foundation.
| What the reader needs | Build this |
|---|---|
| Check a known set of signals repeatedly and decide whether to act | A dashboard. Keep the essential signals, comparisons and exceptions visible together. |
| Read a narrative, reconcile an official figure, or audit complete rows | A report with commentary, fixed reporting scope, and the appropriate detail or export. |
| Ask new questions, change dimensions, or test competing explanations | An exploration workspace with coordinated views and inspectable detail. Link to it from the monitoring screen when the same audience uses both. |
| Navigate different jobs, edit records, assign work, or complete a sequence | A multi-page app with clear navigation and task-specific screens. Give it a dashboard landing view only if there is a recurring monitoring job. |
Name the audience, recurring decision and action. Keep a first-screen item only when the brief identifies the decision it supports.
Move exhaustive records, open-ended exploration and multi-step workflows to the appropriate companion surface. Link from the relevant summary.
Start a requirements meeting with a recent decision. Ask what evidence made someone act, and sketch only the information they used.
For a new build, choose the type, write the brief, agree on measures, and sketch the screen before configuring charts. For a review, use the release checklist and cite the rule ID in each finding. Every numbered rule has a direct link on its ID.
How levels are assigned. Team addition A rule is MUST when breaking it can mislead a reader, hide a signal the decision needs, block access for some readers, or skip a release control. A rule is SHOULD when it is a strong default with legitimate exceptions, or when a MUST rule elsewhere already enforces its core. A rule is GUIDANCE when it calls for judgment rather than a pass-or-fail check. Give new rules a level with the same test.
Attribution. Few supplies the monitoring method, the emphasis on comparison and visual hierarchy, and the compact chart patterns. Team addition marks an implementation choice, an extension, or a release policy specified here. The IDs and rule levels are this standard’s own. The dashboard labels in §03 are used as a working classification. Every example, figure and dataset is original and invented.
| I need to… | Go to |
|---|---|
| Decide what kind of dashboard to build | 03 Dashboard types |
| Interview a stakeholder or write the brief | Interview questions and the brief template |
| Pick a chart | Selection guide and the chart gallery |
| Arrange the screen | 08 Layout and the full-screen wireframes |
| Look up a color or hex value | 09 Color standards |
| Look up a rule by its ID | Rule index |
| Prepare for a design review | Pre-release checklist |
Choose by the decision clock and the first question the screen must answer. The cadences below are examples of how the types differ. Agree on the actual refresh and acceptable delay in the brief.
| Dimension | Strategic | Operational | Analytical monitoring | Tactical Team addition |
|---|---|---|---|---|
| Audience and question | Leaders and program owners. Are we on course to meet our objectives? | Frontline leads and operators. What needs intervention now or this shift? | Analysts and domain owners. Where is an important change concentrated, and what should we examine first? | Department and team managers. Which areas drive the result, and what do we adjust this week? |
| Refresh and detail | Weekly, monthly or at close. Aggregates, targets, trajectories and a small set of exceptions. | By event, every few minutes, or by shift when action permits. Current state, change and enough item detail to route work. | Daily or weekly as sources settle. Comparable segments, distributions and baseline history. | Daily or weekly. Summaries by region, product or team, plus the few items that need a manager’s attention. |
| Interaction | Stable default and a small set of deliberate scope changes. Optional link to commentary or detail. | Immediate reading without clicks. Quick drill to affected records or a response guide. Scope controls only when they help route action. | Visible filters and bounded drill to diagnose a known signal. Use a separate workspace for open-ended exploration. | A few scope filters and a drill to item detail. No cross-filtering on the default view. |
| Alerts | Highlight material misses and sustained changes with an owner and next review. Usually surface them in the review rhythm. | Show active severity, age, affected scope and owner. Notify externally only when timely action and ownership are defined. | Flag qualified deviations, with comparator, minimum sample and timing. Send the analyst to affected segments for validation. | Status marks against targets and a short exception list, worked through in the weekly review. |
| Typical content | Outcomes versus plan, trends, material risks, decisions due. | Queue, capacity, service breaches, exception list, freshness. | Segment ranking, variance, distributions, aligned trend panels, volumes. | Ranked bars, trend lines against last year, deviation bars and short exception tables. |
Team addition Few describes strategic, analytical and operational dashboards. This standard adds tactical for manager views that sit between strategic progress and live operations.
Record one primary type, decision deadline and default audience in the brief. Design the overview around that combination.
Team addition Analytical monitoring here means a fixed starting view for recurring diagnostic questions. Few distinguishes monitoring from free-form analysis. Dashboard Confusion Revisited
Ask a reader to walk through the last time they made the decision. Record the trigger, the evidence they trusted, what action followed, and what they had to look up elsewhere. Ask what would change their mind. Use the answers to separate first-screen signals from supporting detail.
| Area | Ask | Red flag |
|---|---|---|
| Audience | Who reads it, how often, on what screen, and with what training? | “Everyone.” |
| Decisions | What decision does it support, and what action follows a bad reading? | No action can be named. |
| Measures | Which measures, defined how, from which source, and owned by whom? | Two people define it differently. |
| Comparisons | Against what: target, prior period, forecast or norm? | “Just show the number.” |
| Targets | Who sets them? Is higher or lower better? Where do Watch and Act begin? | Nobody owns the target. |
| Time grain | What grain and period: calendar or fiscal, rolling or fixed? | “All history.” |
| Refresh | How fresh must the data be, and how late can it arrive? | “Real time,” with no one able to act in real time. |
| Constraints | Screen size, print or PDF, language, accessibility needs and security. | Unknown. |
| Success | How will we know it works, and when will we review it? | No review date. |
| Answer | Design consequence |
|---|---|
| Decision and action | Becomes the first-screen item list. Items with no decision move to detail (PUR-01, PUR-02). |
| Target and direction | Becomes bullet graphs, bands and status rules (MEA-04, DSP-02). |
| Time grain | Sets the trend axis and the sparkline window (DSP-03, DSP-05). |
| Refresh and delay | Sets the dashboard type, the data-as-of label and the stale state (TYP-01, TXT-03). |
| Screen | Sets the agreed viewport and the grid (LAY-01, LAY-02). |
Get the business owner’s agreement on a one-page brief before building the production view. Record the agreed device and viewport, default scope and test scenarios.
Assign a measure owner and approved source for every headline measure. Record the calculation, inclusion rules, unit, grain, comparator, target direction, period boundaries, refresh and maximum data age.
Agree a measurable acceptance test with typical readers. Time an uncoached read of the default view, then test a threshold case, missing data and the path to action.
Team addition The brief, test evidence and source ownership provide a record reviewers can verify.
Copy the first column as the brief’s headings. Answer each question in one or two lines and link to the approved measure definitions. The right column shows a filled brief.
| Field and interview prompt | Dispatch readiness · completed example |
|---|---|
| Audience and owner. Who acts, where do they view it, and who owns it? | Dispatch leads on a 1,280 × 720 CSS pixel content area. The dispatch manager owns decisions and thresholds. |
| Decision and deadline. What changes and by when? | Reassign work within 10 minutes if dispatches due in the next hour are at risk. |
| Measures and grain. What counts? At what level? | Count distinct open dispatch IDs due in the next 60 minutes that lack a ready vehicle. Show network total and site breakdown. Dispatch events are the approved source. |
| Comparisons and targets. Against what, and in which direction? | Act at 10 or more, watch at 5–9, otherwise within limit. Lower is better. Compare shift completed dispatches with plan at the same point in the shift. |
| Time and freshness. Which clock and how much delay? | Current shift and local site time. Refresh every 2 minutes. Mark stale after 6 minutes without a successful source update. Stop automated escalations while stale. |
| Overview and detail. What is visible, and where does action happen? | At-risk count and threshold, shift progress, site ranking, lead assignment and freshness on the overview. Site detail carries the same time and site scope. |
| Acceptance and approval. What must a reader be able to do? | In 15 seconds, identify whether action is due, the highest-risk site, and when data was current. Demonstrate stale and empty states. Dispatch manager approves the result. |
Choose the measure closest to the decision. Count late dispatches when the job is to prevent delays. Use a proxy only when a direct measure is unavailable, and explain what the proxy misses. Separate the observed value, the comparison and the action threshold.
Label every displayed measure with its unit and time scope. Give decision measures a named comparison or explicitly state that no valid comparator exists and why.
Match the comparator on definition, population, time grain and elapsed period. Label deliberate differences, such as comparing a partial actual with a future target, and show pace or a labeled projection where the decision depends on being on track.
Compute rates from their underlying numerator and denominator at the displayed grain. Apply status and thresholds to full precision. Distinguish actual zero, missing, insufficient, partial and stale values.
When assigning a rating, define the direction of improvement, inclusive and exclusive cutoffs, owner and action. Use words with a distinct mark and give unknown or stale its own state.
Pair summaries with the exception, spread or segment that can change the decision. For a short ranked list, state the sort, count shown and total or link to the full list.
A target is the desired outcome, a forecast is an estimate, a prior period shows observed change, and a norm describes typical behavior under a stated method. Name the comparator directly. For forecasts, show material uncertainty with documented bounds or a range and say what the range means. Compare the same weekdays or elapsed hours where seasonality or shift timing matters.
| Weak | Strong | Why it works |
|---|---|---|
| Health score 71 | At-risk dispatches 12 of 94 due in the next hour · Act at ≥10 | Shows the direct count, population and trigger. |
| On time 93.2867% | On-time dispatch 93.3% (971 of 1,041) · October close · Target 95.0% · 1.7 pp below | Shows volume, useful precision and the gap in percentage points. |
| Revenue $620k vs $1.2M | Quarter to date $620k through 15 Nov · Quarter-end projection $1.04M (scenario range $0.99M–$1.09M) · Quarter-end target $1.20M | Separates actual progress from expected finish. |
| Average wait 8 min · Good | Median wait 8 min · 90th percentile 24 min vs limit 15 min · 42 waits above 15 min of 180 today | Exposes the slow tail an average could hide. |
| Returns 0% | Returns rate unavailable · No delivered orders in scope | Distinguishes an undefined rate from zero returns among delivered orders. |
Start with the comparison. Use a number for reading one value, a table for exact lookup, aligned marks for comparing magnitudes, and time or space only where order or location matters. Reuse the same display for the same task.
Use a labeled scale that preserves the encoded relationship. Every length bar starts at zero, including each signed bar at the visible zero line. A line or dot may use a labeled focused scale if it preserves the intended comparison.
For panels intended to compare magnitudes, align units, time windows, category order and scale domains. If the task compares shapes on independent scales, label each scale and the purpose.
Show missing observations as gaps or explicit states. Distinguish forecasts from actuals with a named boundary and a different line style or mark. Use verified intervals when uncertainty could change the decision.
Choose the simplest display that answers the question. Use aligned position or length before area, and separate series into panels when a combined view obstructs comparison.
| Question | Use | Decision cue |
|---|---|---|
| What is the current value? | Single value or table | Use a table when the reader needs several exact values. |
| How close are we to target? | Bullet graph or bar and line | Use the combination for a comparable target or second measure over time on one scale. |
| Which is larger, smaller or ranked highest? | Bar or dot plot | Use dots when small differences on a clearly labeled focused scale matter. |
| What makes up the whole? | Stacked bar, or restricted treemap | Use a treemap only for many nested, additive parts and rough comparisons. |
| What changed over time? | Line, sparkline, or small multiples | Use sparklines for compact context beside values. Use full axes for estimates or comparisons. |
| How are values distributed? | Histogram or box plot | Show shape with bins or compare medians and spread across groups. |
| Do two numeric measures vary together? | Scatter plot | Use one mark per observation at the stated grain. An association alone cannot establish cause. |
| How far above or below a reference? | Signed deviation bar | Subtract the named reference and center the reading on zero. |
| Does location change the action? | Point map | Use for proximity, barriers or spatial patterns. Rank named places with bars when location is incidental. |
Team addition Chart-by-chart defaults below turn the visual principles into reviewable tests. The geographic and treemap rules set narrow conditions for using area and position.
Use when. The first task is to read one current number or status.
Avoid when. Readers need to compare a set of values or judge a trend.
Common mistake. A large number with no threshold or period.
Pair the value with a name, unit, period and decision context.
Network · 7 Nov 2026, 09:42 CST
At-risk dispatches · Next 60 min
12Act · Trigger ≥10 · 94 due
Use when. One featured value needs comparison with a target or benchmark.
Avoid when. Readers primarily need trend shape, or the comparison mixes units or populations.
Common mistake. Using a cropped baseline for a length bar or adding arbitrary performance bands.
Use a common linear scale, a prominent featured mark and a short perpendicular comparison marker. A length bar starts at zero. Use a point for a focused scale. Include bands only for approved cutoffs and state their direction in text.
North site · 7 Nov shift through 10:00 CST · Dispatches (count)
Use when. A nearby value needs a compact history trace.
Avoid when. The reader needs to estimate intermediate values or compare different scales precisely.
Common mistake. Auto-scaling comparable rows independently or leaving the ending date unclear.
Place the sparkline beside its current value. State the time window and grain, mark the latest point and share scales where magnitudes are compared.
East site · 6 weekly closes, 27 Sep–1 Nov 2026
Late dispatches (count)
927 Sep through 1 Nov, oldest to newest · 4, 6, 5, 7, 8, 9.
Use when. Rank or compare magnitudes across categories.
Avoid when. Shape through ordered time or precise comparison on a focused scale is the primary task.
Common mistake. Truncated bars or alphabetical sorting that obscures rank.
Start the value axis at zero. Keep lengths comparable and sort by value unless a meaningful category order should take precedence.
Network · 7 Nov 2026, 09:42 CST · Due in next 60 min
Use when. Read movement or compare trajectories over ordered time.
Avoid when. The x-axis is unordered categories or only exact values matter.
Common mistake. Connecting across missing periods or evenly spacing irregular dates.
Use equally spaced positions for regular reporting buckets and elapsed-time positions for irregular observations. Show the first and last periods, units and a labeled scale. Leave gaps where data is missing.
Metro network · Monthly closes, May–Oct 2026
May 83%, Jun 87%, Jul 86%, Aug 90%, Sep 88%, Oct 93%.
Use when. Compare a discrete actual to a commensurate benchmark over time.
Avoid when. The measures need different units or independent scales.
Common mistake. Using two y-axes or allowing the benchmark to look like a second actual.
Use one labeled scale and compatible units and time grain. Start bars at zero. Distinguish the benchmark line by style and label.
North site · October 2026 reporting weeks
Week 1 actual 32, week 2 actual 43, week 3 actual 37, week 4 actual 48. Plan was 40 every week.
Use when. Show the whole and a few additive parts.
Avoid when. Readers must compare interior segments precisely or the parts overlap.
Common mistake. Stacking rates with different denominators or omitting the total behind a 100% view.
Stack mutually exclusive parts on a common zero origin. Keep segment order stable. State the denominator for 100% stacks and directly identify the parts.
October 2026 shipments · Total 100
Use when. Compare closely spaced values on a common numeric scale.
Avoid when. Absolute magnitude from zero is the central comparison.
Common mistake. Changing domains across panels or making a tight range look like an absolute scale.
Use one row per category and a labeled numeric axis. Show exact values when a focused domain is needed.
On-time dispatch · October 2026 close
Use when. Read the shape and tail of one numeric distribution.
Avoid when. Readers need individual values or compact comparison across several groups.
Common mistake. Treating numeric bins like sortable categories or using unequal bins with raw count heights.
Use contiguous equal-width numeric bins and a zero count baseline. State the bin width and observation count. State how boundary values are assigned.
East · 44 dispatch waits · 7 Nov 2026
From 0 minutes in five-minute steps · 3, 10, 16, 8, 5, 2. Lower bound included, upper bound excluded.
Use when. Compare median and spread across groups on the same scale.
Avoid when. Exact shape, multimodality or a very small sample drives the decision.
Common mistake. Calling every whisker a minimum or maximum or hiding sample size.
State the box, whisker and outlier definitions. Give each group’s sample size and use a shared scale.
6 Nov 2026 · East n = 44, West n = 39
East whiskers 3 and 24, quartiles 7 and 16, median 11, outlier 32 min. West whiskers 2 and 16, quartiles 5 and 10, median 7, outlier 19 min.
Use when. Assess a relationship between two numeric variables at one observation grain.
Avoid when. One variable is a nominal category, or the task is to show ordered change.
Common mistake. Joining independent observations or treating correlation as causal.
Name both axes with units and name the observation grain. Show overlap with an explicit method when it obscures the pattern.
Network · 12 October 2026 shifts
Pairs in dispatches per staffed hour and median wait minutes · (3, 8), (5, 7), (7, 10), (8, 9), (10, 12), (12, 13), (13, 11), (15, 16), (16, 14), (19, 18), (20, 15), (22, 20).
Use when. Compare the same pattern or magnitude across categories without overlapping series.
Avoid when. The groups require incompatible units or the panels become unreadably small.
Common mistake. Using independent axes for a magnitude comparison.
Repeat the same chart structure, encoding, date range and ordering across panels. Apply VIS-02 to every scale.
Shipments by region · March–October 2026 · thousands per month
Monthly shipments, March to October, in thousands. North 40, 42, 43, 46, 44, 46, 47, 48. East 34, 35, 37, 39, 37, 39, 40, 41. West 27, 29, 30, 32, 31, 33, 33, 34. South 17, 20, 21, 25, 20, 22, 24, 24.
Use when. Look up exact values, names, dates or records.
Avoid when. Spotting a broad pattern or ranking is the first job and a chart communicates it faster.
Common mistake. Using a dense grid or mixing units in one column.
Use explicit headers and units, a known sort order and right-aligned numbers. Add an adjacent visual only when it supports a distinct task.
Network · Dispatch counts due in next 60 min · 7 Nov 2026, 09:42 CST
| Site | Due | At risk | Lead |
|---|---|---|---|
| East | 45 | 7 | Rivera |
| North | 29 | 3 | Singh |
| West | 20 | 2 | Kwan |
Use when. Screen many nested, positive, additive parts for dominant groups.
Avoid when. Close comparisons, negative values or temporal order matter. Prefer bars for shallow hierarchies when space allows.
Common mistake. Using both saturated color and area for the same quantity or placing a zero value in a visible rectangle.
Size each rectangle by its share of the parent. Label hierarchy, units and values. Use boundaries and provide a readable ranked table or list.
Labor by area and task · October 2026 · Total 300 h
Areas in rank order · Warehouse 120 h, Yard 100 h, Office 80 h. Tasks in rank order (hours) · Warehouse Sort 40, Yard Load 31, Warehouse Scan 30, Yard Stage 27, Warehouse Pack 26, Warehouse Check 24, Office Route 24, Yard Move 22, Office Plan 21, Yard Check 20, Office Call 19, Office File 16.
Use when. See which categories are over or under a known reference and by how much.
Avoid when. The absolute level matters more than the difference, or groups use incompatible references.
Common mistake. Displaying a signed difference without naming the comparator.
Use one visible zero reference. Show sign and units directly and name the baseline.
October 2026 · Completed dispatches minus each site’s October plan
Use when. Distance, location or a barrier changes the response.
Avoid when. The place name is only a grouping label.
Common mistake. Assigning random positions or hiding values in pointer-only marks.
Use verified spatial coordinates, show orientation or scale where relevant, and state the decision supported by location. Include a text equivalent.
Alder Reach, fictional local kilometre grid · 7 Nov 2026, 09:42 CST
North bridge closed. South bridge open. West depot is west of the river near the closed bridge. The call and East crew are on the east side, with road access on that side.
Few’s design uses a straight scale to fit actual, comparison and optional ratings in a compact space. Bullet Graph Design Specification
Network · Completed dispatches · 7 Nov 2026 shift through 10:00 CST
76 dispatches completed, plan 84. Scale runs from 0 to 120 in steps of 30. See the caption for the three bands.
East · 7 Nov 2026 current shift
Operating result · October 2026 close · USD thousands
Revenue · Q4 2026 through 15 Nov · USD thousands
Late dispatches · Weekly closes from 27 Sep to 1 Nov 2026
| Site | Count | 6-week trend | Status |
|---|---|---|---|
| East | 9 | Act | |
| North | 5 | Watch | |
| West | 2 | Within limit |
East · 4, 6, 5, 7, 8, 9 (Act). North · 4, 3, 5, 5, 4, 5 (Watch). West · 3, 2, 2, 4, 3, 2 (Within limit).
Keep monitoring charts two-dimensional. Exclude pies, donuts, radial gauges, radar charts and dual-scale charts from the monitoring view. Use the replacements below.
Use a map only when a spatial relationship changes the decision. Plot trustworthy positions on appropriate geometry and make values available as text.
| Avoid or restrict | What it makes harder | Use instead |
|---|---|---|
| Pie and donut | Close shares are hard to compare by angle or arc. Separate pies lack a shared baseline. | Sorted bars for comparing shares. A 100% stacked bar for a few parts of a stated whole. |
| Radial gauge | A large footprint shows a single value, and angled needles make side-by-side comparison slow. | A bullet graph or value with a named target and gap. |
| 3-D chart | Perspective shifts apparent length or area and can hide marks. | Flat bars or dots on a common axis. |
| Radar chart | Shape and area change when category order changes. Adjacent spokes imply a relationship that may be arbitrary. | Aligned dot plots or small multiples with explicit units. Compare only like scales. |
| Dual-axis chart | Independent scale ranges can manufacture crossings and apparent correlation. | Aligned panels with the same time axis, a scatter plot for association, or one clearly labeled index scale when relative movement is the question. |
| Decorative map | Area and geographic position attract attention when the job is simply to compare places. | A sorted bar chart or table. Keep maps for location-driven decisions. |
Team addition The exclusion in AVD-01 is a team release policy. The perceptual rationale follows Few’s writing on dashboard design and dual scales. Reading notes
Design for the viewport agreed in the brief, after browser and product chrome. Everything essential fits that one screen at default zoom, with no scrolling or paging (LAY-01). Build every layout on a shared grid. The starting grid below is a team addition.
Readers of left-to-right languages start at the top left. Put the most important item there, give it the most space, and let importance fall off toward the bottom right (LAY-03). Number the regions in the wireframe, and keep the reading order in the page markup the same as the viewing order.
Group related items by proximity and alignment first. Use whitespace to separate groups. Add a light surface fill only when proximity is not enough, and never box every chart (LAY-02).
Put shared controls in one quiet band beside the heading, or in a left rail. Show their current state as text, with a reset control (LAY-04, INT-01).
Detail lives on drill-through pages reached from a specific item. Each detail page names its scope, keeps the filter context and returns to the same overview state (LAY-05, INT-02).
At the agreed desktop viewport or primary monitoring display, show all essential signals, exceptions and scope in one view at default zoom without scrolling or switching tabs. At narrow widths or zoom, follow ACC-03.
Align components to a shared grid. Use a smaller gap within a group than between groups. Use whitespace and headings first, adding borders only where alignment or grouping needs them.
Place the highest-priority signal or exception before its supporting detail in visual and reading order. Keep a headline measure, its comparison and its status in the same region.
Put shared scope controls next to the heading and local controls inside their group. Preserve labels and reading order when the layout reflows.
Keep detail pages out of the critical overview path. Name the destination, preserve applicable scope on entry, and provide a predictable return to the same overview state.
Show no more than nine data items on the primary view. Count a small-multiples set or a sparkline table as one item.
Team addition A useful desktop starting grid is 12 equal columns, 16 CSS pixels between regions, and 24 around the edge. Begin with 8–12 pixels within a group. Agree the usable viewport after browser and product chrome. Keep horizontal alignment between related charts. Review the screen in reading order, typically left to right and top to bottom for English. Avoid forcing equal-sized tiles when the decision priorities differ.
Start with the decision and scope, then the signals and exceptions, then the explanation or action path. When the overview does not fit, remove duplicate signals, shorten the exception list with a clear route to the rest, and move background material to detail. Keep legible type. The worked examples and the full-screen wireframes below show how the order changes by type.
Each wireframe is drawn at half scale of an example 1920 × 1080 viewport. Circled numbers give the viewing order. Values are sample data.
Team addition This is the team’s default light palette and use contract. Few’s underlying principle is to let context recede and reserve visual emphasis for information that needs attention.
Use muted marks by default and one focal highlight hue per view. Give each color one meaning within a dashboard, including across filters, linked detail and refreshes. Keep status colors reserved for statuses.
Pair status with readable words and a distinct shape. Pair categorical or signed color with labels, line style, marker shape or aligned position. Check grayscale and common color-vision simulations.
Meet 4.5:1 for normal text and 3:1 for large text, essential graphical boundaries and required control states against adjacent colors. For this test, large text is at least 18 pt regular or 14 pt bold, approximately 24 or 18.7 CSS px respectively. Test actual pairs after blending. Use the permitted palette pairings below.
Use a sequential scale for increasing magnitude and a diverging scale only around a meaningful midpoint. Label units and class boundaries, including the midpoint and missing state. Give the same value the same color across comparable views.
| Role | Swatch and hex | Approved use and contrast |
|---|---|---|
| Canvas | #FFFFFF | Default reading and plot surface. |
| Subtle surface | #F2F5F7 | Quiet group or table header fill. |
| Primary text | #17242C | Titles, key values, essential dark outlines. Text on canvas 15.85:1, on subtle 14.47:1. |
| Secondary text | #344955 | Labels, supporting prose and notes. Text on canvas 9.42:1, on subtle 8.60:1. |
| Muted text | #526370 | Metadata and Unknown status. Text on canvas 6.22:1, on subtle 5.68:1. |
| Gridlines | #D7E0E5 | Decorative guides and row dividers only. |
| Axis and boundary | #647581 | Essential axes or control outlines on canvas and subtle surfaces. 4.77:1 on canvas, 4.36:1 on subtle. Use the text tiers for tick labels. |
| Series 1 | #4F6978 | Default bars, lines and points. White text on fill 5.79:1. Mark on canvas 5.79:1, on subtle 5.29:1. |
| Series 2 | #705F7B | Second category when it has a labeled role. White text on fill 5.81:1. Mark on canvas 5.81:1, on subtle 5.31:1. |
| Series 3 | #726541 | Third category when it has a labeled role. White text on fill 5.75:1. Mark on canvas 5.75:1, on subtle 5.25:1. |
| Highlight | #005DA6 | Focus, links and one focal chart role. Text on canvas 6.74:1, on subtle 6.15:1. |
| Within limit | #21604E | Status only. Use a circle and words. Text on canvas 7.37:1, on subtle 6.73:1. |
| Watch | #765210 | Status only. Use a diamond and words. Text on canvas 7.04:1, on subtle 6.43:1. |
| Act | #963528 | Status only. Use a triangle and words. Text on canvas 7.40:1, on subtle 6.76:1. |
| Bullet · Act | #C6D0D8 | Darkest neutral background band. Ink featured bar has 10.12:1 contrast. Ink text on fill 10.12:1. |
| Bullet · Watch | #DFE6EB | Middle neutral background band. Ink featured bar has 12.57:1 contrast. Ink text on fill 12.57:1. |
| Bullet · Within limit | #F0F3F5 | Lightest neutral band. Ink featured bar has 14.22:1 contrast. Ink text on fill 14.22:1. |
| Sequential · 1 low | #D8E7F2 | Lightest magnitude class. Ink text on fill 12.55:1. |
| Sequential · 2 | #91B6D1 | Second magnitude class. Ink text on fill 7.41:1. |
| Sequential · 3 | #426C91 | Third magnitude class. White text on fill 5.55:1. |
| Sequential · 4 high | #245578 | Darkest magnitude class. White text on fill 7.94:1. |
| Diverging · far below | #913E2D | Strong negative deviation. White text on fill 7.17:1. |
| Diverging · below | #DEAA99 | Moderate negative deviation. Ink text on fill 7.78:1. |
| Diverging · midpoint | #F0F3F5 | Neutral midpoint. Label its numeric meaning. Ink text on fill 14.22:1. |
| Diverging · above | #9ABEBB | Moderate positive deviation. Ink text on fill 7.89:1. |
| Diverging · far above | #21605E | Strong positive deviation. White text on fill 7.24:1. |
For scale fills and qualitative bands, color is redundant. Give essential shapes a dark boundary with at least 3:1 contrast against adjacent fills, plus labels or values. Keep quantitative labels outside fills on white or use the specified text pair. Separate dark categorical marks with white space or a white boundary when their adjacency is needed to distinguish them. Gridlines and subtle fills are decorative. Use the axis color for any line needed to interpret the data.
Check the dashboard against each forbidden use below. Remove any occurrence from production views.
Palette applied
Single ink
Use one readable sans-serif family. At the design viewport, start at 24–32 CSS px for the page title, 24–36 for primary values, 16–20 for section titles, 14–16 for data and controls, and 12–14 for secondary notes and ticks. Resolve crowding by removing redundant content.
Right-align numeric table cells within a column and use tabular digits where supported. Keep the same units, precision and sign convention within a comparable series or column.
Display units and time labels without ambiguity. Use the dashboard’s declared locale and currency. Apply display rounding only after calculation and preserve exact source precision in authorized detail when the decision requires it.
| Item | Team default |
|---|---|
| Type and alignment | Use a system sans-serif stack with regular text and semibold labels. Left-align words. Align numbers on the right and use the same decimal precision when comparing a column. |
| Abbreviation and units | Use USD 1.24M for a headline only when its precision is enough. In tables, choose a unit for the whole column, such as Revenue (USD thousands). Show exact values in detail where needed. |
| Negative and positive values | Use a minus sign, as in −42. Use +18 for a positive deviation when it aids reading. Use parentheses only where the audience’s established accounting convention requires them. |
| Percentages and rates | Use 93.3% for a rate. A move from 91.0% to 93.3% is +2.3 pp or about +2.5% relative. Spell out percentage points when the abbreviation is unfamiliar. |
| Rounding | A measured rate of 94.96% can round to 95.0% while missing a 95.0% target. Show 94.96% or 0.04 pp below target if that difference changes action. |
| Dates and times | Use 7 Nov 2026 or Nov 2026 in English labels. For intra-day decisions include the local time and zone or offset, as in 7 Nov 2026, 09:42 CST (UTC−06:00). Name fiscal periods explicitly. |
Team addition The type sizes, locale defaults and sign conventions are implementation defaults. Check them at the actual reading distance. Use larger type for wall displays.
Title each chart with its monitoring question or a supported current finding. If the title states a finding, update it when filters, refresh or missing data invalidate it.
Name every axis, series and marker that would otherwise be ambiguous. Prefer direct labels when legible. If needed, place a compact legend beside the plot with matching order and non-color cues.
Show current scope, latest successful data-as-of time and stale or partial state where the reader makes the decision. Put full definitions, sources and cutoff notes in a labeled footnote or linked definition.
Use the business’s own terms in plain language. Define each acronym at first use on the dashboard or in its linked definitions.
Write for the job. “Dispatches at risk by site” works for a stable monitor. “East has 7 of 12 at-risk dispatches” works if the title follows the data. A short annotation can name a verified event or explain a discontinuity. Anchor it to the affected observation and distinguish observed timing from a proven explanation.
Keep chart labels horizontal where possible. Thin tick labels before shrinking them. State units once per axis or column where that makes every value clear. Keep critical thresholds and warnings next to the affected value, even if the detailed method lives in a note.
| Title type | Example | Use |
|---|---|---|
| Topic only | “Revenue” | Avoid. The reader has to read the chart to learn anything (TXT-01). |
| Monitoring question | “Is revenue on pace for the October target?” | Stable monitors whose answer changes with every refresh. |
| Supported finding | “October revenue 3% below target; margin above plan” | Reviewed or dynamic views. Compute the finding from the data and update it when filters or refresh change the answer. |
A reader who skims only the titles should come away with the story. A title that says revenue is up while the chart shows it down is a defect, not a quirk.
Label the mark, not a key (TXT-02). A line gets its name at its right end, a bar gets its value at its tip, and a dot gets both. Use a legend only when direct labels would collide, and place it beside the plot in the same order as the marks. The reader should never travel from mark to key and back more than once.
Annotate only the one or two points that change how a chart reads, such as a policy change, an outage or a data gap. Keep each note to a few words in the secondary text color, anchor it to the affected observation with a short leader line, and never cover a mark. Say what happened, not what the number is. The number already has a label.
Every dashboard carries a definition note for any measure whose name does not fully define it, and a freshness note with the data-as-of time and refresh behavior (TXT-03).
Sample data On-time delivery = orders delivered by the promised date, net of cancellations. Data as of 7 Nov 2026, 09:41 CST · refreshes every 2 minutes · stale after 6 minutes.
Make the default view answer the primary monitoring question without interaction. Show active filters, affected scope and a reset path when filters are available.
Make drill paths deliberate. Pass the selected entity and relevant scope, show the context at the destination, and keep required information available without hovering.
Reserve tooltips for extra precision or supporting context. Where present, make the same content reachable by keyboard and touch or via an equivalent visible route.
For every automated alert, define trigger, duration or repeat rule, recipient, action, resolution and stale-data behavior. Test that refresh does not cause repeated notifications for the same unresolved event.
| Type | Enable when useful | Keep stable | Do not make interactive |
|---|---|---|---|
| Strategic | Period or portfolio scope where the owner compares recurring views. Link to the approved commentary or measure detail. | Keep the default period and benchmark visible. Avoid selections that silently redefine the scorecard. | Sorting, chart-type switching or swapping measures on the default view. |
| Operational | One deliberate step to affected work, a handoff or a response guide. Optional pause for rapid updates when safe, with a visible paused state. | Keep urgency, age and affected scope visible. Preserve focus and avoid unexpected resorting while someone is reading or acting. | Anything that needs training, or that hides urgency while someone is acting. |
| Analytical monitoring | Explicit filter and reset, meaningful drill or coordinated selection, and optional exact-value details. | Preserve the baseline and show what each control affects. If a selection changes the denominator or target, label the new scope. | Layout rearranging or redefining measures. Send open questions to an exploration workspace. |
| Tactical | Up to three scope filters, such as region, product or team, and a drill to item detail. | Keep the week, the target and the exception list visible while filters change. | Chart-type switching or cross-filtering on the default view. |
Team addition Alert delivery, deduplication and stale-data behavior are operational safeguards. A persistent, explicitly labeled on-screen exception is different from an automated notification.
Team addition Apply WCAG 2.2 AA where the delivery surface supports web content, and verify equivalent access on other surfaces. The checks here cover common dashboard barriers. See the accessibility reference for the complete criteria.
Operate every control with a keyboard. Use a logical focus order, visible and unobscured focus, and a way to leave each control. Preserve focus during refresh and return it predictably after dismissing a panel.
Give each chart a meaningful accessible name and provide an equivalent readable summary or data table. Expose control names, values, states and table headers. Announce meaningful changes without reading every refresh.
Keep text and controls usable at 200% text resize. At 320 CSS px width or equivalent 400% zoom, reflow content in reading order without page-level horizontal scrolling. Where a chart or data table genuinely requires two dimensions, contain its scrolling and provide an equivalent readable alternative.
Verify contrast and non-color cues, complete the core tasks by keyboard, and read the overview, changed filters and detail using a supported screen reader. If the chart surface cannot provide equivalent access, include an accessible route before release.
Make hover or focus disclosures dismissible, hoverable and persistent where WCAG 1.4.13 applies. Provide a visible or equivalent touch path for the same information.
For color vision, test protan, deutan and tritan simulations as well as grayscale. Confirm that labels, shapes and patterns still separate the intended series. At narrow widths prioritize the same critical information and permit vertical reading. A responsive stack can preserve access without changing the purpose of the desktop overview.
Test with real keyboard focus and a screen reader, including status and filter changes. A chart’s alternative should preserve the comparison, values needed for the decision, exceptions and units. Use a data table when exact lookup is part of the job. For operational auto-updates, announce changes that require attention and keep repeated background updates quiet.
Team addition Keep the approved brief, rule findings, test evidence, exception decisions and release version together. Record dashboard and standard versions separately.
Before release, have the dashboard owner validate the decisions and measures, and a BI peer verify the checklist against the actual delivery surface. Close all MUST findings and record evidence for every pass or inapplicable rule.
Record each SHOULD departure with its rule ID, affected audience, evidence, risk, mitigation, approver and expiry or review date. MUST failures block release. Fix the failure or use an approved delivery surface that passes.
Re-review when audience, scope, measure definitions, targets, refresh, major layout or interactions change. Record the dashboard version and the standard version reviewed.
Use Pass, Fail, or Not applicable with evidence for each cited rule. These rows group related checks. Open individual rule links when recording findings.
| Review area | What to verify | Rule IDs |
|---|---|---|
| Purpose and brief | Name the recurring decision, choose the type, agree on the viewport and test it with a typical reader. | |
| Measures and states | Reconcile definitions, comparator, grain and full-precision triggers. Exercise zero, missing, partial and stale cases. | |
| Visual selection | Match the display to the task. Verify axes, aligned scales, label and display rules, forecast boundaries and forbidden forms. | |
| First screen | Inspect the default overview and reading order at the agreed size. Check scope, related groups and return from detail. | |
| Color and type | Check actual contrast pairs, status in grayscale, color consistency, precision, locale and size. | |
| Language and controls | Check dynamic titles, scope and freshness. Exercise reset, drill, tooltip alternatives and alert lifecycle. | |
| Access | Test keyboard, focus, assistive reading, resizing, reflow and alternate data access. | |
| Release evidence | Record owner and peer approval, evidence, departures, expiry and versions. |
The BI design owner maintains this standard. Use major versions for breaking requirements, minor versions for new compatible guidance or rules, and patch versions for clarifications. Keep IDs stable. Retire an ID with a redirect and reason, and never reuse it. In the change log record date, version, affected IDs, decision and migration impact.
| Version | Date | Change and impact |
|---|---|---|
| 1.0.0 | 7 Nov 2026 | Initial standard. Apply to new dashboard releases and review existing dashboards when they undergo material changes. |
| Symptom | Why it hurts | Rule ID | Fix |
|---|---|---|---|
| There are three overview tabs. | The reader must remember earlier signals to compare them. | LAY-01 | Put the recurring decision and its essential signals on one default screen. |
| The title says “Performance.” | Readers cannot tell what they are monitoring. | TXT-01 | Name the measure or state a supported finding. |
| The top tile shows a naked count. | The count cannot distinguish normal from urgent. | MEA-01, DSP-01 | Show scope, comparator and trigger with the count. |
| A partial month is compared with a full prior month. | Elapsed time can explain the apparent change. | MEA-02 | Match elapsed periods or show a clearly labeled pace comparison. |
| A site rate is averaged across sites. | Small sites have the same weight as large sites. | MEA-03 | Sum numerators and denominators before dividing. |
| Unknown data appears as zero. | Readers may take action on a state that has never been observed. | MEA-03, VIS-03 | Show an unavailable state or a gap with an explanation. |
| An exception has a red dot only. | Its meaning depends on seeing and decoding the hue. | COL-02, MEA-04 | Use a triangle and the word Act with the trigger. |
| Bars start at 90%. | Lengths exaggerate small differences. | VIS-01, DSP-04 | Start bars at zero or use labeled dots for the focused comparison. |
| Comparable trend panels use different scales. | Equal-looking slopes or heights can mean different magnitudes. | VIS-02, DSP-12 | Use common domains, units and periods for comparison. |
| The data is hours old and still labeled “Live.” | An operator may respond to a resolved issue. | TXT-03, INT-04 | Show the last successful data time, stale state and alert policy. |
| Every KPI has a bright border and icon. | The urgent exception gets the same weight as everything else. | LAY-03, COL-01 | Use a quiet baseline and reserve emphasis for the action signal. |
| The target is available only on hover. | Readers cannot make the primary comparison from the overview. | INT-01, INT-02, ACC-05 | Place target and gap beside actual. Reserve tooltip for supporting context. |
| A dual-scale chart makes two lines cross. | Axis choices can create a visually persuasive relationship. | AVD-01, DSP-06 | Use time-aligned panels or a scatter plot at a shared grain. |
| Clicking a site opens unfiltered detail. | Readers may investigate the wrong population. | LAY-05, INT-02 | Pass the selected site and period. Display them in detail and preserve return state. |
| Keyboard focus disappears after refresh. | The reader loses their place during a decision. | ACC-01 | Preserve focus and verify focus visibility in the actual delivery surface. |
| A zero-to-one-hundred stack shows averages. | The colored lengths imply a whole that cannot be added. | DSP-07 | Show the averages as aligned positions. Stack only additive, non-overlapping parts. |
| A pie chart shows channel share. | Readers compare angles poorly, especially between slices that do not touch. | AVD-01 | Use sorted bars or a dot plot for the parts. |
| Bars are drawn in 3-D. | Perspective and depth distort length and hide values behind other bars. | AVD-01 | Draw flat bars on a common baseline with direct values. |
| One measure uses a rainbow color scale. | Hue steps imply categories that do not exist and fail color-vision checks. | COL-04, COL-05 | Use one sequential hue with labeled class boundaries. |
| A legend sits below a multi-line chart. | Each reading needs a trip from mark to key and back. | TXT-02 | Label each line at its end. Keep a legend only when labels would collide. |
| Region bars appear in alphabetical order. | The reader has to do the ranking the chart should show. | DSP-04 | Sort by value unless a meaningful category order applies. |
| Revenue shows as $48,214,906.37. | False precision slows reading and implies accuracy the decision does not need. | TYPF-02, TYPF-03 | Round for display after calculation, such as $48.2M. Keep exact values in authorized detail. |
| A map decorates the screen when location does not matter. | Region size and shape draw attention the values do not justify. | AVD-02 | Use a ranked bar or table. Use a map only when a spatial relationship changes the decision. |
| The overview has 20 equal-sized tiles. | With no hierarchy, the urgent signal has nowhere to stand out. | PUR-01, LAY-03, LAY-06 | Rank the measures in the brief. Keep the top signals and move the rest to detail. |
| Blue means North in one panel and Forecast in another. | Readers carry a color meaning between panels and misread the second. | COL-01 | Fix one meaning per color across the dashboard and check it in review. |
The executive team reviews company performance each month and decides where the plan needs action. Both versions use the same invented October 2026 data.
Before · a typical first draft
After · same data, redesigned to this standard
Revenue $4.83M against a $5.00M target (Watch); October 2025 was $4.52M. Gross margin 41.2% against 40.0% (Within limit). On-time delivery 92.4% against 95.0% (Watch). Return rate 4.1% against a 3.5% ceiling (Act). Revenue by region in October: North $1.62M, East $1.31M, West $1.04M, South $0.86M. Open tickets 1,284 against a ceiling of 1,100 (Act).
Every rule in this standard, in the order it appears. Select an ID to jump to the rule in context. 64 rules: 38 MUST, 24 SHOULD, 2 GUIDANCE.
| ID | Level | Rule |
|---|---|---|
| PUR-01 | MUST | Name the audience, recurring decision and action. Keep a first-screen item only when the brief identifies the decision it supports. |
| PUR-02 | SHOULD | Move exhaustive records, open-ended exploration and multi-step workflows to the appropriate companion surface. Link from the relevant summary. |
| PUR-03 | GUIDANCE | Start a requirements meeting with a recent decision. Ask what evidence made someone act, and sketch only the information they used. |
| TYP-01 | MUST | Record one primary type, decision deadline and default audience in the brief. Design the overview around that combination. |
| REQ-01 | MUST | Get the business owner’s agreement on a one-page brief before building the production view. Record the agreed device and viewport, default scope and test scenarios. |
| REQ-02 | MUST | Assign a measure owner and approved source for every headline measure. Record the calculation, inclusion rules, unit, grain, comparator, target direction, period boundaries, refresh and maximum data age. |
| REQ-03 | SHOULD | Agree a measurable acceptance test with typical readers. Time an uncoached read of the default view, then test a threshold case, missing data and the path to action. |
| MEA-01 | MUST | Label every displayed measure with its unit and time scope. Give decision measures a named comparison or explicitly state that no valid comparator exists and why. |
| MEA-02 | MUST | Match the comparator on definition, population, time grain and elapsed period. Label deliberate differences, such as comparing a partial actual with a future target, and show pace or a labeled projection where the decision depends on being on track. |
| MEA-03 | MUST | Compute rates from their underlying numerator and denominator at the displayed grain. Apply status and thresholds to full precision. Distinguish actual zero, missing, insufficient, partial and stale values. |
| MEA-04 | MUST | When assigning a rating, define the direction of improvement, inclusive and exclusive cutoffs, owner and action. Use words with a distinct mark and give unknown or stale its own state. |
| MEA-05 | SHOULD | Pair summaries with the exception, spread or segment that can change the decision. For a short ranked list, state the sort, count shown and total or link to the full list. |
| VIS-01 | MUST | Use a labeled scale that preserves the encoded relationship. Every length bar starts at zero, including each signed bar at the visible zero line. A line or dot may use a labeled focused scale if it preserves the intended comparison. |
| VIS-02 | MUST | For panels intended to compare magnitudes, align units, time windows, category order and scale domains. If the task compares shapes on independent scales, label each scale and the purpose. |
| VIS-03 | MUST | Show missing observations as gaps or explicit states. Distinguish forecasts from actuals with a named boundary and a different line style or mark. Use verified intervals when uncertainty could change the decision. |
| VIS-04 | GUIDANCE | Choose the simplest display that answers the question. Use aligned position or length before area, and separate series into panels when a combined view obstructs comparison. |
| DSP-01 | SHOULD | Pair the value with a name, unit, period and decision context. |
| DSP-02 | MUST | Use a common linear scale, a prominent featured mark and a short perpendicular comparison marker. A length bar starts at zero. Use a point for a focused scale. Include bands only for approved cutoffs and state their direction in text. |
| DSP-03 | SHOULD | Place the sparkline beside its current value. State the time window and grain, mark the latest point and share scales where magnitudes are compared. |
| DSP-04 | MUST | Start the value axis at zero. Keep lengths comparable and sort by value unless a meaningful category order should take precedence. |
| DSP-05 | MUST | Use equally spaced positions for regular reporting buckets and elapsed-time positions for irregular observations. Show the first and last periods, units and a labeled scale. Leave gaps where data is missing. |
| DSP-06 | MUST | Use one labeled scale and compatible units and time grain. Start bars at zero. Distinguish the benchmark line by style and label. |
| DSP-07 | MUST | Stack mutually exclusive parts on a common zero origin. Keep segment order stable. State the denominator for 100% stacks and directly identify the parts. |
| DSP-08 | SHOULD | Use one row per category and a labeled numeric axis. Show exact values when a focused domain is needed. |
| DSP-09 | MUST | Use contiguous equal-width numeric bins and a zero count baseline. State the bin width and observation count. State how boundary values are assigned. |
| DSP-10 | SHOULD | State the box, whisker and outlier definitions. Give each group’s sample size and use a shared scale. |
| DSP-11 | SHOULD | Name both axes with units and name the observation grain. Show overlap with an explicit method when it obscures the pattern. |
| DSP-12 | SHOULD | Repeat the same chart structure, encoding, date range and ordering across panels. Apply VIS-02 to every scale. |
| DSP-13 | SHOULD | Use explicit headers and units, a known sort order and right-aligned numbers. Add an adjacent visual only when it supports a distinct task. |
| DSP-14 | SHOULD | Size each rectangle by its share of the parent. Label hierarchy, units and values. Use boundaries and provide a readable ranked table or list. |
| DSP-15 | SHOULD | Use one visible zero reference. Show sign and units directly and name the baseline. |
| DSP-16 | SHOULD | Use verified spatial coordinates, show orientation or scale where relevant, and state the decision supported by location. Include a text equivalent. |
| AVD-01 | MUST | Keep monitoring charts two-dimensional. Exclude pies, donuts, radial gauges, radar charts and dual-scale charts from the monitoring view. Use the replacements below. |
| AVD-02 | MUST | Use a map only when a spatial relationship changes the decision. Plot trustworthy positions on appropriate geometry and make values available as text. |
| LAY-01 | MUST | At the agreed desktop viewport or primary monitoring display, show all essential signals, exceptions and scope in one view at default zoom without scrolling or switching tabs. At narrow widths or zoom, follow ACC-03. |
| LAY-02 | SHOULD | Align components to a shared grid. Use a smaller gap within a group than between groups. Use whitespace and headings first, adding borders only where alignment or grouping needs them. |
| LAY-03 | MUST | Place the highest-priority signal or exception before its supporting detail in visual and reading order. Keep a headline measure, its comparison and its status in the same region. |
| LAY-04 | SHOULD | Put shared scope controls next to the heading and local controls inside their group. Preserve labels and reading order when the layout reflows. |
| LAY-05 | SHOULD | Keep detail pages out of the critical overview path. Name the destination, preserve applicable scope on entry, and provide a predictable return to the same overview state. |
| LAY-06 | SHOULD | Show no more than nine data items on the primary view. Count a small-multiples set or a sparkline table as one item. |
| COL-01 | MUST | Use muted marks by default and one focal highlight hue per view. Give each color one meaning within a dashboard, including across filters, linked detail and refreshes. Keep status colors reserved for statuses. |
| COL-02 | MUST | Pair status with readable words and a distinct shape. Pair categorical or signed color with labels, line style, marker shape or aligned position. Check grayscale and common color-vision simulations. |
| COL-03 | MUST | Meet 4.5:1 for normal text and 3:1 for large text, essential graphical boundaries and required control states against adjacent colors. For this test, large text is at least 18 pt regular or 14 pt bold, approximately 24 or 18.7 CSS px respectively. Test actual pairs after blending. Use the permitted palette pairings below. |
| COL-04 | SHOULD | Use a sequential scale for increasing magnitude and a diverging scale only around a meaningful midpoint. Label units and class boundaries, including the midpoint and missing state. Give the same value the same color across comparable views. |
| COL-05 | MUST | Check the dashboard against each forbidden use below. Remove any occurrence from production views. |
| TYPF-01 | SHOULD | Use one readable sans-serif family. At the design viewport, start at 24–32 CSS px for the page title, 24–36 for primary values, 16–20 for section titles, 14–16 for data and controls, and 12–14 for secondary notes and ticks. Resolve crowding by removing redundant content. |
| TYPF-02 | SHOULD | Right-align numeric table cells within a column and use tabular digits where supported. Keep the same units, precision and sign convention within a comparable series or column. |
| TYPF-03 | MUST | Display units and time labels without ambiguity. Use the dashboard’s declared locale and currency. Apply display rounding only after calculation and preserve exact source precision in authorized detail when the decision requires it. |
| TXT-01 | SHOULD | Title each chart with its monitoring question or a supported current finding. If the title states a finding, update it when filters, refresh or missing data invalidate it. |
| TXT-02 | MUST | Name every axis, series and marker that would otherwise be ambiguous. Prefer direct labels when legible. If needed, place a compact legend beside the plot with matching order and non-color cues. |
| TXT-03 | MUST | Show current scope, latest successful data-as-of time and stale or partial state where the reader makes the decision. Put full definitions, sources and cutoff notes in a labeled footnote or linked definition. |
| TXT-04 | SHOULD | Use the business’s own terms in plain language. Define each acronym at first use on the dashboard or in its linked definitions. |
| INT-01 | MUST | Make the default view answer the primary monitoring question without interaction. Show active filters, affected scope and a reset path when filters are available. |
| INT-02 | MUST | Make drill paths deliberate. Pass the selected entity and relevant scope, show the context at the destination, and keep required information available without hovering. |
| INT-03 | SHOULD | Reserve tooltips for extra precision or supporting context. Where present, make the same content reachable by keyboard and touch or via an equivalent visible route. |
| INT-04 | MUST | For every automated alert, define trigger, duration or repeat rule, recipient, action, resolution and stale-data behavior. Test that refresh does not cause repeated notifications for the same unresolved event. |
| ACC-01 | MUST | Operate every control with a keyboard. Use a logical focus order, visible and unobscured focus, and a way to leave each control. Preserve focus during refresh and return it predictably after dismissing a panel. |
| ACC-02 | MUST | Give each chart a meaningful accessible name and provide an equivalent readable summary or data table. Expose control names, values, states and table headers. Announce meaningful changes without reading every refresh. |
| ACC-03 | MUST | Keep text and controls usable at 200% text resize. At 320 CSS px width or equivalent 400% zoom, reflow content in reading order without page-level horizontal scrolling. Where a chart or data table genuinely requires two dimensions, contain its scrolling and provide an equivalent readable alternative. |
| ACC-04 | MUST | Verify contrast and non-color cues, complete the core tasks by keyboard, and read the overview, changed filters and detail using a supported screen reader. If the chart surface cannot provide equivalent access, include an accessible route before release. |
| ACC-05 | MUST | Make hover or focus disclosures dismissible, hoverable and persistent where WCAG 1.4.13 applies. Provide a visible or equivalent touch path for the same information. |
| GOV-01 | MUST | Before release, have the dashboard owner validate the decisions and measures, and a BI peer verify the checklist against the actual delivery surface. Close all MUST findings and record evidence for every pass or inapplicable rule. |
| GOV-02 | MUST | Record each SHOULD departure with its rule ID, affected audience, evidence, risk, mitigation, approver and expiry or review date. MUST failures block release. Fix the failure or use an approved delivery surface that passes. |
| GOV-03 | SHOULD | Re-review when audience, scope, measure definitions, targets, refresh, major layout or interactions change. Record the dashboard version and the standard version reviewed. |