Business Intelligence Dashboard Design Work Sample for Analysts and Management Information TeamsAnyone can build a dashboard. This measures whether you would build the right one.
Forty-two exercises on the decisions a dashboard is made of: which decision it is for, which metric answers it, which form shows it, what gets the space, what triggers an alert, and what is left out. Four real reporting briefs, each a named decision with a named audience and a named cadence, ask you to allocate a scarce resource across five candidate elements: the space above the fold on a Monday depot screen, the emphasis on a month-end collections page, twelve alerts a shift on a ward bed board, and the emphasis on a quarterly board page. Your shape is matched against a simulated five-judge expert panel, chance-corrected, and the report prints the panel's own agreement beside your number. Nothing is built or uploaded: this measures the design decisions, not tool skill.
The first dashboard assessment that scores the allocation, not the tool
The Business Intelligence Dashboard Design Work Sample is a fifty-minute applied skill assessment for analysts and management information teams. It measures the design decisions a dashboard is made of, scored as the match between your allocation of fold space, emphasis and alerts across four reporting briefs and a simulated expert panel's allocation, with the panel's agreement printed beside the result.
The failure dashboards actually have is not a wrong chart. It is a right chart in the wrong place: a large tile carrying a total nobody acts on, a ranked table at the decision's grain squeezed below the fold, a pie kept because every page has one, and twelve alerts a shift spent on states the board already shows. Every element can be defended on its own. Together they make a page that is looked at and does not change a decision. Tool courses teach how to build each element; no assessment we know of measures how a designer divides the scarce things, space, emphasis, alerts, across the elements a named decision could use. This one does, and only that.
The construct is grounded on the European Commission Joint Research Centre's DigComp 2.2 framework, competence area 3 on digital content creation and area 1 on information and data literacy, cited as a public reference; the instrument is not affiliated with or endorsed by the JRC. The keying rests on published work cited by author in the methodology note: graphical perception and the ranking of visual encodings, dashboard purpose types and the single screen, data-ink, leading against lagging measures tied to a decision, the measure-becomes-a-target effect, signal against noise and persistence before an alert fires, and alarm fatigue. Every exercise is original, and no business-intelligence, spreadsheet or visualisation product is named or used.
Twenty of the forty-two exercises are allocations. Four briefs, each a real request as it arrives: a depot manager and six route leads deciding every Monday which three routes get an extra van; a finance manager and four collectors deciding each month whom to call first; a ward shift coordinator deciding which patients to move in the next two hours; eight board members deciding on paper whether to approve two more clinics. For each, five candidate elements and one scarce resource: a share of the space above the fold, a place on a six-point emphasis ladder, or a number of the twelve alerts a shift. Twenty-two keyed exercises test the same judgements from the other side: which sentence is complete enough to design from, which metric answers a weekly decision, which form compares fourteen clinics most accurately, which additions make a plot harder to read, and which alerting fix addresses which problem.
The scoring is an expert-vector allocation match, done the way the measurement literature says it has to be done. Your allocation on each brief is correlated with the panel's, so that your overall level and spread drop out and only the shape is scored. The correlation is then read against the correlation an ordinary respondent would show, where ordinary means somebody drawing each element's allocation from the way designers typically allocate it, declared per element and printed on the report. Zero on the scale means indistinguishable from somebody who knows nothing but allocates the way designers typically do. The expert vector is simulated: five named judges, each derived from a distinct published position, carry one value per element; their agreement is computed and printed with the word SIMULATED wherever the panel appears, and an element on which the judges split beyond a declared threshold is printed as contested and excluded from the match. A reader who presses one value throughout is refused a figure, in plain words, rather than scored.
The report is a verdict card. The headline finding in a sentence, then one number, then the 68 per cent band as a shaded block and the 95 per cent band as an outlined box, both written in words beneath. Under the number, the four briefs as contributors, each with its own match, its band and whether an element was contested; beside the number, the panel's own agreement, because a match is only as good as the vector it was matched against. Then a fold mock-up: each brief drawn as the dashboard it would make, the blocks above the fold as wide as the share you gave them, yours beside the panel's. Then every element, sorted by gap, as a labelled bar pair with the five judges drawn as ticks on the panel's bar and the gap written in words.
Around the card: the keyed strand reported beside the index and never folded into it, the five decisions each with its brief and its keyed part, a strength paired with what it costs, one boxed if-then change drawn from your largest gap, the movement a re-sitting would have to clear to count as real change, the reliability table with every refusal printed, what zero means with the declared marginals, and the plain list of what the report does not say. At Rs 999 inclusive of GST in India, or US$9.99 elsewhere, that report is the product: dashboard and business-intelligence courses in India run from a few hundred rupees for a recorded video to many tens of thousands for a classroom programme, and none of them scores the allocation.
What you walk away with
How closely the shape of your allocations across four briefs follows the simulated panel's shape, after your level and spread drop out, chance-corrected so that zero is the designer who allocates the way designers typically allocate and 100 is the panel's shape exactly. The 68 per cent band is drawn as a shaded block, the 95 per cent band as an outlined box, and both are written in words beneath it.
The Monday route stand-up, the month-end collections review, the ward bed board and the quarterly board page: for each, your correlation with the panel, the ordinary respondent's correlation, the chance-corrected index and its band, printed as a contributor and never as a figure of its own. A brief answered with one value throughout is refused a match in words, not scored.
Five simulated judges, each derived from a distinct published position, carry one value per element. Their agreement is computed as an intraclass correlation and printed on the card with the word SIMULATED, together with the elements on which they split beyond the declared threshold, which are printed as contested and excluded from every match.
Each brief drawn as the dashboard it would make, yours beside the panel's: blocks above the fold as wide as the share you gave them, twelve alert slots filled per condition, a six-rung ladder with the elements placed on it. Then every element sorted by gap, as a labelled bar pair with the five judges drawn as ticks, and the gap written in words.
Twenty-two exercises on naming the decision, choosing the metric, choosing the form, giving the space and setting the alert, chance-corrected against declared per-option marginals and reported on the same scale beside the index. Allocation and recognition are different kinds of evidence, and the report does not average them.
A single boxed if-then sentence naming a specific trigger and a specific replacement behaviour, drawn from your largest gap against the panel on a non-contested element. Beside it, the number of points a re-sitting after four to six weeks would have to move to count as real change rather than measurement error.
Inside your report
Illustrative sample — your report is generated from your own responses.
Across 4 briefs and 18 scorable elements, your allocations sat closer to the simulated panel’s shape than a designer who allocates the way designers typically allocate.
| Brief | Match r | Null r | Index | 68% / 95% | Contested |
|---|---|---|---|---|---|
| The Monday route stand-up | 0.86 | -0.02 | 86 | 52 to 100 / 19 to 100 | 1, excluded |
| The month-end collections review | 0.71 | 0.19 | 64 | 30 to 98 / -3 to 100 | none |
| The ward bed board | 0.52 | 0.27 | 34 | 0 to 68 / -33 to 100 | 1, excluded |
| The quarterly board page | 0.48 | 0.19 | 36 | 2 to 70 / -31 to 100 | none |
How to read it:the sentence comes first, then the number, then both bands drawn and written. The panel’s agreement sits beside the number with the word SIMULATED, because a match is only as good as the vector it was matched against. The fold mock-up draws the dashboard you would have made beside the one the panel would have made.
Two bars per element on its own scale: yours above, the SIMULATED panel’s beneath with each judge as a tick, so the panel’s disagreement is visible at the element where it occurs. Sorted by gap, largest first. more than the panel · less · near · contested, excluded.
| Element | You | Panel (SIMULATED) | Gap |
|---|---|---|---|
| A5 the forecast against capacity, per route | 15% of the fold | 34% of the fold | 19 points less than the panel |
| B4 calls made by each collector last month | 5 on the ladder (6 = headline) | 2.2 on the ladder | 56 points more than the panel |
| C4 a new admission request has arrived | 4 of 12 alerts | 2.2 of 12 alerts | contested, excluded |
| D1 a dot plot of the 14 clinics against capacity | 6 on the ladder (6 = headline) | 5.8 on the ladder | near the panel |
How to read it: twenty rows on a full report. A contested element shows its ticks spread wide and is a finding about the panel, not about you; it is left out of every match. The gap is written in words, and the one change on the report is drawn from the largest gap on a non-contested element.
Built for
- Analysts and reporting specialists who build dashboards for other people and want evidence about their design decisions, rather than another certificate in a tool
- Management information and finance teams asked to turn a request for a dashboard into a page somebody acts on, who want to know where their allocation departs from the published literature
- Analytics managers and training providers choosing or developing dashboard designers, who need one specific change per person with the evidence and the panel's disagreement beside it
- Anybody preparing for a business-intelligence or reporting role who wants a work sample on the decisions, scored against a declared panel, without any tool, file or upload
Find out which decision your dashboards are really for before a reader finds out they are for none
42 exercises across seven formats · about 50 minutes · four briefs, twenty allocations, one allocation index with both bands, a simulated five-judge panel with its agreement printed, a fold mock-up per brief, every element against the panel, and twenty-two keyed exercises beside it. Rs 999 in India inclusive of GST, or US$9.99 elsewhere. Sit it again after four to six weeks of the boxed change.
₹999 (incl. GST) · assessment and full report, nothing further to pay
Frequently asked questions
It measures the design decisions a dashboard is made of: which decision it is for and at what cadence, which metric answers it at its grain, which chart form shows it most accurately, what earns the space above the fold, what earns an alert, and what is left out. You do not build, upload or render anything. Four reporting briefs ask you to allocate a scarce resource, fold space, emphasis or twelve alerts a shift, across five candidate elements, and twenty-two keyed exercises test the same judgements from the other side. It does not measure skill with any tool, data modelling, query performance or producing a file in any product.
By an expert-vector allocation match. On each brief your allocation across the elements is correlated with a simulated expert panel's, so your overall level and spread drop out and only the shape is scored. That correlation is read against the correlation an ordinary respondent would show, where ordinary means somebody drawing each element's allocation from the way designers typically allocate it, declared per element and printed on the report. Zero means indistinguishable from somebody who knows nothing but allocates the way designers typically do; 100 is the panel's shape exactly; a negative figure is a shape further from the panel's than ordinary. A reader who presses one value throughout is refused a figure rather than scored.
No panel has met. Five named judges, each derived from a distinct published position in the data-presentation literature, graphical perception, dashboard purpose types, data-ink, decision-and-metric definition, and variation and alerting, carry one value per element, written into the instrument so every number is reproducible. The consensus is their mean. Their agreement is computed as an intraclass correlation and printed on the card beside your number, with the word SIMULATED wherever the panel appears, because a match statistic is only as good as the vector it was matched against. An element on which the judges split beyond a declared threshold is printed as contested and excluded from the match. A real panel of five or more raters replaces the simulated one when it meets.
Data Dashboard Reads measures reading a dashboard somebody else built: estimating values, totals and proportions from drawn charts and judging the soundest response to a reading. Data Visualisation and Storytelling measures the narrative told from data: annotation, audience-first storytelling and misleading-graph pitfalls. This measures the design decision, which element gets the space, the emphasis and the alert for a named decision, and never asks you to read a chart or to say anything about one. The report names both neighbours in its refusal block.
Rs 999 in India inclusive of GST, or US$9.99 elsewhere, one time; 26 credits on a business plan. Forty-two exercises across seven formats take about fifty minutes. The report names a re-sitting after four to six weeks of the boxed change and prints the number of points the allocation index would have to move to count as real change rather than measurement error. No percentile appears anywhere, and you are ranked against nobody. Dashboard and business-intelligence courses in India run from a few hundred rupees for a recorded video to many tens of thousands for a classroom programme, and none of them scores the allocation against a declared panel.
Each one takes a single capability, puts you inside the situations where it is actually tested, and scores your choices against published evidence — with a report designed for that capability alone, not a template. They span hiring, compliance, education, operations and personal skill.
Browse the catalogue →Methodology: Forty-two original exercises across seven formats: twenty allocation exercises (ten sliders and ten six-point emphasis ladders) across four reporting briefs, two first-move scenarios with graded options, seven single-choice exercises, six select-every-line exercises, four match-the-following exercises and three ordering exercises. DECLARED RESPONSE INSTRUCTION: one instruction for the whole instrument, what does this brief WARRANT, judged against the published data-presentation and information-design literature; a knowledge-style instruction about the design, not a behavioural tendency, mixed with nothing. CONSTRUCT STATEMENT: This measures the design decisions a business-intelligence dashboard is made of: which decision the dashboard is for and at what cadence, which metric answers that decision at its grain, which chart form shows it most accurately, what earns the space above the fold, what earns an alert, and what is left out. It is measured as the correlation between your allocation of a scarce resource across the candidate elements of four reporting briefs and the mean allocation of five SIMULATED expert judges, chance-corrected against how designers typically allocate. It does not measure skill with any dashboard or spreadsheet tool, data modelling, query writing or performance, or building a file in any product; nothing is built or uploaded in this sitting. It does not measure reading a dashboard somebody else built, which is the Data Dashboard Reads assessment in this catalogue, nor the narrative told from a chart, which is the Data Visualisation and Storytelling assessment. It measures the decisions, not the read and not the story. THE CAPABILITY GAP, STATED: the pipeline asked for a file-upload work sample. The platform has no file upload and no dashboard execution engine, so this instrument delivers the same construct, the design decisions, on existing item types, and says so in its title, description, report and sales page. Nothing is built, uploaded or rendered by the candidate. FRAMEWORK: the construct is grounded on DigComp 2.2, the European Digital Competence Framework for Citizens, published by the European Commission's Joint Research Centre (Vuorikari, Kluzer and Punie, 2022), competence area 3, digital content creation, and area 1, information and data literacy; the framework is cited as a public reference, the instrument is not affiliated with or endorsed by the JRC or the European Commission, and the framework's name appears in no title, label or call to action. SCORING DESIGN, C4, expert-vector allocation match, chance-corrected as C2: for each brief, your allocation vector across its five elements is correlated (Pearson) with the SIMULATED expert consensus vector over the elements you answered that are not CONTESTED; the expected correlation of a respondent drawing each element's allocation from THAT ELEMENT'S DECLARED MARGINAL (value_base_rate) is estimated by a seeded Monte Carlo of 400 draws inside the builder; the brief's index is 100 (match - E[null]) / (1 - E[null]); the instrument's index is the mean over briefs with at least three scorable elements, with its standard error and 68 and 95 per cent bands from the declared modelled standard deviation and an assumed omega printed unrounded. A correlation is used rather than a distance so that your overall level and spread drop out: a reader who gives everything a little more is scored on shape, not on elevation. THE EXPERT VECTOR IS SIMULATED: C4 requires five or more raters with a reported agreement statistic. No panel has met. Five named, simulated judges, each derived from a distinct published position (graphical perception; dashboard purpose types; data-ink; decision-and-metric definition; variation and alerting), carry one value per element, authored in the seed and printed on the report; the consensus is their mean; inter-rater agreement is computed in the builder as ICC(2,k), two-way random effects, absolute agreement, mean of five raters, on the twenty elements normalised to their own range, with the mean pairwise correlation beside it, and the word SIMULATED is printed wherever the panel is named. An element on which the five judges' standard deviation exceeds 10 per cent of the element's range is CONTESTED: it is printed as a finding and excluded from the match for that brief. Two of the twenty elements are contested by design, the delivery map and the every-admission alert. Flatness is tested on the RAW presses before any mapping or reversing: a brief answered with one value throughout has no shape and is refused a match, never scored; three or more flat briefs refuse the index. An unanswered exercise leaves the numerator, the denominator and the chance term; an empty sitting scores exactly zero and is not reportable; a sitting under 24 answered exercises is withheld with the refusal printed in place of every figure. The twenty-two keyed exercises are chance-corrected against their declared per-option, per-pair and per-position marginals and reported BESIDE the allocation index, never folded into it; a competency carries a number only with at least eight answered exercises and omega at or above .70 unrounded, otherwise a three-way word. No percentile appears anywhere; there is no norm group. REPORT DESIGN, D1, the Verdict Card: one headline sentence, one number, the 68 per cent band as a shaded block and the 95 per cent band as an outlined box, both written in words; the four briefs as contributors with their own match and whether an element was contested; the panel's agreement printed on the card beside the number with the word SIMULATED; what zero means with the declared marginals; the element-level view sorted by gap with the gap in words; a fold mock-up drawing each above-the-fold brief as proportional blocks, yours beside the panel's; the keyed strand beside the index; the reliability table; the refusal block; a strength paired with its downside; and one boxed if-then change with the re-measurement threshold. SOURCES: Cleveland and McGill, Graphical perception: theory, experimentation, and application to the development of graphical methods (Journal of the American Statistical Association, 1984), and Heer and Bostock, Crowdsourcing graphical perception (CHI, 2010), for the ranking of encodings keyed on the chart-form exercises; Few, Information Dashboard Design (O'Reilly, 2006), for dashboard purpose types, the single-screen principle and the glance, and Few, Show Me the Numbers (Analytics Press, 2nd ed., 2012), for table-versus-graph choice and the bullet graph; Tufte, The Visual Display of Quantitative Information (Graphics Press, 1983), for data-ink and the removal of what carries no data; Wexler, Shaffer and Cotgreave, The Big Book of Dashboards (Wiley, 2017), for dashboard scenarios organised by the decision and the audience; Ware, Information Visualization: Perception for Design (Morgan Kaufmann, 3rd ed., 2012), for pre-attentive salience behind the alert and top-left exercises; Shneiderman, The eyes have it: a task by data type taxonomy for information visualizations (IEEE Symposium on Visual Languages, 1996), for overview, zoom and filter as the analytical purpose type; Kaplan and Norton, The balanced scorecard: measures that drive performance (Harvard Business Review, 1992), for leading against lagging measures tied to a decision; Murray and Berwick, Advanced access: reducing waiting and delays in primary care (JAMA, 2003), for the third-next-available-appointment access measure; Strathern, Improving ratings: audit in the British university system (European Review, 1997), for the measure-becomes-a-target effect keyed on the activity-count exercises; Shewhart, Economic Control of Quality of Manufactured Product (Van Nostrand, 1931), and Wheeler, Understanding Variation: the key to managing chaos (SPC Press, 1993), for signal against noise and persistence before a threshold acts; Cvach, Monitor alarm fatigue: an integrative review (Biomedical Instrumentation and Technology, 2012), for alert fatigue and the cost of an alert that changes no action; Shrout and Fleiss, Intraclass correlations: uses in assessing rater reliability (Psychological Bulletin, 1979), and McGraw and Wong, Forming inferences about some intraclass correlation coefficients (Psychological Methods, 1996), for the ICC(2,k) printed on the card; Cronbach and Gleser, Assessing similarity between profiles (Psychological Bulletin, 1953), for the elevation, scatter and shape decomposition that motivates a correlation match; Hattie and Timperley, The power of feedback (Review of Educational Research, 2007), for the feed-up, feed-back and feed-forward structure of the report; Gollwitzer and Sheeran, Implementation intentions and goal achievement: a meta-analysis (Advances in Experimental Social Psychology, 2006), for the if-then form of the one change; Haladyna, Downing and Rodriguez, A review of multiple-choice item-writing guidelines (Applied Measurement in Education, 2002), for cue control. All forty-two items are original works written for this instrument; no item from any published questionnaire, certification programme, vendor curriculum or commercial dashboard course is reproduced or adapted, no business-intelligence, spreadsheet or visualisation product is named or used, no competitor is named anywhere, and the instrument is not affiliated with or endorsed by any author, publisher, professional body, vendor or public institution named above.