An independent study reference written by Dr Phuc V. Nguyen. It is not official subject material — for assessment requirements always follow your subject outline and vUWS.

Visual Communicationintermediate

Dashboards

Q: A monitoring tile shows customer churn for the latest reporting period at three point one per cent, in large type, with no other marking. What is the most important thing missing?

A reference point, so the reader can tell whether that figure is normal. A monitoring tile exists so that someone can tell at a glance whether to act. A bare figure cannot support that judgement, because the reader has no idea whether it is typical, better than the prior period, or past the point of concern. Adding a target, a prior-period comparison or a normal band is what turns the number into a decision cue.

Q: A board asks for a strategic dashboard refreshed every thirty seconds. What is the strongest objection?

The measures a board acts on move on a quarterly cycle, so second-by-second refresh surfaces noise without changing any decision. Refresh frequency should follow the decision cycle. Strategic measures such as market position or margin trend are acted on over quarters, so sub-minute updates show only routine fluctuation and encourage reactions to noise. Technical feasibility is not the issue. The mismatch between the clock of the display and the clock of the decision is.

A dashboard is a monitoring instrument. It places the small set of measures a defined group has to watch onto one screen, so deviations become noticeable without hunting. Three constraints define it. It is standing rather than one-off. It serves specific recurring decisions. Every tile carries a reference point, such as a target, a prior period or a normal range, because a number with nothing to compare it to cannot trigger action. Dashboards are usually classed as operational, on a clock of minutes to hours, tactical, on a clock of days to a month, and strategic, quarterly. Confusing the three is the most common design failure.

Why it matters

Think of the instrument panel in a car. It shows speed, fuel and engine temperature, not the full service history. Each dial has a normal range, so a glance tells you whether to act. A dashboard showing forty numbers with no ranges is a service history bolted to the windscreen. You would still have to read all of it, and you would have to look away from the road to do it.

Before you read on — recall

A monitoring tile shows customer churn for the latest reporting period at three point one per cent, in large type, with no other marking. What is the most important thing missing?

Formulas

Variance against target, relative and absolute

v_{\text{rel}} = \dfrac{A - T}{T}, \qquad v_{\text{abs}} = A - T

A tile comparing an actual

A

to a target

T

can report the gap two ways, and the two say different things. With a conversion target of four per cent and an actual of three point six per cent, the relative variance is minus ten per cent while the absolute variance is minus zero point four percentage points. Reporting only the relative figure makes small-base measures look wildly volatile, and reporting only the absolute figure hides how far off a small target really is. Label which one the tile shows.

Setting an alert threshold from natural variation

\tau = k \, \sigma_{v}

A traffic light only helps if crossing it means something. One workable rule sets the threshold at a multiple

k

of the historical standard deviation of the measure, so the light changes when the value moves further than routine period-to-period variation would explain. If conversion normally wobbles by half a percentage point from one reporting period to the next, a light that turns amber at a tenth of a point will be amber in almost every period, and the team will stop looking at it within a month.

Worked examples

Scenario

A dashboard built for an executive team carries thirty-two tiles across three tabs. Six months on, the team still asks the analytics group for the same three numbers by email.

Solution

The dashboard was built around available data rather than around decisions. The fix starts by naming the recurring decisions the team actually makes, for instance whether to shift marketing spend between channels, whether to escalate a supply issue, and whether the hiring plan is on track. Each decision gets one tile with a reference point and a threshold that says when to act. Everything else moves to a linked detail view. A short dashboard that answers three standing questions gets used, a comprehensive one that answers none does not.

Scenario

A support desk wants a dashboard for team leaders while they are on shift.

Solution

This is an operational dashboard, so its clock is minutes. The tiles should be current queue depth against staffed capacity, longest wait right now against the service commitment, and abandonment rate over the last hour against its normal band. Each carries an attached action, such as moving an agent from email to phones. Monthly satisfaction scores and quarterly cost per ticket belong on a tactical or strategic view, because nothing a team leader can do in the next ten minutes will move them.

Common mistakes

✗A dashboard should contain everything a stakeholder might want. Completeness competes directly with monitoring. Every tile not tied to a decision raises the cost of scanning the ones that are, and the usual result is that the whole screen stops being read.
✗A dashboard is a way to present analysis. A dashboard supports repeated monitoring of a stable set of measures. A one-off investigation with a conclusion is a report or a narrative, and forcing it into tiles strips out the argument that made it useful.
✗Real-time refresh makes a dashboard better. Refresh should match the decision cycle. Updating a measure faster than anyone can act on it adds noise, invites over-reaction to normal variation, and consumes engineering effort better spent on data quality.
✗If the numbers are right, the dashboard is right. A tile with a correct number, no comparison point and no threshold still leaves the reader unable to judge whether anything is wrong. Correctness is necessary and not sufficient.

Revision bullets

•A dashboard is a standing monitoring instrument, not a report
•Every tile needs a reference point: target, prior period, or normal range
•Operational, tactical and strategic dashboards run on different clocks
•Relative and absolute variance say different things, so label which is shown
•Set alert thresholds from natural variation, not from preference
•Refresh rate should follow the decision cycle, not the data feed

Quick check

A monitoring tile shows customer churn for the latest reporting period at three point one per cent, in large type, with no other marking. What is the most important thing missing?

A board asks for a strategic dashboard refreshed every thirty seconds. What is the strongest objection?

Connected topics

BI vs Analytics Star Schema Measures & Dimensions Decision Support KPI Design Visual Variables Choosing a Chart Data Storytelling Visual Mistakes Social Analytics Design Thinking

More in Visual Communication

Purpose of Visuals Truncated Axes Chart Ethics

Sources

Few (2006)
Few, S. Information Dashboard Design: The Effective Visual Communication of Data. O'Reilly Media, 2006.
Sets out the single-screen, at-a-glance definition and the operational, tactical and strategic split.
Eckerson (2010)
Eckerson, W. W. Performance Dashboards: Measuring, Monitoring, and Managing Your Business. 2nd ed. Wiley, 2010.
Treats dashboards as part of a performance-management system rather than as a display technology.
Wexler, Shaffer & Cotgreave (2017)
Wexler, S., Shaffer, J., & Cotgreave, A. The Big Book of Dashboards. Wiley, 2017.
Worked dashboard cases with the decision each one supports stated explicitly.

How to cite this page

Dr. Phil's Quant Lab. (2026). Dashboards. Derivatives Atlas. https://phucnguyenvan.com/concept/ba-dashboards