Understanding Requirements: Who Is the Dashboard For?
Why dashboards fail
In a survey of 450+ dashboard builders (the book Dashboards That Delivered), more than 70% agreed that most dashboards fail. The usual cause is rarely the charts, the colours or too much data. It is that nobody spent enough time upfront understanding who will use the dashboard and what they need from it. Work done before design decides whether the dashboard gets used.
User segments
One topic ("quarterly performance") hides several different dashboards:
| Person at TrendE | What they need |
|---|---|
| CMO | Overall market trends: is TrendE gaining or losing share quarter over quarter? |
| Category manager (women's wear) | Funnel of views, add-to-carts and purchases at SKU (lowest product) level to spot underperformers |
| Warehouse head | Today's fulfilment exceptions, near real time, to act within the hour |
A single master dashboard for all of them would be too dense for the CMO, too shallow for the category manager and useful to nobody. Identifying user segments is the first essential step.
Stakeholders, users, sponsors, pilots, IT
- Stakeholder: usually the person who commissions the dashboard and holds budget or authority (e.g. the CMO). User: the person who opens it regularly (category or regional manager; may include the CMO). They are often different, and their needs differ.
- Stephen Few's point: asking only the requester tends to yield a description of what they are already used to getting, not what they need. The CMO may ask for "the same monthly report, faster"; the daily user knows what is missing. Talk to both the funder and the user.
- Senior sponsor: a senior stakeholder engaged from day 1, not just informed at launch. Adoption problems (distrust of numbers, attachment to old spreadsheets, two departments disputing whose version is right) need someone senior to back the initiative and align people. Involvement makes it partly their initiative.
- Pilot users: a small set from different functions who test the dashboard during the build. Choose people "on your side of the table, not across from it". Every early version has bugs (broken filter, number not matching a spreadsheet, slow chart); an invested pilot reports them constructively, a sceptical one may use the first bug to write the project off, and that reputation is hard to undo. Have the sponsor nominate pilots from each segment, which makes them feel invested.
- IT: controls access to source systems and knows the data. Involve them from week 1, not at the data audit. Otherwise the request lands in their queue as another context-free ticket: slower and lower priority.
Key takeaways
- Most dashboards fail from not understanding users, not from chart or colour choices.
- Different segments need genuinely different dashboards.
- Stakeholder (commissions) is not always the user (opens daily).
- Enrol sponsor, pilots and IT at the start.
Asking the Right Questions
A sentence such as "I want to see performance" is not a requirement; it is the opening line of a conversation. Building a chart straight away is the mistake; the right first move is asking what decision they will make with the number.
Asking the Right Questions
The question checklist: why, what, how and when, where
(Inspired by Stephen Few's work on dashboard requirements.)
| Category | Questions |
|---|---|
| Why (role and objective) | Job title, core responsibilities; business problem being solved; actions or decisions taken from what they see; what they look at now to know the day or week is going well |
| What (data and metrics) | KPIs; benchmark for each (target, same period last year, budget); dimensions to split by (region, product line, location); granularity (daily vs monthly), which strongly shapes chart design |
| How and when | Frequency (every morning, weekly, live stream); alert triggers (e.g. alert when sales fall below 90% of target); next step on spotting a problem (drill down or high-level summary) |
| Where | Device (desktop, laptop, phone, wall screen); presentation use (needs PDF or PowerPoint export); who else needs access and whether they see the same view (security controls) |
On next steps: senior levels typically pass the problem to someone junior and ask for an explanation; mid and junior levels want to drill into detail because they must answer upward.
Test on the sales head's statement ("performance trend line over last 6 months, updated as of yesterday, versus targets, with product-category breakdown"): the what is covered (metric, comparison, one grouping); how and when gets one weak hint ("as of yesterday"); why is blank; where is absent. This is normal: people volunteer the concrete what easily, while the why and the rest take real conversation.
Interviewee types
- Blank slatecannot imagine an ideal dashboard ("just show me sales").
- Laundry listgives about 30 unprioritised metrics, which is not a requirement either.
Interviews (one-on-one with senior stakeholders and a representative user per segment) should be conversations aimed at what the person needs in order to act and decide, not a read-out script. Requirements gathering is as much art as science.
Four elicitation techniques
| Technique | How it works | Best for | TrendE example |
|---|---|---|---|
| Walk me through your day | Ask about a typical day or week and when they look at data | How and when; suits both types (concrete for blank slate; filters the list for laundry list, since rarely-used metrics are forgotten) | Sales head checks only the headline "tracking to target?" every Monday before the leadership call, drilling down only if something looks wrong: summary is enough, refresh is regular |
| Show me what you already look at | Review their actual existing reports | What; also reveals why, pain points, what they act on and ignore | Marketing head's manual spreadsheet of campaign spend vs leads, a data item he had not mentioned. (Reports also reveal effort, such as a PowerPoint taking 2.5 days to compile.) |
| Cross-pollinate across departments | Mention what another department asked for | Sharpens what, sometimes surfaces why. Reaction is either "we need that too" or "no, we need this instead" | Ops head, on hearing sales wants category performance: "we need performance by fulfilment centre and carrier, where delays happen" |
| Rough mock-up | Sketch a rough (even wrong) mock-up mid-interview | Benchmarks and alert triggers; people critique better than they specify | Sales head, shown one quarterly sales number: wants same quarter last year alongside, and a red flag if more than 5% behind target |
Not every question needs a technique: device and presentation use are answered directly. Use techniques where direct questions give vague answers, mainly why and parts of how and when.
Key takeaways
- Use the why / what / how and when / where checklist as a guide, not a script.
- Recognise blank-slate and laundry-list responders and adapt.
- Four techniques: walk through the day, existing reports, cross-pollination, rough mock-up.
- Direct questions suffice for where.
Consolidating Requirements and the Hybrid Dashboard
The consolidated table
The three departments' requirements (typical e-commerce metrics):
| Metric | Who needs it, and split |
|---|---|
| GMV (gross merchandise value, total revenue), total orders, average order value, conversion rate (of those who view a product, how many buy) | All three, as overview |
| Return rate (of 100 delivered orders for a product, how many returned; works like negative sales) | Sales: by product category. Marketing: by channel or campaign. Operations: by fulfilment centre or carrier |
| Gross margin % | Sales only, by product category |
| Sell-through rate (percentage of inventory held that has been sold) | Sales (fast and slow movers) and operations (by fulfilment centre; unsold stock is dead stock). Higher is better |
| Customer acquisition cost, return on ad spend (revenue earned per budget spent), cart abandonment rate (items put in carts that do not convert; informs site or app design and follow-up emails) | Marketing |
| Order cycle time (buy click to delivery at the door), on-time delivery % (share of orders meeting the delivery promise) | Operations only |
| Viewing device or format | All three |
The top five (GMV, total orders, AOV, conversion rate, return rate) are common ground from the same sales data; every metric below them is wanted by only one or two departments. Splits differ by department: for return rate, sales looks for a problem product line; marketing for channels bringing low-quality buyers or campaigns/promo codes (e.g. 70%-off promotions) or creatives that set mismatched expectations; operations for packaging or delivery issues by centre and carrier.
Device matters for layout: sales mostly uses a laptop (trend view sometimes pasted into a leadership slide); operations uses a phone on the warehouse floor. This shapes small-screen and projection design decisions.
Three structuring options
| Option | Pros | Cons | Use when |
|---|---|---|---|
| One size fits all (single shared dashboard) | Single source of truth; lowest maintenance | High clutter; irrelevant data for any viewer; dilutes the sharp view | Departments have heavy overlap in daily workflows and goals, not just a few shared numbers |
| Fragmented (separate dashboards) | Fully tailored | Expensive to maintain; risk of different calculations (e.g. revenue) from different sources; no single leadership overview | Departments share nothing: no overlapping goals or data sources |
| Hybrid (shared core plus department-specific views) | Focused core plus tailored detail | Needs more careful upfront layout planning | Teams share a real core but have distinct operational needs |
TrendE's table (5 common KPIs, the rest custom) points to hybrid: the five common KPIs go on an overview page (perhaps number plus trend line); everything else gets a dedicated page per department, including the common KPIs split by that department's categories (some repetition is intended).
Prioritising and playing back
Users name many metrics. Those a department head calls critical go in phase 1; the rest go to phase 2. Example: the marketing head mentions social media engagement but omits it from his top metrics, so it goes to phase 2. Everything mentioned is logged, not rejected: if people think their idea was quietly dropped they lose trust in the roadmap.
Then reconvene all heads in one room and play back: what we heard from each of you; the shared core on the overview and why; each dedicated tab's contents; what is parked for phase 2 (by their own priorities). Use a rough, sketched mock-up (not built in Power BI or Excel): polished mock-ups make people assume it is nearly final, while an obviously unfinished one invites real critique before committing. Close with sign-off.
Key takeaways
- Shared metrics form the overview; unique metrics get dedicated pages.
- Choose among one-size-fits-all, fragmented, hybrid by the degree of real overlap.
- Prioritised metrics go in phase 1; others are logged for phase 2.
- Play requirements back in a joint meeting using a deliberately rough sketch.
Data Audit and Feasibility
For every key metric: where does the data sit? Users and IT supply this (hence early IT enrolment). At TrendE: sales in the point-of-sale system, marketing spend in ad platforms, fulfilment data in the operations system.
Data Audit and Feasibility
Audit checklist
| Item | What to do and why |
|---|---|
| External sources | Ad data sits with the ad platform vendor; logistics with a third-party shipper. Treat each as its own discovery process: explicit permission, a different team and process, and vendor timelines are often far slower. Flag immediately and prioritise |
| Automated vs manual feeds | Connect directly to source systems, ideally via APIs, rather than file export and upload. Manual handoffs break silently (leave, forgetting, last month's file, an untrained replacement) and stale charts go unnoticed until someone senior questions the numbers. Where manual upload is unavoidable (e.g. quarterly forecast Excel from finance), record it and build upload training into onboarding |
| Excel as source of last resort | Manually edited, columns renamed or reordered, formats drift, nothing enforced. Dashboards need standardised data (e.g. "Bangalore" vs "BLR"). Example: a new team member capitalised all column headers, making the file incompatible and stopping the dashboard update |
| Data dictionary | Document explaining what each field means. "Revenue" may or may not include tax or delivery charges. Never assume from the name; get definitions from the data owner (colleagues often use different ones); definitions are later shown on the dashboard |
| Sample data | Ask for roughly 25 to 100 real rows. Check date formats, blank cells, text in numeric fields: things no description surfaces |
| Refresh frequency | Source must refresh at least as often as users want. Example: marketing wants returns data on the 1st of the month but the logistics partner supplies it only on the 7th; flag the mismatch |
| Known problems | Ask the owner directly (inconsistent fields, broken-tracking periods, unfixed duplicates). They are usually candid and it beats discovering issues three weeks into the build. Also ask the users who built their own reports |
| Revision after landing | Is data accurate at first arrival or revised later? |
| History | How far back does it go, and can past periods be revised? Needed for trends; some systems keep only a rolling window or archive history elsewhere |
| Technical partner | Do not audit alone if uncomfortable with technical talk; an IT person or analyst can spot dates stored as text, suspiciously clean hard-coded fields, or structural mismatches between systems. Cheap at audit, very expensive after build |
Worked example: revised media spend
The media department updates spend at month-end, but fully accurate figures arrive only the following month after vendor credit adjustments. March spend shown on 1 April is not final; it is accurate only on 1 May. Options: (A) permanently show only original figures, (B) let users overwrite history once corrected numbers arrive, (C) hide figures until final. Answer: B. The dashboard needs indicative first numbers (users want likely spend) and a way to overwrite the historical value when the real number arrives, rather than appending forward while last month stays wrong. The original plan (upload only the latest month, never overwrite) had to change for this department.
Feasibility verdict and sign-off
Give every metric an honest verdict: feasible now, feasible with extra work, or not feasible.
- Return rate (marketing wants it by the 1st): feasible with extra work because the logistics partner delivers on the 7th; refresh timing must be aligned. If done quickly it stays in phase 1, otherwise phase 2.
- Sell-through rate by fulfilment centre: not feasible yet. The warehouse system does not track received vs sold inventory consistently across locations, so data is not comparable; moves to phase 2 until fixed.
Reconvene users and stakeholders: what was checked and where data comes from; verdict with brief reasons; ask once more about known issues; get senior sponsors to intervene to speed access and fixes (e.g. the CMO says "get this done fast and it can still be phase 1"); obtain explicit sign-off on version 1 scope.
Key takeaways
- Audit: location, internal vs external, automated access, dictionaries, samples, known issues, first-pass accuracy, history.
- Excel is a last resort; manual handoffs silently fail.
- Every metric gets a feasibility verdict, then sign-off on v1 scope.
Phase-Wise Development
The biggest risk is often not time, changing requirements or messy data, but holding release until the dashboard is "100% complete" and beyond criticism. Seeking perfection is the enemy of progress; if you wait for flawless data you never launch.
Phase-Wise Development
Phase 1 = MVP
It gives stakeholders something simple, clean and functional in weeks, not months. Phasing also guards against clutter (flashy charts, years of history, every available metric). Nothing is deleted: other ideas go to the phase 2 list, scheduled rather than rejected.
Bottlenecks that push a metric to a later phase
| Bottleneck | Meaning | TrendE example |
|---|---|---|
| System | Data not tracked consistently yet | Sell-through by fulfilment centre: warehouse system inconsistent across locations; time for operations to fix tracking |
| Process | Data depends on external vendors or manual handoffs | Return rate: marketing wants the 1st of the month, logistics partner delivers on the 7th; a process mismatch, not a system problem |
| Governance | Data ready but revised after landing | Media spend reported 1 April changes by 1 May from vendor credit adjustments; not anyone's mistake, just how credits work |
The core phase 1 launches on stable data while IT, marketing and logistics clear blockers in the background.
The roadmap
The roadmap lists, per phase, the items, target users, timeline and blockers. Phase 1 normally has none; phase 2 names blockers so everyone knows which team must fix what. It is presented to sponsors and stakeholders for explicit final sign-off before a single visualisation is built.
TrendE case: GMV is 100% clean and feasible, so it is a core KPI on the phase 1 overview. Return rate depends on renegotiating a vendor's delivery schedule (procurement), so it moves to phase 2. Phase 1 launches with GMV, total orders, AOV and conversion rate; procurement now has a clear deadline.
Sign-offs and scope creep
Business users are visual: seeing live data triggers new ideas ("just add one more metric"). It is natural, not a requirements failure, but asks accumulate into chaotic scope creep. The signed-off phase 1 scope is a shield: reply that it is a great idea outside phase 1 scope and log it for phase 2. The timeline is protected and the user feels heard.
A launch is a recurring cycle: tools such as Power BI and Tableau let you release a lean v1, watch usage, adapt and roll out the next phase. Some simple dashboards can be built in one pass; phasing matters most with multiple user segments and senior stakeholders, to manage expectations and show quick progress.
Key takeaways
- Phase 1 is the overlap of top-priority requirements and feasible data.
- Three bottlenecks: system, process, governance.
- Roadmap with blockers needs explicit sign-off before building.
- Signed scope guards against scope creep; later phases are a "parking lot".
Designing the Dashboard
Anatomy of a dashboard
The structure is tool-independent (Power BI, Tableau, Looker or future tools). A dashboard is usually a small connected set of pages or tabs, not one screen.
| Building block | Role |
|---|---|
| Visuals | Charts, tables, numbers; only one of the building blocks |
| KPI cards | The few (under 10) most important figures, prominently placed, usually near the top |
| Filters (slicers) | Let the viewer narrow to a region, date range or category, with other visuals updating; one dashboard serves many views |
| Navigation | Tabs, sidebar menu or buttons to move between pages |
| Hover and click interactions | Pop-ups or expansions revealing detail without redesigning the screen, which contains clutter |
Designing KPI cards
A KPI tells you whether things are broadly on track; a diagnostic tells you why. Senior executives often focus mainly on the KPI cards. An isolated number (net profit margin 12%) is uninterpretable: good, bad or average? A card needs context.
Four parts of a good KPI card:
- Number: large and unmistakable; often the only thing the eye needs.
- Label: plain-language name, using the term users know (GMV vs "total revenue"); spell out on first appearance if needed.
- Comparison or context: the reference point that gives meaning.
- Trend signal: up/down arrow or sparkline, showing direction and whether an issue is persistent, building up or sudden. Senior executives sometimes want a full trend line to gauge seriousness.
Add a tooltip definition on hover to remove ambiguity.
Three comparison types
| Comparison | Question | Notes |
|---|---|---|
| Versus target | Are we hitting what we committed to? | Goal-based; shown as percentage of target achieved; use when an agreed number exists |
| Versus prior period | Improving vs last month, quarter or same period last year? | Month-on-month, quarter-on-quarter, year-on-year; useful without a fixed target or when direction matters more |
| Versus industry standard | How do we stack up against the market? | Less common internally; valuable where meaning depends on external context (customer satisfaction, on-time delivery %, market share) |
Several comparisons on one card (CMO case): the CMO wanted sales versus budget, versus latest forecast (revised quarterly), year-on-year and a trend. Resolution: one card with the absolute sales number and percentage of target achieved (example given: number 200, 30% of target); the latest forecast is the default target; a filter switches the basis between forecast and budget and the target updates; year-on-year in the card's right corner; a small 6-month line chart below. This cut clutter while delivering everything.
TrendE overview cards
| Card | Comparison | Rationale |
|---|---|---|
| GMV | Versus target, plus sparkline or trend | Leadership has a quarterly revenue goal; likely the most-watched card |
| Total orders | Versus prior period | Weak alone; useful read with AOV |
| Average order value | Versus prior period | Place beside total orders: orders rising while AOV falls (more customers but smaller baskets, or existing customers buying less) can hide under stable GMV |
| Conversion rate | Versus target (internal historical benchmark) | Moves before GMV, so an early-warning signal |
| Return rate | Versus target, with trend or sparkline | A rising rate is worth catching early, before it drags GMV or gross margin |
Colouring (red/green) is deferred to the look-and-feel session; here the focus is the information each card carries.
Information flow and visual hierarchy
Eye-tracking findings: big numbers are noticed first; repetition does not help (five identical KPI cards: attention drops sharply by the fourth or fifth). A layout study adds that the most important chart works best left of centre where the eye lands first. Rule: the most important number goes top-left and must look different from its neighbours.
Five tools of hierarchy
| Tool | Effect (TrendE example) |
|---|---|
| Size | Bigger draws the eye first; GMV is the largest number |
| Colour and contrast | A splash of colour on a muted background pulls attention; use sparingly |
| Position | Top-left scanned first |
| Proximity | Close items read as related; total orders sit beside AOV |
| Typography | Weight and size signal importance; bold large number with small label |
Hierarchy concerns what is noticed first within a screen (scannability). Flow is the sequence across pages: start high, then let people drill down; get the broadest read in seconds and move to specific pages only if something looks off.
A dashboard's job is to communicate fast, not to impress; decoration competes with the important numbers. Two failure modes: too much detail with no clear place to look, and a loud headline with no path to investigate.
Three-question test: Do we know where to start? Do we know what is next? Can we tell when we are done? An overview page shows top-level numbers and gives a clear way forward.
Designing for three user levels
| Level | Example | Dashboard needs |
|---|---|---|
| Executive | CMO, CXO (monitors health and opportunities) | Performance vs targets plus trend; very simple, minimalist; not for deep analysis (they ask the next level); minimal learning; printable or exportable to PowerPoint for board meetings |
| Mid-level | Regional or category manager | What their bosses see, plus context and history; highly interactive with drill-down; built for diagnosis and finding action steps |
| Operational | Store manager, customer support | Simple, for immediate decisions; alerts and notifications; some drill-down; for spotting red flags and acting, not analysis |
TrendE's overview is executive tier; the sales, marketing and operations pages are mid-level. Those three pages must be true siblings: comparable slices at the same level below the overview, a top-level summary branching into parallel pages.
Progressive disclosure
Climbing the ladder of diagnostics (KPI to primary diagnostic to deeper) on a real screen without dumping everything: progressive disclosure, showing only what is needed now but making more easy to reveal.
| Mechanism | Example |
|---|---|
| Hover for context | Small detail appears without leaving the visual |
| Expand in place | Click return rate and a category breakdown appears beneath, still on the overview |
| Click through to a detail page | Click conversion rate and land on a page showing it by campaign |
The choice depends on how deep the next rung goes and how often it is needed; a full page is worth the extra click only for genuine deep dives. The same discipline (one headline number, context, easy way deeper) applies to a simple spreadsheet weekly tracker: a habit of thinking, not a software feature.
Wireframe
Before any chart type or colour, put hierarchy, flow, tiers, navigation and progressive disclosure into a structural wireframe: boxes for KPI cards, charts, filters and navigation, with no data, colours or branding. Pen and paper or a simple slide is fine. Ask the stakeholder "does this layout make sense?", not "do you like how it looks?", and get a sign-off. Collaborating with a UX designer helps with layout and navigation.
Key takeaways
- Top-left, biggest, different-looking: the key number.
- Five tools: size, colour and contrast, position, proximity, typography.
- Executive, mid-level and operational users need minimalist, diagnostic and alert-driven designs respectively.
- Progressive disclosure: hover, expand in place, click-through.
- Wireframe sign-off checks structure only.
Choosing appropriate charts
TrendE now has four pages: overview plus sales, marketing and operations. Charts follow the question type:
| Question type | Example | Chart |
|---|---|---|
| Comparison | Top cities by sales | Column chart (under about 6 categories); horizontal bar if more |
| Comparison over time | Sales trend over 6 months | Bar or line for a single series; line better for multiple series (e.g. by product line) |
| Composition | Menu-category share of orders | Pie or donut only if fewer than 4 slices; otherwise stacked bar, or 100% stacked bar for many slices. Position beats angle, so prefer stacked bars |
| Distribution | Delivery times across stores, and consistency | Histogram or box plot (box width shows variability, whiskers show outliers; best for consistency) |
| Relationship | Delivery time vs customer rating | Scatter plot (shows correlation, not necessarily causation) |
TrendE page-by-page
Overview (cards): GMV with callout and quarterly sparkline versus target; total orders and AOV together versus prior quarter; conversion rate and return rate as cards versus target, each with a quarterly sparkline.
Sales page (all comparison questions): gross margin %, sell-through rate and return rate each by product category over the last 6 months. Each gets a bar chart sorted highest to lowest with a trend line below. In Power BI the trend line can be linked to the bar chart so selecting a category updates the line (covered later).
Marketing page (comparison, four times): CAC by channel; ROAS by channel; ROAS by campaign; cart abandonment rate by channel. Four bar charts, each with a trend line.
Operations page:
- Order cycle time by fulfilment centre: bar chart plus trend line; the consistency requirement needs a box plot, showing spread and outliers so the team can tell which centres are consistent, which are too variable, and what outliers to investigate (wider box means more variable).
- On-time delivery % by carrier: bar, trend line and box plot. By region: with 4 regions a bar chart suffices; to find underperforming locations or geographic clusters (e.g. where a new fulfilment centre might help) use a map with a colour gradient.
- Return rate by fulfilment centre and by carrier, with trends and consistency.
Taming an overloaded page
The operations page ends up with too many visuals. Ask users which they want by default. If trend: show bar and trend charts, with a button beside each trend chart "Check performance consistency" that reveals the box plot. If consistency: show box plots with a button "Check performance trend over 6 months". Regularly used views are visible by default; occasional ones sit behind a button (progressive disclosure).
Key takeaways
- Classify the business question first, then pick the chart.
- Most TrendE metrics are comparison questions, answered by sorted bar charts plus trend lines.
- Consistency needs a box plot; geographic clusters need a map.
- Resolve overload via user priorities and buttons.
Designing the look
A dashboard is used when it meets requirements, but look and feel strongly affect appeal. Freeze look and feel upfront and get sign-off: changing colours or fonts after the build is time-consuming, and the choices are subjective, so consensus early keeps everyone sane.
Brand and device
Large-company dashboards must follow brand guidelines. Get the guidelines document, colour palettes and logos from the brand team (if no formal document, at least exact colours, approved logo files and font rules, not guesses from corporate communication). Capture this during requirements. In one project, asking upfront revealed the logo and colours were being redesigned, saving time. Consult a creative brand owner on first designs. Design must also match the device: phone designs differ from laptop designs.
Colour rules
- Purposeful colour: in a chart where every bar has a different colour, the viewer first has to wonder whether colour means anything. Highlighting one underperforming bar answers the question instantly.
- Palette of 3 to 4 colours in total, with a proportion of roughly:
| Share | Role |
|---|---|
| 60% | Neutral (backgrounds, text) |
| 30% | Secondary or brand colour |
| 10% | Accent (highlighting, e.g. the underperforming bar) |
- Brand colour vs status colour: if the brand red clashes with red for "missed target", shift the status colour slightly (e.g. red-orange) and reserve exact brand red for logos and primary actions, so a viewer never confuses brand with warning.
- Reserve discrete red / yellow / green for universally understood signals: below, on, or above target. For degrees rather than pass/fail, use a single-colour gradient from light to dark.
Accessibility
About 1 in 12 men have some red-green colour blindness; a smaller share have blue-yellow. Test rather than assume, e.g. with the free tool Coblis colour-blindness simulator. If colour alone marks the underperformer, the message is lost for some users. Pair colour with an icon, label or pattern; a test: if you could delete the colour and the chart still makes its point, it is accessible. Arrows and small icons help (a down arrow is clearer than a red dot).
Page design
A more professional page: a light canvas on a slightly darker neutral background (a shadow-box effect), consistent spacing between charts, and an invisible underlying grid so charts line up (modern tools such as Power BI make this easy). Include a title and logo styled from the brand guidelines so the page looks branded, not generic.
Wireframe vs mock-up
| Wireframe | Mock-up | |
|---|---|---|
| Question | Does the layout and navigation make sense? | Does the look and feel work? |
| Content | Rough boxes and sketches, grayscale, no specifics | Close to final: real charts, real palette, real logo |
| Data | None | Dummy data; dashboard not live |
| Stage | Before chart selection | Before building |
The dashboard evolves (needs and data availability change) but look, layout and navigation usually stay stable so users need not relearn the tool. This full process applies to important company-wide dashboards with senior sponsors; for a simple, short-term dashboard for one set of users, requirements must still be met but look and feel becomes secondary.
Key takeaways
- Freeze look and feel early with a mock-up sign-off; align with brand guidelines.
- 3 to 4 colours, 60/30/10, purposeful accent; separate brand from status colours.
- Never rely on colour alone; use icons, labels or patterns.
- Grid, spacing, title and logo create a professional page.
The Eight-Step Design Process (Recap)
| Step | Activity | Output and sign-off |
|---|---|---|
| 1 | Identify key stakeholders and user segments (sponsor, department stakeholders, pilot users) | Enrolled people |
| 2 | Gather and consolidate requirements (questioning and techniques) | Alignment on requirements and priorities across departments |
| 3 | Data audit and feasibility with IT and data owners (what is feasible, who owns it, in what format) | Sign-off on v1 scope: requirements specified by users and feasible |
| 4 | Phase-wise plan from user priorities and feasibility | Sponsor and stakeholder sign-off; reference plan for development |
| 5 | Information flow and hierarchy: number of pages, movement between pages, importance within a page | Rough wireframe (no charts) with sign-off from senior stakeholders and users |
| 6 | Chart finalisation: populate wireframe boxes by business question; plan filtering within and across pages | Charts per requirement |
| 7 | Look and feel: from brand guidelines and best practice | Near-final mock-up (colours, logo or images, charts, filters, no real data) signed off by sponsor, stakeholders and pilot users to manage expectations |
| 8 | Ongoing agile development: the dashboard is a living project; needs unfold with use, data arrives later | Test versions to pilot users each sprint, then launch to all; feedback mechanism to report data glitches and bugs |
Key takeaways
- The eight steps run from people and requirements through audit, phasing, structure, charts and look to ongoing agile delivery.
- Each design stage ends with a sign-off to manage expectations.
- A dashboard is never finished; it evolves in sprints with pilot approval before general release.
Revise before your examAll 4 exam tips from this module in one list
- Five prerequisites before design: user segments, stakeholders vs users, a senior sponsor, pilot users on your side, IT in the loop from day 1.
- A KPI card = number + label + comparison + trend signal. A number alone tells you almost nothing.
- Hierarchy = scannability within a screen. Flow = sequence across pages. Do not confuse them.
- Sign-offs occur at scope (step 3), phase plan (4), wireframe (5) and mock-up (7). Wireframe tests structure; mock-up tests look.
Test yourself on this module
Questions from these notes, one at a time, with feedback. Your weak topics are saved.