How to Design a Management Dashboard That Supports Decisions

A management dashboard is the screen a leadership team uses to decide where money, people and attention go next: a handful of measures against target, the exceptions that need a decision, who owns each one, and what was agreed about them last time. If it doesn’t change a decision, it isn’t a management dashboard. It’s a very expensive screensaver.

We have been designing and fixing dashboards for 15 years, and the management ones fail in the most expensive room in the company. Monday, leadership review. Someone shares a screen with thirty-odd tiles. The CFO’s revenue doesn’t match the sales number. Twenty minutes go on whose figure is right, ten on a chart nobody can explain, and the one thing that genuinely needed a decision, the region quietly missing its pipeline, gets “let’s take it offline”. Offline is where decisions go to die.

Now count what that hour costs. Eight of your most expensive people, every week. A data team that spent a quarter wiring sources into a BI tool you pay for per seat. And at the end of it, the decision gets made exactly the way it was made before the dashboard existed: by whoever argues loudest. More heat than light, now with a licence fee.

The fix is not a nicer chart library. It is the order of work. Decisions first, then the measures that inform them, then comparisons, exceptions, actions, data states and permissions. Charts come last, and there are fewer of them than anyone expected. If you need the broader definition first, start with our guide to what a website dashboard is and which type fits which decision. This one is about the screen your leadership team runs the business from.

What a management dashboard is, and who it is for

A management dashboard pulls data from across the business into a short answer to one recurring question: are we on course, and where do we need to step in? It serves people whose job is allocation, of budget, headcount, time and attention, rather than carrying out one task.

That separates it from three things it keeps being confused with:

• An operational dashboard tells a team what needs doing right now: tickets about to breach, orders to ship, shifts to cover. It refreshes by the minute and gets used many times a day. A CEO does not need it, and a CEO who watches it starts micromanaging the support queue.
• An analytical workbench lets an analyst slice data to find out why something moved. Brilliant for the analyst. Handed to a leadership team, it is analysis paralysis with a dropdown.
• A report is a record of what happened, frozen at a date. Useful for the board pack and the auditors. It doesn’t tell anybody what to do on Monday.

The people who use a management dashboard are fewer than you’d think, and more different from each other than they look:

• The founder or CEO wants to know where the company is off course and who is dealing with it.
• The leadership team, finance, sales, operations, product, people, each owns a slice and needs to see it against the whole.
• Department heads need their own numbers with enough company context to argue for resources.
• The board or investors want the trajectory, the risks and evidence that management has a grip on them, usually once a quarter.

Four audiences, four depths, four rhythms. The classic mistake is one screen for all of them. It ends up too shallow for the department head and too noisy for the board, and the CEO gets a screen designed by committee for nobody in particular.

Here’s a test to run before anyone opens a design tool. Could this screen replace the first fifteen minutes of your leadership meeting? If yes, it is a management dashboard. If people still turn up with their own spreadsheets “just in case”, it is a picture of one.

Start with the decisions, not with the data you happen to have

Most management dashboards are built from the inside out. The data team lists what is already in the warehouse, a BI consultant turns each field into a tile, and someone calls the result “a single source of truth”. It is cooking dinner from whatever is left in the cupboard: technically food, and nobody ordered it.

Work from the other end. Sit with each person who will use the screen and list the decisions they make on a cycle. Not the metrics they like. The decisions. Do we hire into the Q1 plan or freeze? Which region gets the extra sales budget? Do we step into this renewal ourselves? Do we keep quoting fixed-price projects? For each one, write down five things:

• The question in plain words, the way the person would say it out loud.
• The owner, the one person who makes the call.
• The cadence: how often the decision comes up.
• The trigger: the threshold or change that means “act now”.
• The action: what the owner can actually do about it.

This decision inventory is the brief. Everything on the dashboard must trace back to a line in it. A measure that informs no decision is decoration, however expensive the pipeline behind it.

The inventory also drags the uncomfortable stuff into daylight early. Decisions with no owner. Decisions that turn out to be made in a corridor without anyone looking at a number. Two leaders who each believe they own pricing. Far better to find that in a workshop than in a meeting where the dashboard gets blamed for a management problem.

And it stops the money leak before it starts. Every tile costs a data pipeline, an agreed definition, a place on the screen and somebody to maintain it for years. Build sixty because the data was available, and you are paying upkeep on fifty-odd numbers nobody decides with, plus the attention tax of making every leader scan past them every single week.

Match the dashboard to the meeting and its rhythm

Management runs on a calendar. There’s the weekly leadership review, the monthly business review after the books close, the quarterly board meeting and the annual plan. Each one has its own questions and its own tolerance for data that is a day or a week old. A management dashboard that ignores the calendar gets the worst of all of them.

Real time is the usual trap. Vendors love it, and it looks fantastic in a demo: numbers ticking up while the sales rep smiles. For management it is mostly noise. Weighing yourself every hour doesn’t make you any thinner. It makes you anxious and slightly unhinged about lunch. A CEO refreshing today’s revenue at 11:40 isn’t managing, they’re fidgeting, and the team gets pinged about a dip that turns out to be Tuesday.

A watched KPI never boils. It does eat the afternoon.

Design for the rhythm instead:

• The weekly leadership review asks what moved against plan since last week, what crossed a threshold and what needs a decision today. Data from yesterday is perfectly fine.
• The monthly business review looks at full-month results against budget and forecast once the close is done, by department and region, with the owners’ commentary.
• The quarterly board view shows the trajectory against the annual plan, the biggest risks and what management is doing about each one. Fewer numbers, more context.
• Daily belongs to operations. If the leadership team needs a daily number, it’s usually because something is on fire, and that deserves its own temporary view with an end date on it.

The single most useful decision we make on management dashboard projects is treating the weekly view as the meeting agenda. The screen opens on what needs a decision, in the order the meeting should take it. Nobody prepares a deck. Nobody argues about whose version is current. The meeting starts where the screen starts, and the hours leadership used to spend building slides go back into running the business.

Which KPIs belong at management level

There is no universal list of management dashboard KPIs, and anyone who sells you one is selling a template. There is a reliable way to decide which measures earn a place, though: they describe outcomes the leadership team is accountable for, or leading signals that predict those outcomes early enough to act on.

Revenue is the rain gauge. It tells you, precisely and too late, that it rained. Pipeline, renewals at risk and projects slipping are the barometer. A management dashboard that only shows the rain gauge is a very accurate record of weather you can no longer do anything about.

What gets measured gets managed, goes the old line. What gets measured badly gets managed badly, with great confidence.

Different areas of the business need different kinds of measure at management level, and each has a natural home one level down:

Notice what the table doesn’t contain: sessions, followers, tickets closed, story points, meetings held. Those are real numbers that someone should watch. That someone is not the leadership team. Push them up to the management view and you get a screen where the marketing head’s favourite vanity chart sits at the same size as cash runway, and gets the same airtime.

The areas also differ in shape, and the design has to respect that:

• Strategic and financial measures move slowly and are judged against a plan. They need targets, forecasts and a clear warning when the month isn’t closed yet.
• Sales and marketing measures are about the future. They need coverage and conversion, not just totals, and the period printed on the tile, because “pipeline” without a date range is a wish list.
• Operational measures are about exceptions. Most projects are fine; the value is in the three that aren’t.
• Team measures carry the most sensitive data in the company and need the strictest aggregation, which we come back to under permissions.

Draw the tree before the screen. At the top go the few outcomes the business is actually run on. Below each, the two or three drivers that explain it. Below those, the department measures that explain the drivers. The management view shows the top two levels. The third sits one click away and belongs to the department. That’s how you keep the screen small without hiding anything.

Management dashboard examples, by the decision they serve

The same company usually needs more than one management view, because the decisions and the rhythms differ. Here are the five we design most often:

• The CEO’s weekly leadership view. Question: where are we off plan, and who is on it? Five to nine headline measures against target, exceptions first, the owners’ notes and the decisions due this week. It opens as the Monday agenda.
• The CFO’s monthly close view. Question: did we land the month, and if not, why not? Results against budget and forecast by cost centre and region, marked provisional until the close, with variances explained by the people who own them.
• The sales leadership pipeline review. Question: will next quarter’s number be there? Coverage by region and segment for a stated period, conversion by stage, deals whose close dates keep slipping, and each regional head’s call.
• The operations leadership view. Question: are we delivering what we sold, and where are we about to stop? On-time delivery against target, projects at risk with their owners, capacity against booked demand for the next eight weeks, and the handful of customers who will notice first.
• The board’s quarterly view. Question: is management in control? Trajectory against the annual plan, the top risks with owners and mitigations, and cash. Fewer numbers than any other view, and more words.

Five examples, five different screens. They share definitions, data sources and a design system. They should not share a layout. The moment one grid tries to serve all five, you’re back to thirty tiles and a meeting that argues about them.

Fewer measures, and a name on every one

Show a leadership team twenty-four tiles of equal size and you’ve recreated Buridan’s donkey: the animal placed exactly between two identical haystacks that starves because it can’t choose. Twenty-four haystacks, same size, same colour, same font. Leaders don’t starve. They do something worse. They read the first row, nod, and move on to the agenda they brought with them.

Our working recommendation for the headline level is five to nine measures. That’s a recommendation, not a law of physics. The right number is the number of outcomes your leadership team is genuinely accountable for. If you can’t get below fifteen, the problem isn’t the design. The company hasn’t agreed what it is optimising for, and no dashboard will settle that argument for you.

Every headline measure gets an owner: one named person who explains it when it moves and proposes what to do. Not a department, not “leadership”, not “the data team”. The buck has to stop at a desk with a name on it, or the red tile becomes everybody’s favourite topic and nobody’s job.

The owner’s name on the tile changes the meeting more than any chart choice ever will.

A red number with an owner is a question. A red number without one is a discussion, and discussions are free to start and expensive to sit through.

Then comes the question every executive asks in the first demo: can I customise it? For the headline level, our answer is no, and it annoys people exactly once. The leadership view is a shared agreement about what the company is run on. If the CEO hides churn, the CFO adds a margin chart and the head of sales resizes pipeline until it looks healthier, you no longer have one screen in the meeting. You have five personal versions of the truth and an argument about which one is on the TV. Keep the headline level fixed and owned. Below it, let people save their own filtered views, pin the regions they care about and build the analysis they like. Personal freedom one level down; one version of the company at the top.

Every number needs a comparison

A child comes home with 72 on a test. Good news? Depends. Out of 80, or out of 200? Was it 90 last term? What did the rest of the class get? A number without a comparison isn’t information. It’s a number.

Management dashboards are full of 72s. “Revenue: $1.6M.” Up or down? Against what? Is that the whole month or the first eleven days? Every headline measure needs comparisons, and the design has to decide which one leads:

• Against target or plan. This is the comparison leadership is accountable for. Where it exists, it goes first.
• Against the previous period. Shows direction, but gets fooled by seasonality every single year.
• Against the same period last year. Takes seasonality out and stops the January panic about December.
• Against forecast, where there is one. Often the most honest view of where the quarter is actually going to land.

Then come the traps that make comparisons lie:

• Partial periods. Month-to-date revenue next to last month’s full total shows a collapse every morning of the first week. Compare like with like, the first eleven days against the first eleven days, or label it so nobody can miss it.
• Percentages without the base. “Churn up 50%” is a crisis if it went from 4% to 6% of revenue, and nothing at all if it went from two small customers to three. Show both.
• Status colours with no rule behind them. If nobody wrote down when amber turns red, the colour is somebody’s mood on the day the tile was configured. Agree thresholds per measure with its owner and show them on the tile.
• Scales chosen for drama. A chart that starts at a convenient number turns a small wobble into a cliff. A leadership team should never have to read an axis to know whether to worry.

Numbers don’t lie, people like to say. No. But people choose the date range.

Show the exceptions first

A good sheepdog doesn’t count sheep. It watches the flock and goes after the one heading for the road. Management by exception works the same way, and it’s the most useful principle a management dashboard can borrow. When everything is within tolerance, the screen should say so in one line and get out of the way. When something isn’t, that thing goes to the top, with its owner and how long it has been out of line.

Most dashboards do the opposite. They show every measure in a fixed grid, in the order someone configured two years ago, and leave leaders to spot the problem themselves. The squeaky wheel gets the grease, usually the metric that belongs to the loudest person in the room, and the quiet wheel that’s actually coming off gets noticed at quarter-end, by which point it’s a write-off.

An exceptions area needs more design than it looks like it does:

• A rule per measure. A threshold against target, a sharp move against trend, a forecast miss. Written down, agreed with the owner, visible on the tile.
• Age. “Below target for three weeks” is a different conversation from “dipped yesterday”.
• Owner and status. Acknowledged, plan requested, plan in place, accepted as is. An exception the leadership team has consciously accepted is no longer an exception, and it should stop shouting.
• A calm all clear. “Everything else within tolerance” is real information. It’s also the line that gets a CEO to trust the screen enough to stop asking for the spreadsheet.

Be careful with alerts. An exception that emails eight people every time a number wobbles trains all eight to route those emails into a folder they never open. Alert the owner when their measure crosses the line. Send leadership the summary, on the meeting’s rhythm.

From signal to action: explanation, owner, decision

A management dashboard that stops at the red number has done the easy half. Picture a credit card statement that shows only the total. You know you spent too much. You have no idea on what, and no way to dispute the charge you don’t recognise. That’s a dashboard with no path from signal to action.

Every exception needs three steps behind it, each one click away:

1. Why it moved. The breakdown by the drivers from your KPI tree, by region, product line, segment or a handful of large accounts, so the explanation starts from data rather than from somebody’s memory.
2. Who is on it and what they propose. The owner’s note, a requested plan and its due date. Not a forty-message comment thread. A status and two sentences.
3. What was decided. Accept, intervene, reallocate, or revisit on a date. Written down with the owner and the review date, where next week’s meeting will find it.

That last step, the decision log, is the part almost nobody designs, and it’s the part that turns a status display into a management tool. Without it, the same red tile gets discussed from scratch every Monday, because nobody remembers what was agreed three weeks ago. With it, the question changes from “why is margin down?” to “we paused large fixed-price quotes on the 5th, is it working?” That’s a leadership team learning from its own decisions instead of rediscovering them every week.

Keep the actions light. A management dashboard is not a project management tool, and trying to turn it into one ends with a second backlog nobody grooms. The actions it needs are few: ask the owner for a plan, add the item to the next review, accept with a note, open the records underneath. Anything heavier links out to the system where the work really happens. For the general mechanics of drill-downs and filters that keep their context, our dashboard design process guide goes further. Here, the point is simpler: the path has to end with a decision, not with a more detailed chart.

Context that prevents misreading

Two builders measure the same wall with two tape measures, and one of the tapes has been stretched. They’ll argue about that wall for an hour. There’s nothing wrong with the wall.

This is the most common failure we find in management dashboards, and you can’t see it in a screenshot: the same word means different things to different people. Finance’s revenue is recognised and net of refunds. Sales’ revenue is whatever was booked this month. Customer success counts an expansion when it’s signed; finance counts it when it’s invoiced. Put two of those on one screen under the same label and the leadership meeting turns into a debating society.

Context has to be designed in, not added as a footnote:

• A definition behind every measure. Formula, source system, what’s excluded, who owns it, when it last changed. One click from the tile, in plain language, not SQL.
• Annotations on the timeline. The price change in March, the outage in June, the one-off enterprise deal that makes Q2 look like a miracle. Without them, every new reader rediscovers the spike and asks the same question in the same meeting.
• The owner’s commentary. Two sentences from the person who owns the number, written before the meeting: what happened, and what they propose. The number tells you what. The owner tells you why and what next.
• Units, currency and period on the tile itself. “$18.4M ARR, end of Q3” is a fact. “18.4” is a riddle.

Every misread number costs twice. Once in the meeting that argues about it, and again in the decision made on the wrong reading: the hire you froze because a partial month looked like a collapse, the region you cut because two teams counted the same deals differently.

Data states: provisional, late, restated and missing

A lab result with no date on it could be from this morning or from your check-up in 2019. You wouldn’t let a doctor act on it. Leadership teams act on undated numbers every week.

Management data has states that operational dashboards rarely deal with, and they belong to the core design, not to an edge-case list for later:

• Provisional. The month isn’t closed and the numbers will move. Say so, give the expected close date, and keep provisional figures out of hard comparisons with closed ones.
• Late source. One system hasn’t synced. Show which one, since when, and what the total is missing, rather than a total that silently drops a region.
• Restated. Last quarter changed after an adjustment. Show the old figure, the new one and the reason, or the board pack and the dashboard will disagree in front of the board.
• No target yet. New measure, new market, a plan nobody has signed off. Say that. Don’t draw a target line at zero.
• Not enough history. A product line launched in spring can’t have “same period last year”. Hide the comparison and explain why, instead of showing a heroic but meaningless percentage.
• Restricted. The viewer isn’t allowed to see part of the data. More on that below; the short rule is that a total must say what it excludes.
• Load failed. Show when the last good data is from and let people use it, clearly labelled, instead of a blank screen five minutes before the board meeting.

And the empty state, which management dashboards have too: the first weeks after launch, a young company with no history, a newly formed department. The general patterns for empty, loading and error states are covered in the process guide mentioned above. For a management dashboard the rule is blunt.

Permissions: who sees which numbers

Parents’ evening works because each parent sees their own child’s report, the head teacher sees the whole school, and nobody reads out the entire class’s grades in the hall. A management dashboard needs the same arrangement, and most get it wrong in one of two directions: everybody sees everything, or everybody sees a slightly different total and nobody knows why.

Management data is the most sensitive data in the company: pay, individual performance, unannounced financials, deal terms, sometimes restructuring plans. Permissions need designing per role, per level of detail and per measure:

• By scope. A regional head sees their region in full and other regions as totals, or not at all.
• By depth. The board sees company-level trajectory. The CFO sees cost centres. Nobody outside the people team sees individual salaries.
• By aggregation. Team measures for groups below a minimum size are suppressed, because the “average salary” of a team of two is two people’s salaries.
• By timing. Financials before results are announced may need a narrower audience than the same numbers a week later.

The design rule that saves the most arguments: when a person sees a filtered total because of their permissions, the screen says so. “Total for UK & Ireland and DACH; 1 region not shown.” Otherwise two leaders walk into the same meeting with two different company totals, both correct, and the first twenty minutes go on discovering that they’re looking at different screens.

Built for the phone, the meeting room and every reader

Management dashboards get read in three places designers rarely test: on a phone in a taxi before the meeting, on a meeting-room screen from three metres away, and through assistive technology. Design it like a motorway sign: readable at speed, from a distance, by someone who is thinking about something else.

• On the phone, don’t shrink the desktop grid. Give a brief: what needs a decision, the headline measures with their status in words, and each owner’s line. The detail can wait for a laptop.
• In the meeting room, use big numbers, few of them, and labels that survive a projector. A screen tested only on a large desktop monitor turns into grey soup on the TV at the end of the table.
• For accessibility, never let colour carry status on its own. Red and green are the pair people with colour vision deficiency most often confuse, so every status colour needs a word or an icon beside it: “Below target”, not a red dot. Keep text contrast readable, make every tile reachable by keyboard with a meaningful label, and give each chart a text summary a screen reader can speak: “Net revenue retention 104%, target 108%, below target for three weeks.” That sentence, incidentally, is the most useful line on the screen for everyone else too.

Validate it before launch, and keep proving it after

Play a pre-season friendly before the league starts. Before anything is built, take the prototype, filled with real numbers from last quarter rather than lorem ipsum and random charts, and replay real leadership meetings against it. Which decisions came up? Would the screen have surfaced them earlier? Which questions did people still need a spreadsheet to answer?

Then test with the actual users on actual tasks, one person at a time, and time it:

• Which area needs a decision this week, and who owns it?
• Is gross margin below target because of price, product mix or delivery costs?
• What did we decide about it last time, and is it working?
• What’s missing from this total, and why?

If a leader can’t answer in a minute or two, the problem is on the screen, not in their head.

After launch, measure usefulness, not traffic. Page views on a management dashboard mean very little. A CEO opening it forty times a day is a sign of anxiety, not value. Look at these instead:

• Meeting use. The share of leadership agenda items that start from the dashboard, and whether people still bring their own decks.
• Decisions recorded. How many exceptions end in a logged decision, and how many get discussed twice without one.
• Time to acknowledge. How long an exception sits before its owner responds.
• Definition disputes. How often the meeting argues about what a number means. This should trend towards zero.
• Dead tiles. Measures nobody opened, filtered or discussed in a quarter.

Then prune, every quarter. Every management dashboard grows a sacred cow: the tile a senior person asked for two years ago that nobody has looked at since. Take it out and see who notices. If someone does, you’ve learned something. If nobody does, you’ve learned more.

If your dashboard is already live and quietly ignored, this is exactly what a UX audit of the existing dashboard looks for: the decisions it misses, the numbers people misread, and the tiles that cost money to maintain and change nothing.

The decision rule

Before anyone draws a chart, the team should be able to answer these questions in order. Stop at the first one without a clear answer:

1. Which decisions does this screen support, for which roles, on what cadence?
2. Which measures inform each decision, and who owns each one?
3. What is each measure compared with, and which threshold means “act now”?
4. What does the path from an exception to a recorded decision look like?
5. Which states can each number be in, provisional, late, restated or missing, and how does the screen say so?
6. Who can see what, and how does a filtered total announce itself?
7. How will you know, three months after launch, that decisions got faster or better?

If any answer is “we’ll see”, don’t design the screen yet. Charts are the last decision, not the first. A dashboard that shows more data is easy to build and easy to ignore. A dashboard that changes a decision is harder, smaller, and the only kind worth paying for. That is the work we do in management dashboard design: decisions, owners, measures, states and permissions first, then the screens.

Define the decisions first. Then the measures, dimensions, comparisons, permissions, data states, and the action each view makes possible. Everything you leave undecided, your leadership team decides by gut in the meeting, which is exactly what they were doing before you paid for the dashboard.