What belongs on a board dashboard: 7 metrics and 5 anti-patterns

At the board meeting a dashboard hangs on the screen. Eighteen charts, filters for every occasion, colours like a command centre. Five minutes in, the CEO still asks the finance director for numbers from his private spreadsheet. We know this picture from many companies, and the conclusion is similar every time: the problem was never the technology, it was what ended up on that dashboard.

A board dashboard is not a gallery of charts, it is a decision-making tool. Before anyone sketches the first mockup, one question needs answering: what decisions does the board take weekly and monthly, and which numbers do those decisions need. Everything else is secondary.

Seven metrics a board actually reads

We treat this set as a starting point, not dogma. In a holding with five companies it will look different than in a forty-person services firm. But in companies from a dozen or so to a few hundred people, these seven items cover most of the decisions made at board level.

1. Revenue against plan. Not sales alone, but the deviation from budget: year to date and in the current month. A hundred thousand below plan in March is a completely different conversation than the same gap in November.

2. Margin, not just revenue. In services: margin per project or client. In trade and manufacturing: gross margin per product group. Growing revenue on a shrinking margin is the most common way companies "grow themselves into bankruptcy", and without this metric nobody spots it in time.

3. Cash and a 13-week forecast. The account balance alone says little. Decisions rest on the forecast: what comes in, what goes out, where the trough falls. A quarter ahead, week by week.

4. Overdue receivables. Who owes, how much and since when, in buckets: 1–30, 31–60, over 60 days. This one tile can pay back the cost of the whole implementation, because conversations about overdue invoices start weeks before they turn into a liquidity problem.

5. A probability-weighted sales pipeline. Not "we have 8 million in offers out", but: how much of that will realistically land this quarter at your historical conversion rate. Without this number the board learns about a revenue hole when it is too late to react.

6. Team utilisation. In services: how much of the teams' time is used and how many weeks ahead sales has sold out the capacity. In manufacturing: line loading. This is the metric that connects sales decisions with hiring decisions, and yet it rarely reaches board level.

7. One operational metric for your industry. On-time delivery in logistics, churn in subscription services, occupancy in hotels, complaint rates in manufacturing. One, chosen deliberately. Not five.

In our view, a screen with these seven items and a simple green/red signal against plan beats an elaborate analytics platform that needs training to use. The reason is mundane: a board has a few minutes a week for the numbers, not half a day. A metric that cannot be read in ten seconds does not, in practice, exist.

Five anti-patterns that kill dashboards

1. Forty charts. The more metrics, the less each one weighs. When everything is important, nothing is. We have seen dashboards where you could find the answer to every question except one: "do we have a problem?".

2. Manually refreshed data. If someone pastes numbers into the feeding spreadsheet every week, that is not automation, it is Excel with a nicer font. Sooner or later someone skips a week, and the board makes a decision on three-week-old numbers without knowing it.

3. Vanity metrics. Page views, follower counts, total downloads. They grow almost always, they lift the mood, and they lead to no decision. The test is simple: if a 20% drop in the metric would trigger no action, the metric has not earned its place on the screen.

4. A metric with no owner. Margin drops three points and... nothing. Because no specific person is responsible for explaining why and preparing a response. Each of the seven numbers should have one name next to it. Without an owner, a metric is decoration.

5. A dashboard nobody opens. The best KPI set dies without a ritual. If the weekly board meeting does not open with the same seven numbers, within a quarter everyone drifts back to private spreadsheets. A tool does not replace a habit; it only makes the habit easier.

And here an honest caveat: a list of metrics will not fix reporting if the source data is poor. A dashboard mercilessly exposes the mess in your systems, so the first weeks after launch can be uncomfortable. That is actually a benefit, though not everyone sees it that way at the start.

Where we start an implementation

At ESKOM AI we start building a dashboard from the list of board decisions, not from the list of available reports, and only then match metrics and data sources to it. We build software in a process supported by a team of specialised AI agents, with a full range of tests: unit, integration, E2E, UI, security and performance, which is why you typically see the first working version with automated data feeds in one to two weeks. We covered the flow and economics of such a project in a separate piece on the CFO dashboard. And if you want to try the seven-metric list on for size in your own company — write to us via the form at eskom.ai/pl/kontakt and we will check, free of charge, which of them can be fed from your current systems.