Dashboards Owners Actually Use

The short answer

Most dashboards get opened twice: once at handover and once when someone asks whether anybody uses the dashboard. They fail because they answer questions nobody asked, at a frequency nobody needs, in a form that requires interpretation.

A dashboard an owner actually uses has three properties: it answers a question they already ask out loud, it tells them when something is wrong rather than waiting to be visited, and it fits on one screen.

Start from the question, not the data

The usual process is backwards. Someone looks at what the database contains and builds charts for it, which produces a page of things that are true and useless.

Instead, write down the questions the owner asks their staff every week. "Are we ahead of last month?" "What is sitting unpaid?" "Which jobs are late?" "Did anything break yesterday?" Those are the dashboard. If a chart does not answer one of them, it is decoration.

Four or five numbers is usually the whole thing. Anyone asking for twenty is describing a report, not a dashboard.

A number alone means nothing

"Revenue this month: NPR 840,000" tells you nothing without a comparison. Against last month, against the same month last year, against target. The comparison is what turns a figure into a decision.

The same applies to anything counted. Forty-three open orders is good or terrible depending on whether it was twelve last week. Always ship the context next to the number, because the owner will otherwise supply it from memory, and memory is generous.

Push, do not wait to be visited

The dashboard nobody opens is still useful if it sends something. A short weekly email with five numbers and the direction each moved will be read by people who have never once loaded the page.

Better still, alert on the exception. Nobody needs a daily message saying everything is fine. They need one message the day the payment queue stopped draining, or the day an order has been sitting unacknowledged for forty-eight hours. We built exactly that logic into this site's deployment checks for the same reason: a check that only speaks up when something is wrong gets trusted, and a check that always speaks gets ignored.

Things that quietly kill a dashboard

• It is slow. Eight seconds to load and nobody comes back. Precompute rather than running heavy queries on every view.
• It disagrees with the accountant. Once the numbers differ from the official ones, trust is gone permanently. Agree the definitions first: does revenue include VAT, is it booked on order or on payment.
• Dates are ambiguous. If the business thinks in Bikram Sambat and the dashboard reports Gregorian months, every figure needs translating before it can be used. See handling BS dates.
• It needs explaining. If a chart requires a sentence of interpretation, replace it with the sentence.

Frequently asked questions

What should go on a business dashboard?

The four or five questions the owner already asks aloud each week, each with a comparison. Everything else belongs in a report that is run when needed.

How often should it update?

Match it to how often someone could act. Daily is right for most small businesses. Real-time looks impressive and changes no decisions, while costing considerably more to build.

Should it be a dashboard or a report?

Dashboard for the handful of numbers watched continuously. Report for detail someone occasionally needs. Trying to make one thing do both produces something that does neither.

Why does nobody use ours?

Usually because it answers questions nobody asked, or because it once disagreed with the accounts and lost its credibility. Ask the owner what they check manually instead, and build that.

Can it send a weekly summary by email?

Yes, and it is often the highest-return part of the whole thing. Five numbers and their direction, every Monday, read on a phone.