Custom reporting and dashboard development means the numbers assemble themselves, from the systems that already hold them, on the schedule you need. The value is not the chart. It is that the figure is current, defined the same way every time, and traceable back to the records it came from.
First question
Trust, and only trust. A dashboard is used when people believe the numbers, and they believe the numbers when they can click one and see where it came from. Presentation is a distant second.
The real work
Most of the difficulty in reporting projects is not technical. It is that two departments have been quietly using the same word for different things.
Invoiced or collected? Before or after refunds? Which date does it belong to — order, delivery, or payment? Three teams, three answers, all defensible.
Bought in the last month, the last year, or has an open contract? The number moves by a factor of several depending on which one you pick.
Work not finished, or work not invoiced? The gap between those two is often where the argument in the Monday meeting actually lives.
So the deliverable includes a written definition for every metric. Unglamorous, and the reason the dashboard survives its first disagreement.
Questions
If your data is already tidy and in one place, use them — they are excellent and cheaper than a custom build. Custom reporting earns its place when the numbers live in three systems that disagree, when the calculation itself is unusual enough that a generic tool cannot express it, or when the report needs to be embedded in software your staff or customers already use.
Yes, but the fix is a decision before it is a build. When two systems report different revenue, someone has to define which is authoritative and why the other differs — usually timing, or a different treatment of refunds and adjustments. We surface those differences explicitly rather than averaging them away, because a reconciled number nobody understands is worse than two numbers you can explain.
As current as the slowest source allows. Data you own can be near real-time; a system that only exports nightly caps everything downstream at a day old. We show the freshness of each figure on the dashboard itself, so nobody makes a decision on a number they assumed was live and wasn't.
They should be able to, and it is the feature that decides whether a dashboard gets trusted. The first thing anyone does with a surprising figure is try to work out where it came from. If they cannot, they stop believing the dashboard and go back to their spreadsheet — so every headline number should open into the records behind it.
The current version, however messy. It tells us more than a requirements document would.