Most companies that decide to "get Power BI" already own a dashboard nobody opens. The licence was the easy part. Someone built a few pages, the numbers were close but not quite right, and within a quarter people went back to exporting to Excel.
That is not a tool failure. It is what happens when a dashboard is treated as a drawing job instead of a data job. Here is what a dashboard expert actually changes, how to tell whether you need one, and the cases where you do not.
A dashboard expert decides which question each dashboard answers, builds the data layer underneath it, gets the metric definitions agreed, and hands over something your team can run. The charts are the smallest part of the work.
The Tool Is Not the Problem
Power BI, Tableau and Looker will all draw you a chart in an afternoon. Anyone can drag a column onto a canvas and get a bar graph, which is exactly why dashboards look finished long before they are correct.
What the tool cannot do is decide that "revenue" means invoiced revenue and not booked revenue. It cannot notice that two source systems record the same customer under different names. It cannot tell you that the figure on the executive page took twenty seconds to load and nobody waited.
Those are the problems that decide whether a dashboard is used. They live in the data and in the definitions, below the part everyone sees.
What a Dashboard Expert Actually Does
Starts from the decision, not the data
The first question is not "what do you have?" but "what will somebody do differently after reading this?" A dashboard with no decision behind it becomes wallpaper.
This sounds small and it removes more wasted work than anything else. Half the requested metrics usually drop out once someone has to name the action they support.
Builds the data layer properly
Most of the effort is invisible: pulling from the ERP or database, cleaning it, joining sources, and shaping it into a model that stays fast. This is where projects are won or lost, and it is the part a quote often leaves out.
A well-shaped model, usually a star schema, also makes every later report cheaper to build and quicker to open. A poorly shaped one is why dashboards slow down as data grows. We cover the symptoms in why your Power BI dashboard is slow.
Gets the definitions agreed
When finance and sales report different revenue, the argument is about meaning, not mathematics. An expert runs that conversation, writes the answer down, and builds it into the model once, so the number is the same everywhere.
If this is already your problem, why Power BI disagrees with your Excel is a good place to start.
Makes it reliable after launch
Refreshes fail. A source changes a column name. A licence lapses. The difference is whether you find out from monitoring or from an angry director. Refresh alerts and a documented fix path are part of the job; see what to do when a refresh fails and nobody notices.
Hands over something you own
Documentation, the transformation logic written down, and training for a named person on your side. Without that you have not bought a dashboard; you have bought a dependency on whoever built it.
Seven Signs You Need One
- Two dashboards disagree on the same number. Usually a definitions problem, and it erodes trust faster than any bug.
- Reports take a long time to open. Often a modelling problem rather than a hardware one.
- Month-end takes days of manual work. A good analyst is spending their week on copy, paste and reconciliation.
- Data arrives by manual export. Someone downloads a file from the ERP and pastes it in. That is a process waiting to break.
- Refreshes fail and nobody notices. The dashboard quietly shows last week's numbers.
- One person understands the whole thing. If they leave, so does your reporting.
- People export to Excel anyway. The clearest sign the dashboard does not answer the question they have.
One or two of these is normal. Three or more means the problem is structural, and more chart-building will not fix it.
What It Costs You Not To
The cost of a poor dashboard rarely appears as a line item, which is why it goes unchallenged.
- Analyst time. Skilled people spend days on reconciliation instead of analysis.
- Decisions on the wrong number. If leadership does not trust the figures, they decide on instinct, or they decide on a figure that is quietly wrong.
- Key-person risk. One departure can leave you with a model nobody understands.
- Rework. Dashboards built without a data layer get rebuilt, often more than once.
Set that against a project, and the comparison is usually less lopsided than it first looks. For realistic figures, read what a Power BI project actually costs.
When You Do Not Need One
An honest answer, because it is not every company.
- You have one clean data source and the definitions are not in dispute.
- You have someone internal with time to learn the tool properly, not as a side task.
- You need a single, simple report, and the cost of it being slightly wrong is low.
In that case, build it yourselves, then pay for a short review once it exists. A Power BI audit checklist gets you a long way for free. An expert becomes worth the money when data comes from several systems, when numbers are contested, or when the report is one that leadership actually acts on.
Expert, Freelancer, Agency or In-House?
| Option | Works best when | Watch for |
|---|---|---|
| Independent expert | Defined project, direct access to the person doing the work | Availability; make sure there is documentation |
| Agency | Large, multi-team programmes | Layers of people between you and the practitioner |
| Hire in-house | Continuous BI work for years | Hiring time; a single person becoming a single point of failure |
| Learn it yourselves | One clean source, simple needs | The first model is rarely the one you keep |
A common sensible pattern: bring in an expert to build the first model properly, then have an internal hire inherit something well structured instead of starting from nothing.
How to Choose One
Ask five things, and be wary if the answers are vague.
- Can I see a data model, not only screenshots? Screenshots show the surface. The model shows the thinking.
- What happens when a refresh fails? You want a specific answer, not "we'll look into it".
- Who owns the definitions? If the answer is "we decide", expect a dispute later.
- What does handover include? Documentation and training, not just workspace access.
- Is the price fixed against a written scope? Open-ended day rates with no defined outcome shift the risk to you.
Quick Answers
What does a dashboard expert do?
Decides which question each dashboard answers, builds the data layer, agrees metric definitions, and makes the result fast, reliable and documented.
Is that the same as a Power BI developer?
Not always. A developer builds what is specified. An expert also challenges the specification and designs the model so numbers stay correct after the first month.
Can my own team do this?
Yes, with one clean source, agreed definitions and time to learn properly. Trouble starts with several systems, contested numbers or a single person who understands the model.
How long does it take?
A single dashboard on one clean source typically takes three to four weeks. Three or four systems usually take six to eight, because reconciling the sources is most of the work.
How do I choose a consultant?
Ask to see a data model, how they handle failed refreshes, who owns definitions, what the handover includes, and whether the price is fixed against a written scope.
If three or more of the signs above sound familiar, tell us how many source systems you have and which numbers people argue about. That is enough for a realistic view before anyone writes a proposal. See how I work and what it costs, or describe your reporting problem.