Ten filters is a design flaw

You open a report and at the top there is a row of filters. Period, branch, department, product group, supplier, sales rep, customer type, channel. Sometimes a second row. The report itself only starts halfway down the screen.

It looks flexible. You can look at anything. In practice nobody looks at anything, because whoever opens the report first has to know which buttons to set to find his answer.

A filter is a question you pass on

Every filter on a page really says: we did not know which selection you needed, so pick it yourself. Once or twice that is reasonable. Ten times means the page does not know which question it answers.

The user then carries the work the designer should have done. And he does it worse, because he does not know the model. He does not know that branch and department affect each other, or that a filter on supplier makes half the revenue disappear because not every item has a supplier.

Filtering hides the comparison

The biggest problem with filters is not the hassle. It is that with a filter you only see one thing at a time.

If you want to know how store A is doing against store B, you set the filter to A, remember the number, set it to B, and compare in your head. With three stores that still works. With eight it does not.

Show the eight stores side by side in one chart and you see at a glance which one stands out. That is not a detail. It is exactly what a dashboard is for. The store with the best margin is rarely your best store, but you only see that when they are side by side.

Visualise more, filter less

For every filter you want to add, you can ask: can I not just show the breakdown?

• Instead of a filter on branch: a bar chart with all branches, sorted.
• Instead of a filter on month: a line across the months, with last year next to it.
• Instead of a filter on product group: a small table with the groups as rows.

That takes more thought in the design. You have to choose which breakdown matters most. But someone has to make that choice anyway, and the designer does it better than the user who wants to know something quickly on a Monday morning.

Which filters can stay

Filters are not bad. They just belong in the right place.

Period can almost always stay, with a sensible default. The last closed week, or the current month. That way the report shows something meaningful without anyone touching anything.

Who you are does not need to be a filter. A branch manager who may only see his own store gets that through access rules, not through a slicer he can forget. See who sees what in your report.

For detail you can use filters on a separate page. Whoever wants to zoom in on one supplier clicks through to a page where that is possible. The main page stays calm.

A good rule of thumb: two, at most three filters on a page where decisions are made. If you need more, it is really two pages.

The hidden problem: filters that stick

There is another reason to be sparing. In Power BI, filters are saved. Someone sets the filter to one branch last week, closes the report, and this week still sees one branch without noticing. He thinks revenue has dropped.

That is the most dangerous kind of error: a correct number read the wrong way. The fewer filters, the smaller that chance. And show visibly on every page which selection is active, in the title or in a small box. Then the reader always knows what he is looking at.

What you can automate

You can make the default selection move with the calendar, with a measure or a relative date filter, so nobody has to change the period every Monday. A button that resets all filters to the default takes five minutes to build and prevents a lot of confusion. And the visible summary of active filters is one measure that returns the selection as text.

The claim

A row of filters is a question the designer passes on to the user. Show the comparison that matters straight away, keep filters for period and detail, and never let a page ask for more than two or three buttons before it says anything.