Learn More...

Our sample PBIX file illustrates a number of real-world use cases where Autocomplete makes a difference. If you'd like to see more concrete examples, send us an email and there's a good chance we'll add your scenario to our sample!

New in version 6.0: per-term operators, match reasons, a paste report, Also Search, quick picks, Copy as DAX and more. See New in Version 6.0, further down this page.

Definitions

This diagram labels (in red) different data entry mechanisms that the visual supports:

Select Single or Multiple? Filter Styles.

This fundamental setting controls whether the visual allows entry of one filter term, or multiple.

Single selection only needs to show a single text box, where multiple entry offers two modes of entry: the search text box (aka Entry Text Box) and the accumulated filter term area (aka Filter List Area). The accumulated filter area can either be a text area, or a list of values which resemble rounded "pills".

In version 3.0, a new entry style was added: "Text Pop-up". This hides the filter list area until explicitly invoked. The button used to invoke it will show a red circle glyph when there is at least one filter term in place.

It's possible to let the user toggle the entry style between Text and Pills:

Filtering Behavior and Entry

At the most fundamental level, you can control whether you're filtering to match values or exclude values.

You can also control whether a) the visual only starts to filter when all data for it is fully loaded, and b) whether finding no matches leads to cross filtering out all data or simply ignores the fact there are no matches.

As you type a value in the entry text box, autocompletion may try to append the "top match" to what you're actively typing. You can disable this behavior with the 'Attempt Completion' setting.

With the 'Word Splitting' setting, if you type multiple words in the entry text box, either the whole phrase (as typed) must match (Disabled), all words must match (Intersect), or any word must match (Union). (The example below shows Union enabled.)

The 'Casing' setting can force the case of text entered in the entry text box to upper or lower case, if desired. 'Auto-detect' (the default) examines the source data to determine if it is all upper or lower case to determine what to enforce.

When 'Can Change Case' is enabled (the default), text typed in the entry text box can have its case changed, to match source values. For example below we typed 'oscar c' and the case has been changed to match the source's value of 'Oscar Carson'.

Buttons / Links

Most buttons can be switched out to use links, instead. (The default is to use buttons.)

Text that's shown for button tool-tips or link text can be changed using "message" setting, under Formatting. (Note that virtually all text messages can be overridden in settings.)

Button images can be overridden with your own images of choice. You'll find one setting per button. This setting would be used for the Reset button:

Buttons / links for different features can be selectively shown, using toggle settings. For example, the "Add All Loaded" button is invisible by default, unless it is explicitly enabled here:

Buttons / links available:

• Add Current: Adds value currently in the entry text box to the filter area. How the add is performed is controlled by settings such as 'Retain Partial Match on Add' and 'Expand on Add'.
• Reset: (aka 'Clear''.) Clears any current filters. (Both in the entry text box and filter list.)
• Can Toggle Show Blanks: When enabled, shows a toggle button that switches between including or excluding blank values in the filter.
• Can Toggle List Style: When enabled, shows a toggle button that switches between presenting the filter list as an editable text area or a set of read-only text values.
• Can Toggle Include Exclude: When enabled, shows a toggle button that switches from including the filter items to excluding them.
• Add All: When enabled and there are source values loaded into the visual, shows a button that will take all distinct source values and place them in the filter list. This can be used to identify source values that are 'currently filtered' based on report cross filters.
• Get List: When enabled, a button is shown that, when clicked, brings up the pick list. (As opposed to requiring typing of new values in the entry text box to bring up the pick list.)
• Copy Clipboard: This button, when clicked, copies all values in the filter list to the clipboard.

Using Multiple Filter Terms

Filter terms can be combined using logical "AND" or "OR". (Default is "OR" = Union.) Union means any filter list item match allows the source row to be "seen", where Intersect means all filter list items must match in each source row.

Starting in version 3.2, you can have finer-grained control over filter term combining. Enabling the 'Allow Filter Prefixes' setting lets you prefix filter list items with '+' (OR), '&' (AND), or '-' (NOT). Consider this example:

Here we have enabled "Allow Filter Prefixes", and added "&le" and "&nn" as filter terms. Notice that the cross-filter to the table on the right shows only names that contain both "le" and "nn" in their body. Now let's change the set of filter terms to include "-qu":

Note how the table on the right has changed to no longer include items like "Quinn Cole", which contains "qu" as text. If we instead use "+" as a prefix, this example shows names that contain either "kim" or "chen" but not "bob":

When multiple filter items are allowed, you can control whether what you're actively typing in the entry text box will "filter out" to other visuals on the page, or whether filtering only happens for "committed values" (i.e. are in the filter list text area / pill list).

When 'Selecting Auto-Adds' is enabled (the default), picking an item in the pick list will add it to the filter list. (It will also close the pick list if 'Keep List Open' is false.) If it's disabled, then the picked value will only be placed in the entry text box, offering the user the option to add it to the filter list using Enter or clicking the Add button/link.

When 'Retain Partial Match on Add' is enabled, the pick list can remain open, even when an item is picked. In this case, the item is added to the filter list, and removed from the pick list. This makes it easy to select other values to add which share the same partial match text. In the example below, 'osca' was typed, 'Oscar Chen' was picked, and then 'Oscar Duarte' was picked. The pick list remained open until focus was lost on the entry text box. Notice that the two picked values were added to the filter list.

'Keep List Open' is similar, where the list will remain open after a selection is made, and focus must be lost or a click on the Close pick list link. In this example, we're picking two names in sequence.

When 'Enter Adds Single' is enabled, if the visual has suggested additional, non-typed text and you press 'Enter', the full match will be added to the filter list. Otherwise, only the typed portion will be added. Notice how partial matches in the filter list area, by default, filter all rows that relate to the filter text (using all other settings that control matching behavior).

'Expand on Add', when enabled, changes the nature of the Add button / link, and potentially the Enter key if 'Enter Adds Single' is disabled. Instead of adding the item being entered, the visual identifies all actual matches in the source data and adds them to the filter area. This has the effect of letting you get a full list of matches, from a partial search term. This list can in turn be copied out of the visual, using the clipboard.

When 'Only Allow Matched' is enabled, partial matches are not used for filtering. This generally requires use of the pick list, or pasting in full values.

'Maximum to Filter' controls the maximum number of distinct values in the value source that can be filtered by the visual. The default is 10,000. Increase this at the risk of performance issues.

When 'Filter Select All' is enabled, putting focus in the filter list area (when text), the visual will automatically select all text in that text area. (Makes copying it or clearing it easier.)

Things to Know

Version 6.0 lets one list mix match rules, shows why a suggestion matched, checks a pasted list, searches several columns from one box, and keeps searches one click away. It is an in-place upgrade: reports that use the visual pick it up automatically, and no setting was renamed or removed. Most of it stays out of sight until you switch it on; the few things that do change are listed under What Changes When You Upgrade from 5.x.

• Per-term operators — exact, starts-with, ends-with, fuzzy and exclude terms side by side in one list.
• Match reasons — a small mark says why a sounds-like, close-spelling or whole-word suggestion is in the pick list.
• Paste report — after pasting a list, see which terms matched exactly, inexactly or not at all.
• Unmatched terms — spot selections that match nothing under the report's other filters.
• Also Search — one box that searches several columns and filters the one that matched.
• Quick picks and recent searches — named term sets and each viewer's recent searches, one click away.
• Copy as DAX — take the applied filter straight into a measure.
• Accessibility — screen readers, keyboard-only use, reduced motion and contrast.
• Power Search styling presets — the new search features in one step.

Per-term Operators

Since version 3.2, 'Allow Filter Prefixes' has let a filter term start with '+', '&' or '-' (see Using Multiple Filter Terms). Version 6.0 turns this into a small syntax that gives each term its own rule, so one list can mix them. A term can start with one of '+ & -', then one of '= ^ $ ~', then the text itself:

Write	Meaning	Example
=	Exact value	=Contoso Ltd matches that customer only, not every name that contains it.
^	Starts with	^Contoso matches Contoso Ltd and Contoso Northwest, but not The Contoso Group.
$	Ends with	$Group matches The Contoso Group.
~	Fuzzy, for this term	~kontoso filters by fuzzy matching even while 'List Not Fuzzy' keeps the other terms literal. 'Fuzzy Matching' must be on; with a sounds-like 'Fuzzy Algorithm' such as Double-metaphone it finds Kontoso Limited and the three Contoso customers.
-	Exclude	-The Contoso Group removes it from what the other terms match.
+ &	Union, intersect	Exactly as since version 3.2.
"…"	No operators	"-5 Discount" is searched as ordinary text instead of being read as an exclusion (it finds -5 Discount Co).
Combined	Exclude, with a match rule	-=Contoso Ltd excludes exactly that value; -^Contoso excludes every value that starts with Contoso.

For example, Contoso, -The Contoso Group and =Northwind Traders in one list filter to Contoso Ltd, Contoso Northwest and Northwind Traders: everything containing Contoso, minus The Contoso Group, plus exactly Northwind Traders.

How much of this syntax is read is set by 'Term Operators' (Data and Interaction → Filtering), which supersedes 'Allow Filter Prefixes': a new report no longer needs that switched on.

• Automatic (the default): a report that had 'Allow Filter Prefixes' switched on keeps exactly its older '+ & -' behavior, and every other report reads operators exactly as Data-aware does.
• Data-aware (all operators): a leading operator character is only read as an operator when the whole term is not itself one of your values, so a customer really called '-5 Discount Co' is left alone, as are negative numbers, Excel exports that begin with '=' and file paths that begin with '~'. Wrap a term in double quotes to force it to be taken literally.
• Legacy (+ & - only): exactly the version 3.2 prefixes. Off: no operators are read at all.

For example, with a customer really called '-5 Discount Co', typing -5 Discount Co adds that customer as it is (its pill has no badge), while =Contoso Ltd is read as an exact-value operator, because no customer is called '=Contoso Ltd'.

In the Pills style, a term that uses an operator shows its value with a small badge for the operator — -The Contoso Group appears as 'The Contoso Group' with a '-' badge — and hovering the badge gives its meaning. Pills without an operator look as before, and the Text list styles show every term as typed. The badge is a label, not a control: to change a pill's operator, remove the pill and add the term again with the new operator. The badges are controlled by 'Show Pill Operator Badges' (on by default) and 'Pill Operator Badge Color'. With 'Show Operator Picker' on (off by default), a small button to the left of the entry text box sets the operator for the next value added, typed or picked: each click moves it along — a dot for none, then '=', '^', '$', '~' and '-' — and it returns to none after each add unless 'Operator Picker Is Sticky' is on.

With 'Include Current as Type' on (the default), the text you are still typing already filters the report with the operator the picker shows, exactly as it will once added: with a Contoso pill in the list and the picker on '-', typing The Contoso Group already leaves just Contoso Ltd and Contoso Northwest.

More detail

• The operator picker applies to a value picked from the suggestion list as well as to a typed one. A picked value is always taken as that value, even one that begins with an operator character; one that is itself wrapped in double quotes ("Classic", quotes included) is added quoted once more, so it keeps its quotes.
• Choosing another operator while text is typed re-applies that text with the new operator, and leaving the entry text box without adding the text keeps that filter, operator included.
• An exclusion subtracts from what the other terms include, as it always has: a list holding only -Contoso selects nothing, so under the default 'Load Behavior' the report shows no data. To show everything except some values, use 'Filter Mode' = Exclude (or the include / exclude button).
• Under 'Only Allow Exact', an operator overrides exact matching for its own term: '^', '$' and '~' match as they would under contains matching, and '=' stays a whole-value lookup. '-' then excludes by exact value unless a match operator follows it (-^Contoso), and '+' and '&' have no effect.
• While data is still loading, the visual cannot yet tell whether a term is one of your values, so Data-aware reads the operator.
• The operator picker appears only with 'Support Multiple' on and 'Term Operators' reading the full operator set. With 'Allow CSV Entry', its operator applies to every comma-separated term of the entry.
• 'Show Pill Operator Badges' and 'Show Operator Picker' are under Formatting → Display & Layout; 'Pill Operator Badge Color' is under Colors → Buttons, Links & Pills.

Match Reasons in the Pick List

A pick list row that matched for a reason you cannot see in its text now carries a small mark: '~' when it sounds like what you typed (the Double-metaphone algorithms), '≈' for a close spelling (the Standard algorithm), and '·' when what you typed is a whole word further inside the value (typing 'contoso' marks The Contoso Group). With the Custom Mapping or Lemmatized algorithm a fuzzy row also carries '≈', described as a fuzzy match rather than a close spelling, since those rules are not spelling corrections. Rows that start with what you typed, or contain it inside a word, carry no mark. Hover a mark for its wording — Sounds like what you typed, Close spelling, Fuzzy match or Whole word inside this value — which a screen reader hears after the value. For example, with 'Fuzzy Matching' on and 'Fuzzy Algorithm' set to Double-metaphone, typing 'kontoso' lists Kontoso Limited without a mark, and Contoso Ltd, Contoso Northwest and The Contoso Group each with '~'.

The setting is 'Show Match Reason' (Formatting → Display & Layout): 'Inexact only' (the default, also on upgraded reports), 'All' (the same marks today) or 'Off'. The marks are not available with the 'Native (can overflow the visual)' list placement, where the browser draws the list.

Paste Report

After a viewer pastes several terms into the Text list, pastes a comma-separated list into the entry text box (with 'Allow CSV Entry' on), or uses 'Expand Batch with Mappings', a line under the list now says how the terms matched — for example 2 exact · 1 inexact · 2 unmatched — with a Details link. Exact means the term is one of your values, inexact that it matched values that start with it or contain it (or, with fuzzy matching, sound or are spelled like it), and unmatched that the filter got nothing from it. The filter is applied first, exactly as before: the report only describes it.

Say a viewer pastes five customer names from an email: Contoso Ltd, Northwind Traders, Fabrikam, Contso and Acme Corp. Details lists the unmatched terms first (Contso, a typo, and Acme Corp), then the inexact one (Fabrikam matched Fabrikam Inc and Fabrikam Supply). Its buttons are Copy unmatched, Copy inexact mapping (ready to paste into Excel), Remove unmatched from list, and Close.

The visual itself, rendered in a browser with sample data (not a Power BI Desktop capture).

'Paste Report' (Formatting → Paste Report) offers 'Summary with Details link' (the default), 'Summary and details' (the details open by themselves) and 'Off'.

More detail

• No report appears for a single term, for typed text or for picks from the pick list.
• A term that matched but did not take effect — removed by a '-' term, say, or past 'Maximum to Filter' — says so, and the summary counts it as not applied. With 'Paste Report Lists Exact Terms' on, Details lists the exact terms too.
• 'Paste Report Rows per Group' (25) caps each list in the details; the copy buttons always copy every term. 'Paste Report Auto-hide (seconds)' hides the report after a delay (0, the default, keeps it until it is closed).
• The report follows the filter, not the pick list: with 'List Not Fuzzy' on (the default), a misspelling the pick list would suggest is still unmatched, because the filter does not apply it. Prefix it with '~', or turn 'List Not Fuzzy' off, to let the filter accept it.
• While data is still loading, the summary says (index partial) and leaves out the terms it cannot yet vouch for, then updates when loading completes.
• Nothing about the report is saved in the Power BI report, and it never changes the visual's size: in a small visual the details open over the list. There, after a paste into the filter list, 'Summary and details' opens with just its summary line, so the details do not cover the list you are typing in; Details opens the rest.
• The auto-hide countdown pauses while the pointer is over the report or keyboard focus is inside it. Remove unmatched checks each term again against the data as it is at that moment, and never removes a term it cannot vouch for.

Unmatched Terms

A term in the list can match nothing under the report's other filters — a customer that only has sales in Germany, while a Region slicer is set to USA. The term stays in the list but contributes nothing, and until now nothing said so. Three settings, all off by default, can point it out:

• 'Unmatched Term Style' (Formatting → Display & Layout), for the Pills style: 'Dashed border' draws the pill's border dashed, in the pill's text color, and 'Dashed border + strike-through' also strikes out its text. The text itself is never faded: it keeps the color every other pill has, so a marked pill stays just as readable. While any pill is marked, a Remove unmatched button under the list removes them all.
• 'Show Unmatched Terms Line', for the Text list styles: a line under the list such as 2 terms match nothing under the current filters., with the same Remove unmatched button.
• 'Remove Unmatched When Adding' (Data and Interaction → Multiple Values): a term that matches nothing at the moment it is added is dropped straight away, and a message says how many were dropped.

For example, with 'Unmatched Term Style' set to Dashed border and Contoso Ltd and Fabrikam Supply in the list, where Fabrikam Supply only has rows in Germany, picking USA in a Region slicer gives the Fabrikam Supply pill a dashed border and shows Remove unmatched under the list. Clear the Region slicer and the pill is back to normal. The mark is information only: nothing is ever removed because another slicer changed, since a customer unmatched under one region can match under the next.

In the Text list styles the same situation shows as one line under the list, with the same Remove unmatched button. The line and the button update when terms are added, pasted or removed and when you leave the list, never while you are still typing in it.

More detail

• With 'Allow Tooltips' on, hovering a marked pill says No rows match under the current filters, and a screen reader always hears it. The Remove unmatched button is not shown when 'Filter Text Read Only' is on, and the line's wording can be changed under Message Text Overrides.
• A term is judged exactly the way the filter judges it, so every matching setting applies. Excluding ('-') terms are never marked, and nothing is marked while 'Filter Mode' (or the include / exclude button) is Exclude.
• While data is still loading, after a cancelled load, when the visual has reached its capacity, or when the other filters leave the visual no data at all, it cannot know that a term matches nothing, so nothing is marked or dropped.
• 'Remove Unmatched When Adding' checks only the terms being added; a term already in the list is never removed.

Also Search: One Box, Several Columns

Place up to four more columns in the new Also Search (optional) field well to search them from the same box as the Value Source — for example Customer as the Value Source, with Product and Order ID in Also Search. The pick list then groups the suggestions under a header for each column, and picking one filters the column it came from. Typing 'wid' lists Wide World Importers under Customer and Widget under Product. Pick Widget and add Contoso Ltd, and the page narrows to Contoso Ltd's Widget orders: filters on different columns combine with AND.

Here the Value Source has two columns, Customer and Region, so its header reads Customer / Region.

One search bar for the whole page. Put the columns people actually look things up by — customer name, product, city, order number — in Also Search, make the visual wide and short, and it becomes a global search bar: the viewer types a few letters without deciding which field to use, and the list sorts out which column each match came from. Pick a customer, a product or a city and the whole page filters on that column.

One box, three columns: the matches are grouped by the column they came from. (The visual rendered in a browser with sample data.)

Picks become pills that name their column, with quick picks one click away. (The visual rendered in a browser with sample data.)

To aim a typed term at one column, start it with the column's name and a colon: Product:wid searches Product only, and operators go after the colon (Product:-Widget). The column name can be typed in upper or lower case, in the entry text box or in the Text list, and it is recognized whatever 'Term Operators' says.

Product:wid searches the Product column only. (The visual rendered in a browser with sample data.)

The column headers are set in bold small capitals on a very light gray band, with a thin rule between groups, so they cannot be mistaken for values; they are not rows you can pick, and the arrow keys skip them. To give the band your own color, set 'Group Header Background Color' (Colors → Entry & List Areas); left unset, the band is the list's background moved a small step toward its text, so it stays a light band on a colored or dark list too. Either way the header text is darkened (or lightened) as far as needed to stay readable against the band (4.5:1, the WCAG AA level for text), and a high-contrast theme always wins. When only the Value Source has matches, no header is shown and the list looks as it always has.

Nothing changes until a column is placed in the Also Search well. Its settings are under Data and Interaction → Also Search.

More detail

• 'Group Headers' (on; turned off, a suggestion's column name follows it in muted text instead), 'Matches Shown per Column' (5, so one column cannot push the others out of the list; 'Matches Shown' still limits the whole list), 'Value Source First' (on; turned off, the column with the closest match to what you typed is listed first) and 'Show Column on Pills' (on; hover the column name on a pill to see which column it filters).
• A screen reader hears a suggestion's column after its value; the headers themselves are not read as options. With the 'Native (can overflow the visual)' list placement the browser draws the list, so there are no headers: each Also Search suggestion carries its column name instead.
• An Also Search column can come from another table than the Value Source, such as a related Products table's Category, and a number column works too. With several columns in the Value Source, the Value Source keeps its combined filter and each Also Search column adds its own; Exclude is still refused for such a Value Source (the visual says so), while the Also Search columns are still excluded.
• A column with more values than the visual can hold is searched only in part, and a message names it.
• A term typed and added without picking filters the Value Source, as before, unless it is exactly a suggestion shown under another column's header. The Text list shows the full term (Product:Widget), and renaming a column in the field well keeps its terms on it.
• Each column is matched against its own values with every matching setting, 'Filter Mode' and the include / exclude button apply to all of them, and 'Maximum to Filter' applies to each column's filter.
• Because the columns combine with AND, terms aimed at one column that match nothing leave the report with no data under a 'No Match, No Data' 'Load Behavior'; under 'No Match, All Data' that column is simply not filtered.
• Secondary Values, Sort Order, Additional Tool Tips, Filter Default and blanks stay with the Value Source, 'Add All Loaded' adds Value Source values only, and date or time columns placed in Also Search are not searched.
• The visual receives one row per combination of Value Source and Also Search values, so loading takes longer on large tables, a Secondary Value that is a measure is combined by 'Secondary Aggregation' (right for a sum, not for an average or a distinct count), and a Sort Order measure and Additional Tool Tips are taken from the first combination loaded.
• A report saved with Also Search terms and opened with a version before 6.0 reads a term such as Product:Widget as a Value Source value.

Quick Picks and Recent Searches

Quick picks are named sets of terms that you, the report author, define. They appear as a row of chips between the entry text box and the filter list, and one click applies a set. Switch on 'Show Quick Picks' (Data and Interaction → Quick Picks; it requires 'Support Multiple') and write one quick pick per line in 'Quick Picks', as Name=term1|term2:

Key accounts=Contoso Ltd|Northwind Traders|Blue Yonder Airlines
Contoso family=Contoso
Contoso, not the group=Contoso|-The Contoso Group

Each term is written exactly as you would type it, so the operators above and Also Search column names (Widgets=Product:Widget) work here too. 'Quick Pick Action' decides whether a click replaces the filter list ('Replace', the default) or adds to it ('Add'), and with 'Show Save Quick Pick' on, a '+' at the end of the chip row saves the current list as a new quick pick.

The '+' button opens a small name box inside the visual: type a name and press Enter (or Save) to save the current filter list, or Escape (or Cancel) to close it. A quick pick with exactly that name gets the new terms; otherwise a line is added at the end of 'Quick Picks', and every other line is kept as written. A saved line starts with '=', as in =Two picks=Northwind Traders|Proseware: its chip puts back exactly the list that was saved, the way a recent search does, instead of adding the terms one by one (with 'Expand on Add' on, adding them again would bring back values that had been removed). Delete the leading '=' to have its terms added like typed terms instead.

More detail

• Terms are separated by '|' because commas appear in values too often. To put a '|' inside a term, wrap the term in double quotes; the quotes are an escape whenever the full operator set is read ('Term Operators' Data-aware, or Automatic in a report that never switched on 'Allow Filter Prefixes').
• Each term goes in exactly as if it were typed and added, so 'Expand on Add', 'Only Allow Exact', 'Casing' and the other add settings apply, and the filter is applied once, after the last term. 'Replace' first clears the filter list and the entry text box.
• Only the first '=' ends the name, so a term may itself begin with '=' (Exact==Contoso Ltd), and a name that contains '=' can be wrapped in double quotes. Blank lines, and lines without a name or without any terms, are ignored.
• A quick pick saved with the '+' button is saved with the report in Power BI Desktop. A viewer in the Power BI Service cannot change the report, so there the new chip lasts only for that viewer's session. A term that cannot be written into a line — one containing a line break or an unquoted '|' — is left out of the saved quick pick, and a message names it.
• The chip row wraps onto a second line and then scrolls; in the Pills style the pills give up its height. 'List Spacing' applies above the row as well as below it.
• The chips use the filter list's font, and their colors are 'Quick Pick Background Color' and 'Quick Pick Text Color' (Colors → Buttons, Links & Pills); left unset, they follow the report theme when 'Follow Report Theme' is on.

After the quick picks come each viewer's own recent searches, in italics and marked with ⟲ ('Recent Searches Shown', 5 by default; 0 turns them off). Clicking one puts that exact list back, replacing the filter list or adding to it as 'Quick Pick Action' says. A recent chip shows its terms cut at 40 characters, with all of them in its tooltip. Recents are kept in the viewer's own browser, never in the report.

More detail

• A search is remembered when it is committed with at least one term - except one committed while the column is still loading under a 'Fully Loaded' Load Behavior, and a very large one (more than about 4,000 characters of terms, typically a big paste or an expanded list). Refining a search updates its entry rather than adding one per step, and a search identical to one of your quick picks is not kept as a recent, so clicking a quick pick never pushes a viewer's own searches out.
• A search is committed by Enter, Add, a pick from the suggestion list, a paste into the Text list, leaving the Text list after editing it, removing a pill or clicking a chip; Reset starts a new search.
• Recents are kept per table and column, so reports that filter the same column share them. A recent search that used an Also Search column (a Product:Widget term) appears only in visuals that search that column too. Power BI clears a visual's stored data 29 days after it was last written.

Recent searches and your tenant. Recents use Power BI's local-storage permission for custom visuals, which the visual requests as optional and uses only while 'Show Quick Picks' is on and 'Recent Searches Shown' is above 0. A Power BI administrator can turn local storage off for custom visuals, and some hosts or browsers do not allow it: then no recent chips appear, your quick picks work as usual, and viewers see no warning or message about it.

Copy as DAX

'Show Copy as DAX' (Formatting → Display & Layout, off by default) adds a DAX button to the entry row. It copies the filter the report is actually under — not what is typed in the box — as a DAX predicate you can paste into a measure, a calculated column or a calculated table. With Contoso in the list (matching three customers), it copies 'Shipments'[Customer] IN { "Contoso Ltd", "Contoso Northwest", "The Contoso Group" }, so CALCULATE ( COUNTROWS ( Shipments ), … ) with that pasted in counts those customers' rows.

More detail

• Terms that matched nothing are not included. An exclude filter is wrapped in NOT ( … ), and several Value Source columns give the tuple form.
• With Also Search columns, every filtered column's predicate is copied, joined with &&. That works in FILTER, in measures and in calculated columns, and in CALCULATE when the columns come from one table (otherwise split it into one argument per column).
• While the visual is blanking the report ('Must Select' with nothing entered, or a 'No Match, No Data' 'Load Behavior' with nothing matched) it copies FALSE(), or NOT ( FALSE() ) in Exclude mode.
• When nothing is filtered, nothing is copied and the message says so. The confirmation follows 'Copied Message Duration' and is read out by screen readers; a copy the browser refuses shows 'Failed to copy to clipboard.' With 'Show Links' on, the button is a 'Copy filter as DAX' link.
• The button sits to the left of the entry text box and follows 'Button Background Color' and 'General Style' like the other entry-row buttons. Tab reaches it right before the entry text box (after the operator picker).

Accessibility

Version 6.0 brings the visual's screen reader and keyboard support up to the standard patterns, with nothing to switch on:

• The entry text box, the pick list and the filter pills describe themselves to screen readers with the standard combo box, list box and list roles, and the highlighted suggestion is announced, with its position, as you arrow through the list. Match counts and messages are read out as they change (the count is not read while data is still loading, nor again on each arrow key).
• The active suggestion — the one 'Attempt Completion' has filled into the entry text box, or the one the arrow keys moved to — is outlined in the list, so the row Enter adds can be seen as well as heard. The keys work as before.
• Each pill's remove button is now a real button: Tab moves through the pills, and Enter, Space, Delete or Backspace removes the focused one, after which focus moves on to the next pill (or the previous one when the last pill is removed, or the entry text box when none is left). Before 6.0 it responded only to the mouse. Escape closes the suggestion list without leaving the entry text box.
• The entry text box and the Text list are named for screen readers (the box after the visual's in-frame title, or as 'Search' and the Value Source's name), and a pill's operator badge or a suggestion's match-reason mark is read out as words after the value — 'Fabrikam Supply, Excluded' — rather than as a symbol. Pressed from the keyboard, the operator picker keeps focus and announces the operator it moved to.
• The operator picker and the DAX button sit left of the entry text box, and Tab reaches them right before it; every control that was there before 6.0 keeps its place in the Tab order.
• Quick-pick chips and the paste report's controls are buttons too: Tab reaches the chips after the buttons of the entry row, Enter or Space applies one and keyboard focus stays on the chip, and a recent-search chip is read as 'Recent search:' followed by all of its terms. The paste report's headings can be reached as headings, and an unmatched pill is marked by its dashed border and a sentence for screen readers, never by color alone.
• The loading spinner stops turning when the operating system is set to reduce motion.
• Text on filter pills and quick-pick chips meets the WCAG AA contrast level (4.5:1) in the default look and when following the report theme. Colors you set yourself are used as you set them, so choose ones that reach 4.5:1 as well.

Power Search Styling Presets

Two new styling presets switch the 6.0 search features on in one step: under Presets → Styling, pick 'Text, Power Search' or 'Pills, Power Search'. Each sets its List Style (Text or Pills) and turns on 'Show Operator Picker', 'Show Pill Operator Badges', 'Paste Report' = Summary and details, 'Unmatched Term Style' = Dashed border, 'Show Unmatched Terms Line', 'Show Match Reason' = Inexact only, 'Show Quick Picks' and 'Keep List Open' (the suggestion list stays up even when typing narrows it to the one value the box has filled in), with the spacing and title typography of 'Text, Native Compact'.

Illustration — not a screenshot

More detail

• The spacing is 'Container Padding' 5, 'Frame Title Padding' 0, 'Frame Title Font Size' 12 and 'Frame Title Font' Segoe UI Semibold. Like every styling preset, it also turns on 'Follow Report Theme' and sets 'General Style' to Native Slicer.
• A styling preset changes how the visual looks and what it shows, never what it filters: 'Support Multiple', 'Term Operators' and the matching settings are left alone. So the operator picker and the chip row appear only where 'Support Multiple' is on (the picker also needs 'Term Operators' reading the full operator set), and chips only once there are quick picks or recent searches to show.
• A preset is applied at the moment you pick it, not when the report opens. Once a Power Search preset has been applied, the report stores 'Styling' as None, so versions before 6.0 can still open it: the Styling box keeps showing the preset while the report stays open, reads None after it is reopened, and the settings the preset switched on stay as they are.
• 'Keep List Open' is not new, and it stays off by default. With 'Attempt Completion' on and 'Keep List Open' off, the suggestion list closes once typing narrows it to the single value the entry text box has filled in; the Power Search presets turn 'Keep List Open' on so the list stays up while you type and, with 'Support Multiple' on, after each pick.
• For an exact width match with the built-in slicers, also set the visual's own Padding to 0 under General → Properties → Padding, which a preset cannot change.

What Changes When You Upgrade from 5.x