What a grouping is
Most marketing dimensions have long tails. A traffic_source column might have hundreds of distinct values: google, google.com, www.google.com, facebook.com, m.facebook.com, t.co, l.facebook.com, instagram.com, and on and on. None of those rows are wrong, they're just too granular for a high-level chart.
A grouping lets you collapse those values into named buckets. Roll all the Google-ish entries into Organic Search, all the Meta-owned domains into Paid Social, t.co into Twitter, and the remaining strays into Other. Charts now show four bars instead of four hundred.
Create a grouping
Go to Lens > Groupings and click Add Grouping. Pick the base table and the dimension you want to group. The form lets you build named groups one at a time.
For each group, add one or more rules. Each rule matches dimension values by equals, contains, starts with, or in (multi-value). Values that match any rule fall into that group. Group order matters: the first rule that matches wins, so put the most specific rules at the top of the list.
Set a default for values that don't match any group. Common choices: Other, Unknown, or keep the original value. Save the grouping.
Use a grouping on a tile
On any board tile built from the same base table, swap the raw dimension for the grouping. The dimension dropdown lists the grouping alongside the table's normal columns. The tile re-renders bucketed instead of raw, while the underlying data stays untouched.
Build the grouping once, use it on every board. When the rules need updating (a new traffic source, a vendor merger, a renamed UTM convention), change the grouping definition in one place and every chart updates on next render.
Combining groupings with segments
Groupings change how rows are bucketed. Segments change which rows are included. They're independent and stack cleanly: a tile can apply Channel (a grouping) and US Mobile (a segment) at the same time. Use groupings to shape the chart axis, segments to narrow the data set.
Troubleshooting
Some values still show up un-bucketed
Either the rules don't actually match those values (check capitalization on equals, character set on contains), or the values do match but a higher-priority group is catching them first. Reorder rules with the most specific at the top.
Default bucket is too large
Look at what's in it. If you see a recognizable pattern, add another group rule to capture it. If it's truly long-tail noise, that's what the default is for. Renaming it Other or Long Tail makes that explicit.