Most of the Group by options in a report do the obvious thing — group by campaign, by ad set, by format, and Atria collapses everything with the same value into one row. Custom groups are for the comparison that doesn't already exist as a column: your two angles, your three funnel stages, evergreen versus seasonal, the ads you've made in-house against the ads the agency made. You define what each bucket contains, and Atria reports on the buckets.
It's the option we reach for most, and the one people find last.
When to use a custom group instead of a normal group-by
Quick rule of thumb, because this is really the whole decision:
If the thing you want to compare already exists as a value on your ads — a campaign name, a format, a custom tag, a naming convention position — use the ordinary Group by option for it. Less work, and it keeps updating itself as you launch. The full list is in Group-by options in Analytics reports.
If the thing you want to compare is a judgement call that only lives in your head — "our three testing angles", "the ads that mention price" — that's a custom group. You're defining the categories by describing them in filters.
The second one is where the interesting analysis usually hides, because the comparison your team argues about in the Monday meeting is almost never a field Meta hands over.
Building one
Open or create a custom report.
Click Group by and choose Define custom groups from the dropdown.
3. Define each group by setting filters — the conditions an ad has to meet to fall into that bucket.
4. Add as many groups as the comparison needs, then click the orange Save in the report header. Reports don't save themselves, and a set of custom groups is a painful thing to lose to a closed tab.
From then on the report treats your groups as rows: spend, ROAS, CPA and every other metric you've added, per group, side by side.
Advanced filters inside your group definitions
This is the meaningful change since this article was last written, and it widens what a custom group can be.
A filter used to match all of your conditions or any of them, never a mix. That made some groups impossible to express — "the Summer Sale or Back to School campaigns, but only the video and carousel ads inside them" needed two groups and a calculator.
Now you can nest conditions into groups, each with its own And / Or logic, and those condition groups combine with And so each one narrows the result further. So the bucket above becomes one definition: (Summer Sale Or Back to School) And (Video Or Carousel). Condition groups nest one level deep, and anything you'd already built keeps working exactly as before.
You'll see the same controls on the Dimension filter and Metric filter in any report
+ Add filter rule for a single condition, and Add filter group — the button with the small panel icon — for a nested one, both inside modals headed Dimension filter records and Metric filter records. A nested group lands as an indented, orange-bordered box with its own Where row and its own + Add filter rule, joined to the rules above it by an And dropdown, so you can see the shape of your logic rather than having to hold it in your head.
A worked example: comparing creators
The request we get most is some version of "can I build a report with only certain creators' ads?" — and if you want them compared rather than merely isolated, custom groups are the tool.
Make sure each creator's ads are identifiable — a custom tag per creator is the reliable way, or a naming convention position if your ad names already carry it.
In a custom report, Group by → Define custom groups.
Group one: filter to creator A. Group two: creator B. Group three: creator C.
Narrow all three the same way if you need a fair fight — adding And (Video) to each group keeps you from comparing one creator's videos against another's statics.
That last step is the bit people skip, and it's the bit that makes the answer trustworthy. A custom group is only as honest as the filters you gave it, so make the groups differ in exactly one dimension — the one you're testing — and hold everything else steady.
Or just ask Raya
If describing it is quicker than clicking it, say so in plain English — "compare ads in the Summer Sale or Back to School campaigns that are video or carousel, build an Analytics report" — and Raya will build the grouped filter and save it into a report for you. Handy when you're exploring and don't yet know how many groups you want.
A few things worth knowing
Your groups travel with the report. Duplicating a report carries the group-by settings, filters, metrics, date range and attribution settings across — so build the comparison once and clone it per account. See Duplicate analytics reports.
Groups are only as current as their definitions. A custom group is a set of conditions, not a fixed list of ads, so new ads that match will join the group on their own. If you defined a group by naming specific campaigns, though, next quarter's campaigns won't be in it — worth a diary note.
Overlapping definitions. If you've got a comparison in your head and you're not sure how to express it as filters, send it over in the words you'd use to say it out loud — that's usually the fastest way for us to turn it into a set of groups for you.