tl;dr
A group is a new dimension you define. The clusters inside it are the values it can take.
Each cluster has rules. Anything that matches no rule appears as
n/a.Clusters are checked in order — the first match wins.
Rules are set per report area, and each area offers different fields.
New rules take effect from your next data refresh. Changes to existing groups apply retroactively.
What is a Custom Group?
Klar reports let you slice your numbers by dimensions like channel, campaign or product type. Those come out of the box. A Custom Group is a dimension you define, using data Klar already has.
Say you sell across three business lines, but nothing in your shop marks which order belongs to which. You know the pattern: products tagged bundle-kit are Starter Kits, product type Refill means Subscriptions, and the rest is your Core Range. Write that down once, and "Business Line" becomes a dimension you can group by, filter by and track over time.
Business Line — the group
Name the group like a column heading ("Business Line") and the clusters like values ("Starter Kits", not "Products tagged bundle-kit"). Cluster names show up in your reports.
Where can I use them?
Rules are written per report area, and the fields on offer differ. That's because the areas count different things: a row of ad spend is a day of budget on a campaign, with no product or customer attached. A row of sales is a product inside an order.
But you can also combine them. Set up a cluster with sales and attribution data based on the product type and marketing spend based on the campaign name, and you can analyze ROAS etc for those clusters.
Report area | Each row is | Fields you can build rules on |
Sales | a product in an order | Order date, sales channel, channel, campaign, ad group, ad, country, discount code, order tags, customer tags, vendor, product title, product type, product tags, variant title, SKU |
Marketing | one day on one campaign, ad group or ad | Date, channel, campaign, ad group, ad, country, data source |
Attribution — orders | a touchpoint on the way to an order | Order date, channel, campaign, ad group, ad, country, sales channel, discount code, order tags, customer tags, product title, product type, variant title, SKU |
Attribution — visits | a visit, which may never become an order | Channel, campaign, ad group, ad, product title, product type, variant title, SKU |
Influencer reporting uses the same fields as Marketing. A group only has something to say in an area you gave it rules for — set it up for Sales only, then group a Marketing report by it, and everything lands in n/a.
How do I set one up?
Go to Settings → Store Configurator → Groups and click Add Group.
Create the group and give it a clear name — this is what your team will look for in the dimension picker months from now.
Add your clusters, ordered most specific first. See the warning below about why order matters.
Write each cluster's rules. A rule is a field, an operator (e.g. contains, is, starts with) and a value. Combine several with AND when all must be true, or OR when any will do.
Repeat for every area you need. Rules don't travel between report areas — usually you'll write different fields for each.
Save and wait for the next data refresh, then check your work: open a recent report grouped by the new group, and look at how much sits in
n/a. If it's more than you'd expect, go back and check your rules — a field value that doesn't match exactly, or a broad cluster sitting too high, is usually the cause.
What should I watch out for?
Cluster order decides the result
Clusters are checked from top to bottom and the first match wins — nothing after it is considered. Keep your most specific clusters high and your broadest ones low. A wide "Everything else" sitting first will swallow the whole group.
One cluster per row, per group. A row can only carry one value for a group — that's what keeps the reporting safe, since an order in two clusters would have its revenue counted twice. If you need labels that genuinely overlap, use two groups: "Business Line" and "Promotion Type" can both label the same order.
Changes apply retroactively. When you edit an existing group — renaming a cluster, adding a rule, changing the order — Klar recalculates your historical data at the next refresh. You don't need to re-import anything.